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)
156 ops->exit_batch(net_exit_list);
159 static void ops_free_list(const struct pernet_operations *ops,
160 struct list_head *net_exit_list)
163 if (ops->size && ops->id) {
164 list_for_each_entry(net, net_exit_list, exit_list)
169 /* should be called with nsid_lock held */
170 static int alloc_netid(struct net *net, struct net *peer, int reqid)
172 int min = 0, max = 0;
179 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
182 /* This function is used by idr_for_each(). If net is equal to peer, the
183 * function returns the id so that idr_for_each() stops. Because we cannot
184 * returns the id 0 (idr_for_each() will not stop), we return the magic value
185 * NET_ID_ZERO (-1) for it.
187 #define NET_ID_ZERO -1
188 static int net_eq_idr(int id, void *net, void *peer)
190 if (net_eq(net, peer))
191 return id ? : NET_ID_ZERO;
195 /* Must be called from RCU-critical section or with nsid_lock held. If
196 * a new id is assigned, the bool alloc is set to true, thus the
197 * caller knows that the new id must be notified via rtnl.
199 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
201 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
202 bool alloc_it = *alloc;
206 /* Magic value for id 0. */
207 if (id == NET_ID_ZERO)
213 id = alloc_netid(net, peer, -1);
215 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
218 return NETNSA_NSID_NOT_ASSIGNED;
221 /* Must be called from RCU-critical section or with nsid_lock held */
222 static int __peernet2id(struct net *net, struct net *peer)
226 return __peernet2id_alloc(net, peer, &no);
229 static void rtnl_net_notifyid(struct net *net, int cmd, int id, gfp_t gfp);
230 /* This function returns the id of a peer netns. If no id is assigned, one will
231 * be allocated and returned.
233 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
235 bool alloc = false, alive = false;
238 if (refcount_read(&net->count) == 0)
239 return NETNSA_NSID_NOT_ASSIGNED;
240 spin_lock_bh(&net->nsid_lock);
242 * When peer is obtained from RCU lists, we may race with
243 * its cleanup. Check whether it's alive, and this guarantees
244 * we never hash a peer back to net->netns_ids, after it has
245 * just been idr_remove()'d from there in cleanup_net().
247 if (maybe_get_net(peer))
248 alive = alloc = true;
249 id = __peernet2id_alloc(net, peer, &alloc);
250 spin_unlock_bh(&net->nsid_lock);
251 if (alloc && id >= 0)
252 rtnl_net_notifyid(net, RTM_NEWNSID, id, gfp);
257 EXPORT_SYMBOL_GPL(peernet2id_alloc);
259 /* This function returns, if assigned, the id of a peer netns. */
260 int peernet2id(struct net *net, struct net *peer)
265 id = __peernet2id(net, peer);
270 EXPORT_SYMBOL(peernet2id);
272 /* This function returns true is the peer netns has an id assigned into the
275 bool peernet_has_id(struct net *net, struct net *peer)
277 return peernet2id(net, peer) >= 0;
280 struct net *get_net_ns_by_id(struct net *net, int id)
288 peer = idr_find(&net->netns_ids, id);
290 peer = maybe_get_net(peer);
297 * setup_net runs the initializers for the network namespace object.
299 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
301 /* Must be called with pernet_ops_rwsem held */
302 const struct pernet_operations *ops, *saved_ops;
304 LIST_HEAD(net_exit_list);
306 refcount_set(&net->count, 1);
307 refcount_set(&net->passive, 1);
308 get_random_bytes(&net->hash_mix, sizeof(u32));
309 net->dev_base_seq = 1;
310 net->user_ns = user_ns;
311 idr_init(&net->netns_ids);
312 spin_lock_init(&net->nsid_lock);
313 mutex_init(&net->ipv4.ra_mutex);
315 list_for_each_entry(ops, &pernet_list, list) {
316 error = ops_init(ops, net);
320 down_write(&net_rwsem);
321 list_add_tail_rcu(&net->list, &net_namespace_list);
322 up_write(&net_rwsem);
327 /* Walk through the list backwards calling the exit functions
328 * for the pernet modules whose init functions did not fail.
