1 // SPDX-License-Identifier: GPL-2.0-only
2 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
4 #include <linux/workqueue.h>
5 #include <linux/rtnetlink.h>
6 #include <linux/cache.h>
7 #include <linux/slab.h>
8 #include <linux/list.h>
9 #include <linux/delay.h>
10 #include <linux/sched.h>
11 #include <linux/idr.h>
12 #include <linux/rculist.h>
13 #include <linux/nsproxy.h>
15 #include <linux/proc_ns.h>
16 #include <linux/file.h>
17 #include <linux/export.h>
18 #include <linux/user_namespace.h>
19 #include <linux/net_namespace.h>
20 #include <linux/sched/task.h>
21 #include <linux/uidgid.h>
24 #include <net/netlink.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
29 * Our network namespace constructor/destructor lists
32 static LIST_HEAD(pernet_list);
33 static struct list_head *first_device = &pernet_list;
35 LIST_HEAD(net_namespace_list);
36 EXPORT_SYMBOL_GPL(net_namespace_list);
38 /* Protects net_namespace_list. Nests iside rtnl_lock() */
39 DECLARE_RWSEM(net_rwsem);
40 EXPORT_SYMBOL_GPL(net_rwsem);
43 static struct key_tag init_net_key_domain = { .usage = REFCOUNT_INIT(1) };
46 struct net init_net = {
47 .count = REFCOUNT_INIT(1),
48 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
50 .key_domain = &init_net_key_domain,
53 EXPORT_SYMBOL(init_net);
55 static bool init_net_initialized;
57 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
58 * init_net_initialized and first_device pointer.
59 * This is internal net namespace object. Please, don't use it
62 DECLARE_RWSEM(pernet_ops_rwsem);
63 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
65 #define MIN_PERNET_OPS_ID \
66 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
68 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
70 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
72 static struct net_generic *net_alloc_generic(void)
74 struct net_generic *ng;
75 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
77 ng = kzalloc(generic_size, GFP_KERNEL);
79 ng->s.len = max_gen_ptrs;
84 static int net_assign_generic(struct net *net, unsigned int id, void *data)
86 struct net_generic *ng, *old_ng;
88 BUG_ON(id < MIN_PERNET_OPS_ID);
90 old_ng = rcu_dereference_protected(net->gen,
91 lockdep_is_held(&pernet_ops_rwsem));
92 if (old_ng->s.len > id) {
93 old_ng->ptr[id] = data;
97 ng = net_alloc_generic();
102 * Some synchronisation notes:
104 * The net_generic explores the net->gen array inside rcu
105 * read section. Besides once set the net->gen->ptr[x]
106 * pointer never changes (see rules in netns/generic.h).
108 * That said, we simply duplicate this array and schedule
109 * the old copy for kfree after a grace period.
112 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
113 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
116 rcu_assign_pointer(net->gen, ng);
117 kfree_rcu(old_ng, s.rcu);
121 static int ops_init(const struct pernet_operations *ops, struct net *net)
126 if (ops->id && ops->size) {
127 data = kzalloc(ops->size, GFP_KERNEL);
131 err = net_assign_generic(net, *ops->id, data);
137 err = ops->init(net);
148 static void ops_free(const struct pernet_operations *ops, struct net *net)
150 if (ops->id && ops->size) {
151 kfree(net_generic(net, *ops->id));
155 static void ops_pre_exit_list(const struct pernet_operations *ops,
156 struct list_head *net_exit_list)
161 list_for_each_entry(net, net_exit_list, exit_list)
166 static void ops_exit_list(const struct pernet_operations *ops,
167 struct list_head *net_exit_list)
171 list_for_each_entry(net, net_exit_list, exit_list) {
177 ops->exit_batch(net_exit_list);
180 static void ops_free_list(const struct pernet_operations *ops,
181 struct list_head *net_exit_list)
184 if (ops->size && ops->id) {
185 list_for_each_entry(net, net_exit_list, exit_list)
190 /* should be called with nsid_lock held */
191 static int alloc_netid(struct net *net, struct net *peer, int reqid)
193 int min = 0, max = 0;
200 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
203 /* This function is used by idr_for_each(). If net is equal to peer, the
204 * function returns the id so that idr_for_each() stops. Because we cannot
205 * returns the id 0 (idr_for_each() will not stop), we return the magic value
206 * NET_ID_ZERO (-1) for it.
