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/sched/mm.h>
22 #include <linux/uidgid.h>
23 #include <linux/cookie.h>
26 #include <net/netlink.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
31 * Our network namespace constructor/destructor lists
34 static LIST_HEAD(pernet_list);
35 static struct list_head *first_device = &pernet_list;
37 LIST_HEAD(net_namespace_list);
38 EXPORT_SYMBOL_GPL(net_namespace_list);
40 /* Protects net_namespace_list. Nests iside rtnl_lock() */
41 DECLARE_RWSEM(net_rwsem);
42 EXPORT_SYMBOL_GPL(net_rwsem);
45 static struct key_tag init_net_key_domain = { .usage = REFCOUNT_INIT(1) };
49 EXPORT_SYMBOL(init_net);
51 static bool init_net_initialized;
53 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
54 * init_net_initialized and first_device pointer.
55 * This is internal net namespace object. Please, don't use it
58 DECLARE_RWSEM(pernet_ops_rwsem);
59 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
61 #define MIN_PERNET_OPS_ID \
62 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
64 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
66 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
68 DEFINE_COOKIE(net_cookie);
70 static struct net_generic *net_alloc_generic(void)
72 struct net_generic *ng;
73 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
75 ng = kzalloc(generic_size, GFP_KERNEL);
77 ng->s.len = max_gen_ptrs;
82 static int net_assign_generic(struct net *net, unsigned int id, void *data)
84 struct net_generic *ng, *old_ng;
86 BUG_ON(id < MIN_PERNET_OPS_ID);
88 old_ng = rcu_dereference_protected(net->gen,
89 lockdep_is_held(&pernet_ops_rwsem));
90 if (old_ng->s.len > id) {
91 old_ng->ptr[id] = data;
95 ng = net_alloc_generic();
100 * Some synchronisation notes:
102 * The net_generic explores the net->gen array inside rcu
103 * read section. Besides once set the net->gen->ptr[x]
104 * pointer never changes (see rules in netns/generic.h).
106 * That said, we simply duplicate this array and schedule
107 * the old copy for kfree after a grace period.
110 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
111 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
114 rcu_assign_pointer(net->gen, ng);
115 kfree_rcu(old_ng, s.rcu);
119 static int ops_init(const struct pernet_operations *ops, struct net *net)
124 if (ops->id && ops->size) {
125 data = kzalloc(ops->size, GFP_KERNEL);
129 err = net_assign_generic(net, *ops->id, data);
135 err = ops->init(net);
146 static void ops_pre_exit_list(const struct pernet_operations *ops,
147 struct list_head *net_exit_list)
152 list_for_each_entry(net, net_exit_list, exit_list)
157 static void ops_exit_list(const struct pernet_operations *ops,
158 struct list_head *net_exit_list)
162 list_for_each_entry(net, net_exit_list, exit_list) {
168 ops->exit_batch(net_exit_list);
171 static void ops_free_list(const struct pernet_operations *ops,
172 struct list_head *net_exit_list)
175 if (ops->size && ops->id) {
176 list_for_each_entry(net, net_exit_list, exit_list)
177 kfree(net_generic(net, *ops->id));
181 /* should be called with nsid_lock held */
182 static int alloc_netid(struct net *net, struct net *peer, int reqid)
184 int min = 0, max = 0;
191 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
194 /* This function is used by idr_for_each(). If net is equal to peer, the
195 * function returns the id so that idr_for_each() stops. Because we cannot
196 * returns the id 0 (idr_for_each() will not stop), we return the magic value
197 * NET_ID_ZERO (-1) for it.
199 #define NET_ID_ZERO -1
200 static int net_eq_idr(int id, void *net, void *peer)
202 if (net_eq(net, peer))
203 return id ? : NET_ID_ZERO;
207 /* Must be called from RCU-critical section or with nsid_lock held */
208 static int __peernet2id(const struct net *net, struct net *peer)
210 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
212 /* Magic value for id 0. */
213 if (id == NET_ID_ZERO)
218 return NETNSA_NSID_NOT_ASSIGNED;
221 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
222 struct nlmsghdr *nlh, gfp_t gfp);
223 /* This function returns the id of a peer netns. If no id is assigned, one will
224 * be allocated and returned.
226 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
230 if (refcount_read(&net->ns.count) == 0)
231 return NETNSA_NSID_NOT_ASSIGNED;
233 spin_lock_bh(&net->nsid_lock);
234 id = __peernet2id(net, peer);
236 spin_unlock_bh(&net->nsid_lock);
240 /* When peer is obtained from RCU lists, we may race with
241 * its cleanup. Check whether it's alive, and this guarantees
242 * we never hash a peer back to net->netns_ids, after it has
243 * just been idr_remove()'d from there in cleanup_net().