330 list_add(&net->exit_list, &net_exit_list);
332 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
333 ops_exit_list(ops, &net_exit_list);
336 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
337 ops_free_list(ops, &net_exit_list);
343 static int __net_init net_defaults_init_net(struct net *net)
345 net->core.sysctl_somaxconn = SOMAXCONN;
349 static struct pernet_operations net_defaults_ops = {
350 .init = net_defaults_init_net,
353 static __init int net_defaults_init(void)
355 if (register_pernet_subsys(&net_defaults_ops))
356 panic("Cannot initialize net default settings");
361 core_initcall(net_defaults_init);
364 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
366 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
369 static void dec_net_namespaces(struct ucounts *ucounts)
371 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
374 static struct kmem_cache *net_cachep __ro_after_init;
375 static struct workqueue_struct *netns_wq;
377 static struct net *net_alloc(void)
379 struct net *net = NULL;
380 struct net_generic *ng;
382 ng = net_alloc_generic();
386 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
390 rcu_assign_pointer(net->gen, ng);
399 static void net_free(struct net *net)
401 kfree(rcu_access_pointer(net->gen));
402 kmem_cache_free(net_cachep, net);
405 void net_drop_ns(void *p)
408 if (ns && refcount_dec_and_test(&ns->passive))
412 struct net *copy_net_ns(unsigned long flags,
413 struct user_namespace *user_ns, struct net *old_net)
415 struct ucounts *ucounts;
419 if (!(flags & CLONE_NEWNET))
420 return get_net(old_net);
422 ucounts = inc_net_namespaces(user_ns);
424 return ERR_PTR(-ENOSPC);
431 refcount_set(&net->passive, 1);
432 net->ucounts = ucounts;
433 get_user_ns(user_ns);
435 rv = down_read_killable(&pernet_ops_rwsem);
439 rv = setup_net(net, user_ns);
441 up_read(&pernet_ops_rwsem);
445 put_user_ns(user_ns);
448 dec_net_namespaces(ucounts);
455 * net_ns_get_ownership - get sysfs ownership data for @net
456 * @net: network namespace in question (can be NULL)
457 * @uid: kernel user ID for sysfs objects
458 * @gid: kernel group ID for sysfs objects
460 * Returns the uid/gid pair of root in the user namespace associated with the
461 * given network namespace.
463 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
466 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
467 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
469 if (uid_valid(ns_root_uid))
472 if (gid_valid(ns_root_gid))
475 *uid = GLOBAL_ROOT_UID;
476 *gid = GLOBAL_ROOT_GID;
479 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
481 static void unhash_nsid(struct net *net, struct net *last)
484 /* This function is only called from cleanup_net() work,
485 * and this work is the only process, that may delete
486 * a net from net_namespace_list. So, when the below
487 * is executing, the list may only grow. Thus, we do not
488 * use for_each_net_rcu() or net_rwsem.
493 spin_lock_bh(&tmp->nsid_lock);
494 id = __peernet2id(tmp, net);
496 idr_remove(&tmp->netns_ids, id);
497 spin_unlock_bh(&tmp->nsid_lock);
499 rtnl_net_notifyid(tmp, RTM_DELNSID, id,
504 spin_lock_bh(&net->nsid_lock);
505 idr_destroy(&net->netns_ids);
506 spin_unlock_bh(&net->nsid_lock);
509 static LLIST_HEAD(cleanup_list);
511 static void cleanup_net(struct work_struct *work)
513 const struct pernet_operations *ops;
514 struct net *net, *tmp, *last;
515 struct llist_node *net_kill_list;
516 LIST_HEAD(net_exit_list);
518 /* Atomically snapshot the list of namespaces to cleanup */
519 net_kill_list = llist_del_all(&cleanup_list);
521 down_read(&pernet_ops_rwsem);
523 /* Don't let anyone else find us. */
524 down_write(&net_rwsem);
525 llist_for_each_entry(net, net_kill_list, cleanup_list)
526 list_del_rcu(&net->list);
527 /* Cache last net. After we unlock rtnl, no one new net
528 * added to net_namespace_list can assign nsid pointer
529 * to a net from net_kill_list (see peernet2id_alloc()).
530 * So, we skip them in unhash_nsid().
532 * Note, that unhash_nsid() does not delete nsid links
533 * between net_kill_list's nets, as they've already
534 * deleted from net_namespace_list. But, this would be
535 * useless anyway, as netns_ids are destroyed there.
537 last = list_last_entry(&net_namespace_list, struct net, list);
538 up_write(&net_rwsem);
540 llist_for_each_entry(net, net_kill_list, cleanup_list) {
541 unhash_nsid(net, last);
542 list_add_tail(&net->exit_list, &net_exit_list);
546 * Another CPU might be rcu-iterating the list, wait for it.
547 * This needs to be before calling the exit() notifiers, so
548 * the rcu_barrier() below isn't sufficient alone.