208 #define NET_ID_ZERO -1
209 static int net_eq_idr(int id, void *net, void *peer)
211 if (net_eq(net, peer))
212 return id ? : NET_ID_ZERO;
216 /* Must be called from RCU-critical section or with nsid_lock held. If
217 * a new id is assigned, the bool alloc is set to true, thus the
218 * caller knows that the new id must be notified via rtnl.
220 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
222 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
223 bool alloc_it = *alloc;
227 /* Magic value for id 0. */
228 if (id == NET_ID_ZERO)
234 id = alloc_netid(net, peer, -1);
236 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
239 return NETNSA_NSID_NOT_ASSIGNED;
242 /* Must be called from RCU-critical section or with nsid_lock held */
243 static int __peernet2id(struct net *net, struct net *peer)
247 return __peernet2id_alloc(net, peer, &no);
250 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
251 struct nlmsghdr *nlh, gfp_t gfp);
252 /* This function returns the id of a peer netns. If no id is assigned, one will
253 * be allocated and returned.
255 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
257 bool alloc = false, alive = false;
260 if (refcount_read(&net->count) == 0)
261 return NETNSA_NSID_NOT_ASSIGNED;
262 spin_lock_bh(&net->nsid_lock);
264 * When peer is obtained from RCU lists, we may race with
265 * its cleanup. Check whether it's alive, and this guarantees
266 * we never hash a peer back to net->netns_ids, after it has
267 * just been idr_remove()'d from there in cleanup_net().
269 if (maybe_get_net(peer))
270 alive = alloc = true;
271 id = __peernet2id_alloc(net, peer, &alloc);
272 spin_unlock_bh(&net->nsid_lock);
273 if (alloc && id >= 0)
274 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
279 EXPORT_SYMBOL_GPL(peernet2id_alloc);
281 /* This function returns, if assigned, the id of a peer netns. */
282 int peernet2id(struct net *net, struct net *peer)
287 id = __peernet2id(net, peer);
292 EXPORT_SYMBOL(peernet2id);
294 /* This function returns true is the peer netns has an id assigned into the
297 bool peernet_has_id(struct net *net, struct net *peer)
299 return peernet2id(net, peer) >= 0;
302 struct net *get_net_ns_by_id(struct net *net, int id)
310 peer = idr_find(&net->netns_ids, id);
312 peer = maybe_get_net(peer);
319 * setup_net runs the initializers for the network namespace object.
321 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
323 /* Must be called with pernet_ops_rwsem held */
324 const struct pernet_operations *ops, *saved_ops;
326 LIST_HEAD(net_exit_list);
328 refcount_set(&net->count, 1);
329 refcount_set(&net->passive, 1);
330 get_random_bytes(&net->hash_mix, sizeof(u32));
331 net->dev_base_seq = 1;
332 net->user_ns = user_ns;
333 idr_init(&net->netns_ids);
334 spin_lock_init(&net->nsid_lock);
335 mutex_init(&net->ipv4.ra_mutex);
337 list_for_each_entry(ops, &pernet_list, list) {
338 error = ops_init(ops, net);
342 down_write(&net_rwsem);
343 list_add_tail_rcu(&net->list, &net_namespace_list);
344 up_write(&net_rwsem);
349 /* Walk through the list backwards calling the exit functions
350 * for the pernet modules whose init functions did not fail.