552 /* Run all of the network namespace exit methods */
553 list_for_each_entry_reverse(ops, &pernet_list, list)
554 ops_exit_list(ops, &net_exit_list);
556 /* Free the net generic variables */
557 list_for_each_entry_reverse(ops, &pernet_list, list)
558 ops_free_list(ops, &net_exit_list);
560 up_read(&pernet_ops_rwsem);
562 /* Ensure there are no outstanding rcu callbacks using this
567 /* Finally it is safe to free my network namespace structure */
568 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
569 list_del_init(&net->exit_list);
570 dec_net_namespaces(net->ucounts);
571 put_user_ns(net->user_ns);
577 * net_ns_barrier - wait until concurrent net_cleanup_work is done
579 * cleanup_net runs from work queue and will first remove namespaces
580 * from the global list, then run net exit functions.
582 * Call this in module exit path to make sure that all netns
583 * ->exit ops have been invoked before the function is removed.
585 void net_ns_barrier(void)
587 down_write(&pernet_ops_rwsem);
588 up_write(&pernet_ops_rwsem);
590 EXPORT_SYMBOL(net_ns_barrier);
592 static DECLARE_WORK(net_cleanup_work, cleanup_net);
594 void __put_net(struct net *net)
596 /* Cleanup the network namespace in process context */
597 if (llist_add(&net->cleanup_list, &cleanup_list))
598 queue_work(netns_wq, &net_cleanup_work);
600 EXPORT_SYMBOL_GPL(__put_net);
603 * get_net_ns - increment the refcount of the network namespace
604 * @ns: common namespace (net)
606 * Returns the net's common namespace.
608 struct ns_common *get_net_ns(struct ns_common *ns)
610 return &get_net(container_of(ns, struct net, ns))->ns;
612 EXPORT_SYMBOL_GPL(get_net_ns);
614 struct net *get_net_ns_by_fd(int fd)
617 struct ns_common *ns;
620 file = proc_ns_fget(fd);
622 return ERR_CAST(file);
624 ns = get_proc_ns(file_inode(file));
625 if (ns->ops == &netns_operations)
626 net = get_net(container_of(ns, struct net, ns));
628 net = ERR_PTR(-EINVAL);
635 struct net *get_net_ns_by_fd(int fd)
637 return ERR_PTR(-EINVAL);
640 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
642 struct net *get_net_ns_by_pid(pid_t pid)
644 struct task_struct *tsk;
647 /* Lookup the network namespace */
648 net = ERR_PTR(-ESRCH);
650 tsk = find_task_by_vpid(pid);
652 struct nsproxy *nsproxy;
654 nsproxy = tsk->nsproxy;
656 net = get_net(nsproxy->net_ns);
662 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
664 static __net_init int net_ns_net_init(struct net *net)
667 net->ns.ops = &netns_operations;
669 return ns_alloc_inum(&net->ns);
672 static __net_exit void net_ns_net_exit(struct net *net)
674 ns_free_inum(&net->ns);
677 static struct pernet_operations __net_initdata net_ns_ops = {
678 .init = net_ns_net_init,
679 .exit = net_ns_net_exit,
682 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
683 [NETNSA_NONE] = { .type = NLA_UNSPEC },
684 [NETNSA_NSID] = { .type = NLA_S32 },
685 [NETNSA_PID] = { .type = NLA_U32 },
686 [NETNSA_FD] = { .type = NLA_U32 },
689 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
690 struct netlink_ext_ack *extack)
692 struct net *net = sock_net(skb->sk);
693 struct nlattr *tb[NETNSA_MAX + 1];
698 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
699 rtnl_net_policy, extack);
702 if (!tb[NETNSA_NSID]) {
703 NL_SET_ERR_MSG(extack, "nsid is missing");
706 nsid = nla_get_s32(tb[NETNSA_NSID]);
708 if (tb[NETNSA_PID]) {
709 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
710 nla = tb[NETNSA_PID];
711 } else if (tb[NETNSA_FD]) {
712 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
715 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
719 NL_SET_BAD_ATTR(extack, nla);
720 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
721 return PTR_ERR(peer);
724 spin_lock_bh(&net->nsid_lock);
725 if (__peernet2id(net, peer) >= 0) {
726 spin_unlock_bh(&net->nsid_lock);
728 NL_SET_BAD_ATTR(extack, nla);
729 NL_SET_ERR_MSG(extack,
730 "Peer netns already has a nsid assigned");
734 err = alloc_netid(net, peer, nsid);
735 spin_unlock_bh(&net->nsid_lock);
737 rtnl_net_notifyid(net, RTM_NEWNSID, err, GFP_KERNEL);
739 } else if (err == -ENOSPC && nsid >= 0) {
741 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
742 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
749 static int rtnl_net_get_size(void)
751 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
752 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
756 static int rtnl_net_fill(struct sk_buff *skb, u32 portid, u32 seq, int flags,
757 int cmd, struct net *net, int nsid)
759 struct nlmsghdr *nlh;
760 struct rtgenmsg *rth;
762 nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rth), flags);
766 rth = nlmsg_data(nlh);
767 rth->rtgen_family = AF_UNSPEC;
769 if (nla_put_s32(skb, NETNSA_NSID, nsid))
770 goto nla_put_failure;
776 nlmsg_cancel(skb, nlh);
780 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
781 struct netlink_ext_ack *extack)
783 struct net *net = sock_net(skb->sk);