352 list_add(&net->exit_list, &net_exit_list);
354 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
355 ops_pre_exit_list(ops, &net_exit_list);
360 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
361 ops_exit_list(ops, &net_exit_list);
364 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
365 ops_free_list(ops, &net_exit_list);
371 static int __net_init net_defaults_init_net(struct net *net)
373 net->core.sysctl_somaxconn = SOMAXCONN;
377 static struct pernet_operations net_defaults_ops = {
378 .init = net_defaults_init_net,
381 static __init int net_defaults_init(void)
383 if (register_pernet_subsys(&net_defaults_ops))
384 panic("Cannot initialize net default settings");
389 core_initcall(net_defaults_init);
392 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
394 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
397 static void dec_net_namespaces(struct ucounts *ucounts)
399 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
402 static struct kmem_cache *net_cachep __ro_after_init;
403 static struct workqueue_struct *netns_wq;
405 static struct net *net_alloc(void)
407 struct net *net = NULL;
408 struct net_generic *ng;
410 ng = net_alloc_generic();
414 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
419 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
420 if (!net->key_domain)
422 refcount_set(&net->key_domain->usage, 1);
425 rcu_assign_pointer(net->gen, ng);
431 kmem_cache_free(net_cachep, net);
439 static void net_free(struct net *net)
441 kfree(rcu_access_pointer(net->gen));
442 kmem_cache_free(net_cachep, net);
445 void net_drop_ns(void *p)
448 if (ns && refcount_dec_and_test(&ns->passive))
452 struct net *copy_net_ns(unsigned long flags,
453 struct user_namespace *user_ns, struct net *old_net)
455 struct ucounts *ucounts;
459 if (!(flags & CLONE_NEWNET))
460 return get_net(old_net);
462 ucounts = inc_net_namespaces(user_ns);
464 return ERR_PTR(-ENOSPC);
471 refcount_set(&net->passive, 1);
472 net->ucounts = ucounts;
473 get_user_ns(user_ns);
475 rv = down_read_killable(&pernet_ops_rwsem);
479 rv = setup_net(net, user_ns);
481 up_read(&pernet_ops_rwsem);
486 key_remove_domain(net->key_domain);
488 put_user_ns(user_ns);
491 dec_net_namespaces(ucounts);
498 * net_ns_get_ownership - get sysfs ownership data for @net
499 * @net: network namespace in question (can be NULL)
500 * @uid: kernel user ID for sysfs objects
501 * @gid: kernel group ID for sysfs objects
503 * Returns the uid/gid pair of root in the user namespace associated with the
504 * given network namespace.
506 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
509 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
510 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
512 if (uid_valid(ns_root_uid))
515 if (gid_valid(ns_root_gid))
518 *uid = GLOBAL_ROOT_UID;
519 *gid = GLOBAL_ROOT_GID;
522 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
524 static void unhash_nsid(struct net *net, struct net *last)
527 /* This function is only called from cleanup_net() work,
528 * and this work is the only process, that may delete
529 * a net from net_namespace_list. So, when the below
530 * is executing, the list may only grow. Thus, we do not
531 * use for_each_net_rcu() or net_rwsem.
536 spin_lock_bh(&tmp->nsid_lock);
537 id = __peernet2id(tmp, net);
539 idr_remove(&tmp->netns_ids, id);
540 spin_unlock_bh(&tmp->nsid_lock);
542 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
547 spin_lock_bh(&net->nsid_lock);
548 idr_destroy(&net->netns_ids);
549 spin_unlock_bh(&net->nsid_lock);
552 static LLIST_HEAD(cleanup_list);
554 static void cleanup_net(struct work_struct *work)
556 const struct pernet_operations *ops;
557 struct net *net, *tmp, *last;
558 struct llist_node *net_kill_list;
559 LIST_HEAD(net_exit_list);
561 /* Atomically snapshot the list of namespaces to cleanup */
562 net_kill_list = llist_del_all(&cleanup_list);
564 down_read(&pernet_ops_rwsem);
566 /* Don't let anyone else find us. */
567 down_write(&net_rwsem);
568 llist_for_each_entry(net, net_kill_list, cleanup_list)
569 list_del_rcu(&net->list);
570 /* Cache last net. After we unlock rtnl, no one new net
571 * added to net_namespace_list can assign nsid pointer
572 * to a net from net_kill_list (see peernet2id_alloc()).
573 * So, we skip them in unhash_nsid().
575 * Note, that unhash_nsid() does not delete nsid links
576 * between net_kill_list's nets, as they've already
577 * deleted from net_namespace_list. But, this would be
578 * useless anyway, as netns_ids are destroyed there.
580 last = list_last_entry(&net_namespace_list, struct net, list);
581 up_write(&net_rwsem);
583 llist_for_each_entry(net, net_kill_list, cleanup_list) {
584 unhash_nsid(net, last);
585 list_add_tail(&net->exit_list, &net_exit_list);
588 /* Run all of the network namespace pre_exit methods */
589 list_for_each_entry_reverse(ops, &pernet_list, list)
590 ops_pre_exit_list(ops, &net_exit_list);
593 * Another CPU might be rcu-iterating the list, wait for it.