784 struct nlattr *tb[NETNSA_MAX + 1];
790 err = nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, NETNSA_MAX,
791 rtnl_net_policy, extack);
794 if (tb[NETNSA_PID]) {
795 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
796 nla = tb[NETNSA_PID];
797 } else if (tb[NETNSA_FD]) {
798 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
801 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
806 NL_SET_BAD_ATTR(extack, nla);
807 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
808 return PTR_ERR(peer);
811 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
817 id = peernet2id(net, peer);
818 err = rtnl_net_fill(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
819 RTM_NEWNSID, net, id);
823 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
833 struct rtnl_net_dump_cb {
836 struct netlink_callback *cb;
841 /* Runs in RCU-critical section. */
842 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
844 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
847 if (net_cb->idx < net_cb->s_idx)
850 ret = rtnl_net_fill(net_cb->skb, NETLINK_CB(net_cb->cb->skb).portid,
851 net_cb->cb->nlh->nlmsg_seq, NLM_F_MULTI,
852 RTM_NEWNSID, net_cb->net, id);
861 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
863 struct net *net = sock_net(skb->sk);
864 struct rtnl_net_dump_cb net_cb = {
869 .s_idx = cb->args[0],
873 idr_for_each(&net->netns_ids, rtnl_net_dumpid_one, &net_cb);
876 cb->args[0] = net_cb.idx;
880 static void rtnl_net_notifyid(struct net *net, int cmd, int id, gfp_t gfp)
885 msg = nlmsg_new(rtnl_net_get_size(), gfp);
889 err = rtnl_net_fill(msg, 0, 0, 0, cmd, net, id);
893 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, gfp);
899 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
902 static int __init net_ns_init(void)
904 struct net_generic *ng;
907 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
909 SLAB_PANIC|SLAB_ACCOUNT, NULL);
911 /* Create workqueue for cleanup */
912 netns_wq = create_singlethread_workqueue("netns");
914 panic("Could not create netns workq");
917 ng = net_alloc_generic();
919 panic("Could not allocate generic netns");
921 rcu_assign_pointer(init_net.gen, ng);
923 down_write(&pernet_ops_rwsem);
924 if (setup_net(&init_net, &init_user_ns))
925 panic("Could not setup the initial network namespace");
927 init_net_initialized = true;
928 up_write(&pernet_ops_rwsem);
930 if (register_pernet_subsys(&net_ns_ops))
931 panic("Could not register network namespace subsystems");
933 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
934 RTNL_FLAG_DOIT_UNLOCKED);
935 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
936 RTNL_FLAG_DOIT_UNLOCKED);
941 pure_initcall(net_ns_init);
944 static int __register_pernet_operations(struct list_head *list,
945 struct pernet_operations *ops)
949 LIST_HEAD(net_exit_list);
951 list_add_tail(&ops->list, list);
952 if (ops->init || (ops->id && ops->size)) {
953 /* We held write locked pernet_ops_rwsem, and parallel
954 * setup_net() and cleanup_net() are not possible.
957 error = ops_init(ops, net);
960 list_add_tail(&net->exit_list, &net_exit_list);
966 /* If I have an error cleanup all namespaces I initialized */
967 list_del(&ops->list);
968 ops_exit_list(ops, &net_exit_list);
969 ops_free_list(ops, &net_exit_list);
973 static void __unregister_pernet_operations(struct pernet_operations *ops)
976 LIST_HEAD(net_exit_list);
978 list_del(&ops->list);
979 /* See comment in __register_pernet_operations() */
981 list_add_tail(&net->exit_list, &net_exit_list);
982 ops_exit_list(ops, &net_exit_list);
983 ops_free_list(ops, &net_exit_list);
988 static int __register_pernet_operations(struct list_head *list,
989 struct pernet_operations *ops)
991 if (!init_net_initialized) {
992 list_add_tail(&ops->list, list);
996 return ops_init(ops, &init_net);
999 static void __unregister_pernet_operations(struct pernet_operations *ops)
1001 if (!init_net_initialized) {
1002 list_del(&ops->list);
1004 LIST_HEAD(net_exit_list);
1005 list_add(&init_net.exit_list, &net_exit_list);
1006 ops_exit_list(ops, &net_exit_list);
1007 ops_free_list(ops, &net_exit_list);
1011 #endif /* CONFIG_NET_NS */
1013 static DEFINE_IDA(net_generic_ids);
1015 static int register_pernet_operations(struct list_head *list,
1016 struct pernet_operations *ops)
1021 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1026 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1028 error = __register_pernet_operations(list, ops);
1032 ida_free(&net_generic_ids, *ops->id);
1038 static void unregister_pernet_operations(struct pernet_operations *ops)
1040 __unregister_pernet_operations(ops);
1043 ida_free(&net_generic_ids, *ops->id);
1047 * register_pernet_subsys - register a network namespace subsystem
1048 * @ops: pernet operations structure for the subsystem
1050 * Register a subsystem which has init and exit functions
1051 * that are called when network namespaces are created and
1052 * destroyed respectively.
1054 * When registered all network namespace init functions are
1055 * called for every existing network namespace. Allowing kernel
1056 * modules to have a race free view of the set of network namespaces.
1058 * When a new network namespace is created all of the init
1059 * methods are called in the order in which they were registered.
1061 * When a network namespace is destroyed all of the exit methods
1062 * are called in the reverse of the order with which they were
1065 int register_pernet_subsys(struct pernet_operations *ops)
1068 down_write(&pernet_ops_rwsem);
1069 error = register_pernet_operations(first_device, ops);
1070 up_write(&pernet_ops_rwsem);
1073 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1076 * unregister_pernet_subsys - unregister a network namespace subsystem
1077 * @ops: pernet operations structure to manipulate
1079 * Remove the pernet operations structure from the list to be
1080 * used when network namespaces are created or destroyed. In
1081 * addition run the exit method for all existing network
1084 void unregister_pernet_subsys(struct pernet_operations *ops)
1086 down_write(&pernet_ops_rwsem);
1087 unregister_pernet_operations(ops);
1088 up_write(&pernet_ops_rwsem);
1090 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1093 * register_pernet_device - register a network namespace device
1094 * @ops: pernet operations structure for the subsystem
1096 * Register a device which has init and exit functions
1097 * that are called when network namespaces are created and
1098 * destroyed respectively.
1100 * When registered all network namespace init functions are
1101 * called for every existing network namespace. Allowing kernel
1102 * modules to have a race free view of the set of network namespaces.
1104 * When a new network namespace is created all of the init
1105 * methods are called in the order in which they were registered.
1107 * When a network namespace is destroyed all of the exit methods
1108 * are called in the reverse of the order with which they were
1111 int register_pernet_device(struct pernet_operations *ops)
1114 down_write(&pernet_ops_rwsem);
1115 error = register_pernet_operations(&pernet_list, ops);
1116 if (!error && (first_device == &pernet_list))
1117 first_device = &ops->list;
1118 up_write(&pernet_ops_rwsem);
1121 EXPORT_SYMBOL_GPL(register_pernet_device);
1124 * unregister_pernet_device - unregister a network namespace netdevice
1125 * @ops: pernet operations structure to manipulate
1127 * Remove the pernet operations structure from the list to be
1128 * used when network namespaces are created or destroyed. In
1129 * addition run the exit method for all existing network
1132 void unregister_pernet_device(struct pernet_operations *ops)
1134 down_write(&pernet_ops_rwsem);
1135 if (&ops->list == first_device)
1136 first_device = first_device->next;
1137 unregister_pernet_operations(ops);
1138 up_write(&pernet_ops_rwsem);
1140 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1142 #ifdef CONFIG_NET_NS
1143 static struct ns_common *netns_get(struct task_struct *task)
1145 struct net *net = NULL;
1146 struct nsproxy *nsproxy;
1149 nsproxy = task->nsproxy;
1151 net = get_net(nsproxy->net_ns);
1154 return net ? &net->ns : NULL;
1157 static inline struct net *to_net_ns(struct ns_common *ns)
1159 return container_of(ns, struct net, ns);
1162 static void netns_put(struct ns_common *ns)
1164 put_net(to_net_ns(ns));
1167 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1169 struct net *net = to_net_ns(ns);
1171 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1172 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1175 put_net(nsproxy->net_ns);
1176 nsproxy->net_ns = get_net(net);
1180 static struct user_namespace *netns_owner(struct ns_common *ns)
1182 return to_net_ns(ns)->user_ns;
1185 const struct proc_ns_operations netns_operations = {
1187 .type = CLONE_NEWNET,
1190 .install = netns_install,
1191 .owner = netns_owner,