594 * This needs to be before calling the exit() notifiers, so
595 * the rcu_barrier() below isn't sufficient alone.
596 * Also the pre_exit() and exit() methods need this barrier.
600 /* Run all of the network namespace exit methods */
601 list_for_each_entry_reverse(ops, &pernet_list, list)
602 ops_exit_list(ops, &net_exit_list);
604 /* Free the net generic variables */
605 list_for_each_entry_reverse(ops, &pernet_list, list)
606 ops_free_list(ops, &net_exit_list);
608 up_read(&pernet_ops_rwsem);
610 /* Ensure there are no outstanding rcu callbacks using this
615 /* Finally it is safe to free my network namespace structure */
616 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
617 list_del_init(&net->exit_list);
618 dec_net_namespaces(net->ucounts);
620 key_remove_domain(net->key_domain);
622 put_user_ns(net->user_ns);
628 * net_ns_barrier - wait until concurrent net_cleanup_work is done
630 * cleanup_net runs from work queue and will first remove namespaces
631 * from the global list, then run net exit functions.
633 * Call this in module exit path to make sure that all netns
634 * ->exit ops have been invoked before the function is removed.
636 void net_ns_barrier(void)
638 down_write(&pernet_ops_rwsem);
639 up_write(&pernet_ops_rwsem);
641 EXPORT_SYMBOL(net_ns_barrier);
643 static DECLARE_WORK(net_cleanup_work, cleanup_net);
645 void __put_net(struct net *net)
647 /* Cleanup the network namespace in process context */
648 if (llist_add(&net->cleanup_list, &cleanup_list))
649 queue_work(netns_wq, &net_cleanup_work);
651 EXPORT_SYMBOL_GPL(__put_net);
654 * get_net_ns - increment the refcount of the network namespace
655 * @ns: common namespace (net)
657 * Returns the net's common namespace.
659 struct ns_common *get_net_ns(struct ns_common *ns)
661 return &get_net(container_of(ns, struct net, ns))->ns;
663 EXPORT_SYMBOL_GPL(get_net_ns);
665 struct net *get_net_ns_by_fd(int fd)
668 struct ns_common *ns;
671 file = proc_ns_fget(fd);
673 return ERR_CAST(file);
675 ns = get_proc_ns(file_inode(file));
676 if (ns->ops == &netns_operations)
677 net = get_net(container_of(ns, struct net, ns));
679 net = ERR_PTR(-EINVAL);
686 struct net *get_net_ns_by_fd(int fd)
688 return ERR_PTR(-EINVAL);
691 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
693 struct net *get_net_ns_by_pid(pid_t pid)
695 struct task_struct *tsk;
698 /* Lookup the network namespace */
699 net = ERR_PTR(-ESRCH);
701 tsk = find_task_by_vpid(pid);
703 struct nsproxy *nsproxy;
705 nsproxy = tsk->nsproxy;
707 net = get_net(nsproxy->net_ns);
713 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
715 static __net_init int net_ns_net_init(struct net *net)
718 net->ns.ops = &netns_operations;
720 return ns_alloc_inum(&net->ns);
723 static __net_exit void net_ns_net_exit(struct net *net)
725 ns_free_inum(&net->ns);
728 static struct pernet_operations __net_initdata net_ns_ops = {
729 .init = net_ns_net_init,
730 .exit = net_ns_net_exit,
733 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
734 [NETNSA_NONE] = { .type = NLA_UNSPEC },
735 [NETNSA_NSID] = { .type = NLA_S32 },
736 [NETNSA_PID] = { .type = NLA_U32 },
737 [NETNSA_FD] = { .type = NLA_U32 },
738 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
741 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
742 struct netlink_ext_ack *extack)
744 struct net *net = sock_net(skb->sk);
745 struct nlattr *tb[NETNSA_MAX + 1];
750 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
751 NETNSA_MAX, rtnl_net_policy, extack);
754 if (!tb[NETNSA_NSID]) {
755 NL_SET_ERR_MSG(extack, "nsid is missing");
758 nsid = nla_get_s32(tb[NETNSA_NSID]);
760 if (tb[NETNSA_PID]) {
761 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
762 nla = tb[NETNSA_PID];
763 } else if (tb[NETNSA_FD]) {
764 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
767 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
771 NL_SET_BAD_ATTR(extack, nla);
772 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
773 return PTR_ERR(peer);
776 spin_lock_bh(&net->nsid_lock);
777 if (__peernet2id(net, peer) >= 0) {
778 spin_unlock_bh(&net->nsid_lock);
780 NL_SET_BAD_ATTR(extack, nla);
781 NL_SET_ERR_MSG(extack,
782 "Peer netns already has a nsid assigned");
786 err = alloc_netid(net, peer, nsid);
787 spin_unlock_bh(&net->nsid_lock);
789 rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
792 } else if (err == -ENOSPC && nsid >= 0) {
794 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
795 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
802 static int rtnl_net_get_size(void)
804 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
805 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
806 + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
810 struct net_fill_args {
820 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
822 struct nlmsghdr *nlh;
823 struct rtgenmsg *rth;
825 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
830 rth = nlmsg_data(nlh);
831 rth->rtgen_family = AF_UNSPEC;
833 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
834 goto nla_put_failure;
837 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
838 goto nla_put_failure;
844 nlmsg_cancel(skb, nlh);
848 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
849 const struct nlmsghdr *nlh,
851 struct netlink_ext_ack *extack)
855 if (!netlink_strict_get_check(skb))
856 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
857 tb, NETNSA_MAX, rtnl_net_policy,
860 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
861 NETNSA_MAX, rtnl_net_policy,
866 for (i = 0; i <= NETNSA_MAX; i++) {
874 case NETNSA_TARGET_NSID:
877 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
885 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
886 struct netlink_ext_ack *extack)
888 struct net *net = sock_net(skb->sk);
889 struct nlattr *tb[NETNSA_MAX + 1];
890 struct net_fill_args fillargs = {
891 .portid = NETLINK_CB(skb).portid,
892 .seq = nlh->nlmsg_seq,
895 struct net *peer, *target = net;
900 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
903 if (tb[NETNSA_PID]) {
904 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
905 nla = tb[NETNSA_PID];
906 } else if (tb[NETNSA_FD]) {
907 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
909 } else if (tb[NETNSA_NSID]) {
910 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
912 peer = ERR_PTR(-ENOENT);
913 nla = tb[NETNSA_NSID];
915 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
920 NL_SET_BAD_ATTR(extack, nla);
921 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
922 return PTR_ERR(peer);
925 if (tb[NETNSA_TARGET_NSID]) {
926 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
928 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
929 if (IS_ERR(target)) {
930 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
931 NL_SET_ERR_MSG(extack,
932 "Target netns reference is invalid");
933 err = PTR_ERR(target);
936 fillargs.add_ref = true;
937 fillargs.ref_nsid = peernet2id(net, peer);
940 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
946 fillargs.nsid = peernet2id(target, peer);
947 err = rtnl_net_fill(msg, &fillargs);
951 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
957 if (fillargs.add_ref)
963 struct rtnl_net_dump_cb {
967 struct net_fill_args fillargs;
972 /* Runs in RCU-critical section. */
973 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
975 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
978 if (net_cb->idx < net_cb->s_idx)
981 net_cb->fillargs.nsid = id;
982 if (net_cb->fillargs.add_ref)
983 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
984 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
993 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
994 struct rtnl_net_dump_cb *net_cb,
995 struct netlink_callback *cb)
997 struct netlink_ext_ack *extack = cb->extack;
998 struct nlattr *tb[NETNSA_MAX + 1];
1001 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
1002 NETNSA_MAX, rtnl_net_policy,
1007 for (i = 0; i <= NETNSA_MAX; i++) {
1011 if (i == NETNSA_TARGET_NSID) {
1014 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
1016 NL_SET_BAD_ATTR(extack, tb[i]);
1017 NL_SET_ERR_MSG(extack,
1018 "Invalid target network namespace id");
1019 return PTR_ERR(net);
1021 net_cb->fillargs.add_ref = true;
1022 net_cb->ref_net = net_cb->tgt_net;
1023 net_cb->tgt_net = net;
1025 NL_SET_BAD_ATTR(extack, tb[i]);
1026 NL_SET_ERR_MSG(extack,
1027 "Unsupported attribute in dump request");
1035 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
1037 struct rtnl_net_dump_cb net_cb = {
1038 .tgt_net = sock_net(skb->sk),
1041 .portid = NETLINK_CB(cb->skb).portid,
1042 .seq = cb->nlh->nlmsg_seq,
1043 .flags = NLM_F_MULTI,
1047 .s_idx = cb->args[0],
1051 if (cb->strict_check) {
1052 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1058 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1061 cb->args[0] = net_cb.idx;
1063 if (net_cb.fillargs.add_ref)
1064 put_net(net_cb.tgt_net);
1065 return err < 0 ? err : skb->len;
1068 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
1069 struct nlmsghdr *nlh, gfp_t gfp)
1071 struct net_fill_args fillargs = {
1073 .seq = nlh ? nlh->nlmsg_seq : 0,
1077 struct sk_buff *msg;
1080 msg = nlmsg_new(rtnl_net_get_size(), gfp);
1084 err = rtnl_net_fill(msg, &fillargs);
1088 rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, gfp);
1094 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1097 static int __init net_ns_init(void)
1099 struct net_generic *ng;
1101 #ifdef CONFIG_NET_NS
1102 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1104 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1106 /* Create workqueue for cleanup */
1107 netns_wq = create_singlethread_workqueue("netns");
1109 panic("Could not create netns workq");
1112 ng = net_alloc_generic();
1114 panic("Could not allocate generic netns");
1116 rcu_assign_pointer(init_net.gen, ng);
1118 down_write(&pernet_ops_rwsem);
1119 if (setup_net(&init_net, &init_user_ns))
1120 panic("Could not setup the initial network namespace");
1122 init_net_initialized = true;
1123 up_write(&pernet_ops_rwsem);
1125 if (register_pernet_subsys(&net_ns_ops))
1126 panic("Could not register network namespace subsystems");
1128 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1129 RTNL_FLAG_DOIT_UNLOCKED);
1130 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1131 RTNL_FLAG_DOIT_UNLOCKED);
1136 pure_initcall(net_ns_init);
1138 #ifdef CONFIG_NET_NS
1139 static int __register_pernet_operations(struct list_head *list,
1140 struct pernet_operations *ops)
1144 LIST_HEAD(net_exit_list);
1146 list_add_tail(&ops->list, list);
1147 if (ops->init || (ops->id && ops->size)) {
1148 /* We held write locked pernet_ops_rwsem, and parallel
1149 * setup_net() and cleanup_net() are not possible.
1152 error = ops_init(ops, net);
1155 list_add_tail(&net->exit_list, &net_exit_list);
1161 /* If I have an error cleanup all namespaces I initialized */
1162 list_del(&ops->list);
1163 ops_pre_exit_list(ops, &net_exit_list);
1165 ops_exit_list(ops, &net_exit_list);
1166 ops_free_list(ops, &net_exit_list);
1170 static void __unregister_pernet_operations(struct pernet_operations *ops)
1173 LIST_HEAD(net_exit_list);
1175 list_del(&ops->list);
1176 /* See comment in __register_pernet_operations() */
1178 list_add_tail(&net->exit_list, &net_exit_list);
1179 ops_pre_exit_list(ops, &net_exit_list);
1181 ops_exit_list(ops, &net_exit_list);
1182 ops_free_list(ops, &net_exit_list);
1187 static int __register_pernet_operations(struct list_head *list,
1188 struct pernet_operations *ops)
1190 if (!init_net_initialized) {
1191 list_add_tail(&ops->list, list);
1195 return ops_init(ops, &init_net);
1198 static void __unregister_pernet_operations(struct pernet_operations *ops)
1200 if (!init_net_initialized) {
1201 list_del(&ops->list);
1203 LIST_HEAD(net_exit_list);
1204 list_add(&init_net.exit_list, &net_exit_list);
1205 ops_pre_exit_list(ops, &net_exit_list);
1207 ops_exit_list(ops, &net_exit_list);
1208 ops_free_list(ops, &net_exit_list);
1212 #endif /* CONFIG_NET_NS */
1214 static DEFINE_IDA(net_generic_ids);
1216 static int register_pernet_operations(struct list_head *list,
1217 struct pernet_operations *ops)
1222 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1227 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1229 error = __register_pernet_operations(list, ops);
1233 ida_free(&net_generic_ids, *ops->id);
1239 static void unregister_pernet_operations(struct pernet_operations *ops)
1241 __unregister_pernet_operations(ops);
1244 ida_free(&net_generic_ids, *ops->id);
1248 * register_pernet_subsys - register a network namespace subsystem
1249 * @ops: pernet operations structure for the subsystem
1251 * Register a subsystem which has init and exit functions
1252 * that are called when network namespaces are created and
1253 * destroyed respectively.
1255 * When registered all network namespace init functions are
1256 * called for every existing network namespace. Allowing kernel
1257 * modules to have a race free view of the set of network namespaces.
1259 * When a new network namespace is created all of the init
1260 * methods are called in the order in which they were registered.
1262 * When a network namespace is destroyed all of the exit methods
1263 * are called in the reverse of the order with which they were
1266 int register_pernet_subsys(struct pernet_operations *ops)
1269 down_write(&pernet_ops_rwsem);
1270 error = register_pernet_operations(first_device, ops);
1271 up_write(&pernet_ops_rwsem);
1274 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1277 * unregister_pernet_subsys - unregister a network namespace subsystem
1278 * @ops: pernet operations structure to manipulate
1280 * Remove the pernet operations structure from the list to be
1281 * used when network namespaces are created or destroyed. In
1282 * addition run the exit method for all existing network
1285 void unregister_pernet_subsys(struct pernet_operations *ops)
1287 down_write(&pernet_ops_rwsem);
1288 unregister_pernet_operations(ops);
1289 up_write(&pernet_ops_rwsem);
1291 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1294 * register_pernet_device - register a network namespace device
1295 * @ops: pernet operations structure for the subsystem
1297 * Register a device which has init and exit functions
1298 * that are called when network namespaces are created and
1299 * destroyed respectively.
1301 * When registered all network namespace init functions are
1302 * called for every existing network namespace. Allowing kernel
1303 * modules to have a race free view of the set of network namespaces.
1305 * When a new network namespace is created all of the init
1306 * methods are called in the order in which they were registered.
1308 * When a network namespace is destroyed all of the exit methods
1309 * are called in the reverse of the order with which they were
1312 int register_pernet_device(struct pernet_operations *ops)
1315 down_write(&pernet_ops_rwsem);
1316 error = register_pernet_operations(&pernet_list, ops);
1317 if (!error && (first_device == &pernet_list))
1318 first_device = &ops->list;
1319 up_write(&pernet_ops_rwsem);
1322 EXPORT_SYMBOL_GPL(register_pernet_device);
1325 * unregister_pernet_device - unregister a network namespace netdevice
1326 * @ops: pernet operations structure to manipulate
1328 * Remove the pernet operations structure from the list to be
1329 * used when network namespaces are created or destroyed. In
1330 * addition run the exit method for all existing network
1333 void unregister_pernet_device(struct pernet_operations *ops)
1335 down_write(&pernet_ops_rwsem);
1336 if (&ops->list == first_device)
1337 first_device = first_device->next;
1338 unregister_pernet_operations(ops);
1339 up_write(&pernet_ops_rwsem);
1341 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1343 #ifdef CONFIG_NET_NS
1344 static struct ns_common *netns_get(struct task_struct *task)
1346 struct net *net = NULL;
1347 struct nsproxy *nsproxy;
1350 nsproxy = task->nsproxy;
1352 net = get_net(nsproxy->net_ns);
1355 return net ? &net->ns : NULL;
1358 static inline struct net *to_net_ns(struct ns_common *ns)
1360 return container_of(ns, struct net, ns);
1363 static void netns_put(struct ns_common *ns)
1365 put_net(to_net_ns(ns));
1368 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1370 struct net *net = to_net_ns(ns);
1372 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1373 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1376 put_net(nsproxy->net_ns);
1377 nsproxy->net_ns = get_net(net);
1381 static struct user_namespace *netns_owner(struct ns_common *ns)
1383 return to_net_ns(ns)->user_ns;
1386 const struct proc_ns_operations netns_operations = {
1388 .type = CLONE_NEWNET,
1391 .install = netns_install,
1392 .owner = netns_owner,