245 if (!maybe_get_net(peer)) {
246 spin_unlock_bh(&net->nsid_lock);
247 return NETNSA_NSID_NOT_ASSIGNED;
250 id = alloc_netid(net, peer, -1);
251 spin_unlock_bh(&net->nsid_lock);
255 return NETNSA_NSID_NOT_ASSIGNED;
257 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
261 EXPORT_SYMBOL_GPL(peernet2id_alloc);
263 /* This function returns, if assigned, the id of a peer netns. */
264 int peernet2id(const struct net *net, struct net *peer)
269 id = __peernet2id(net, peer);
274 EXPORT_SYMBOL(peernet2id);
276 /* This function returns true is the peer netns has an id assigned into the
279 bool peernet_has_id(const struct net *net, struct net *peer)
281 return peernet2id(net, peer) >= 0;
284 struct net *get_net_ns_by_id(const struct net *net, int id)
292 peer = idr_find(&net->netns_ids, id);
294 peer = maybe_get_net(peer);
299 EXPORT_SYMBOL_GPL(get_net_ns_by_id);
302 * setup_net runs the initializers for the network namespace object.
304 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
306 /* Must be called with pernet_ops_rwsem held */
307 const struct pernet_operations *ops, *saved_ops;
309 LIST_HEAD(net_exit_list);
311 refcount_set(&net->ns.count, 1);
312 ref_tracker_dir_init(&net->refcnt_tracker, 128);
314 refcount_set(&net->passive, 1);
315 get_random_bytes(&net->hash_mix, sizeof(u32));
317 net->net_cookie = gen_cookie_next(&net_cookie);
319 net->dev_base_seq = 1;
320 net->user_ns = user_ns;
321 idr_init(&net->netns_ids);
322 spin_lock_init(&net->nsid_lock);
323 mutex_init(&net->ipv4.ra_mutex);
325 list_for_each_entry(ops, &pernet_list, list) {
326 error = ops_init(ops, net);
330 down_write(&net_rwsem);
331 list_add_tail_rcu(&net->list, &net_namespace_list);
332 up_write(&net_rwsem);
337 /* Walk through the list backwards calling the exit functions
338 * for the pernet modules whose init functions did not fail.
340 list_add(&net->exit_list, &net_exit_list);
342 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
343 ops_pre_exit_list(ops, &net_exit_list);
348 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
349 ops_exit_list(ops, &net_exit_list);
352 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
353 ops_free_list(ops, &net_exit_list);
359 static int __net_init net_defaults_init_net(struct net *net)
361 net->core.sysctl_somaxconn = SOMAXCONN;
362 net->core.sysctl_txrehash = SOCK_TXREHASH_ENABLED;
367 static struct pernet_operations net_defaults_ops = {
368 .init = net_defaults_init_net,
371 static __init int net_defaults_init(void)
373 if (register_pernet_subsys(&net_defaults_ops))
374 panic("Cannot initialize net default settings");
379 core_initcall(net_defaults_init);
382 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
384 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
387 static void dec_net_namespaces(struct ucounts *ucounts)
389 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
392 static struct kmem_cache *net_cachep __ro_after_init;
393 static struct workqueue_struct *netns_wq;
395 static struct net *net_alloc(void)
397 struct net *net = NULL;
398 struct net_generic *ng;
400 ng = net_alloc_generic();
404 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
409 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
410 if (!net->key_domain)
412 refcount_set(&net->key_domain->usage, 1);
415 rcu_assign_pointer(net->gen, ng);
421 kmem_cache_free(net_cachep, net);
429 static void net_free(struct net *net)
431 if (refcount_dec_and_test(&net->passive)) {
432 kfree(rcu_access_pointer(net->gen));
433 kmem_cache_free(net_cachep, net);
437 void net_drop_ns(void *p)
439 struct net *net = (struct net *)p;
445 struct net *copy_net_ns(unsigned long flags,
446 struct user_namespace *user_ns, struct net *old_net)
448 struct ucounts *ucounts;
452 if (!(flags & CLONE_NEWNET))
453 return get_net(old_net);
455 ucounts = inc_net_namespaces(user_ns);
457 return ERR_PTR(-ENOSPC);
464 refcount_set(&net->passive, 1);
465 net->ucounts = ucounts;
466 get_user_ns(user_ns);
468 rv = down_read_killable(&pernet_ops_rwsem);
472 rv = setup_net(net, user_ns);
474 up_read(&pernet_ops_rwsem);
479 key_remove_domain(net->key_domain);
481 put_user_ns(user_ns);
484 dec_net_namespaces(ucounts);
491 * net_ns_get_ownership - get sysfs ownership data for @net
492 * @net: network namespace in question (can be NULL)
493 * @uid: kernel user ID for sysfs objects
494 * @gid: kernel group ID for sysfs objects
496 * Returns the uid/gid pair of root in the user namespace associated with the
497 * given network namespace.
499 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
502 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
503 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
505 if (uid_valid(ns_root_uid))
508 if (gid_valid(ns_root_gid))
511 *uid = GLOBAL_ROOT_UID;
512 *gid = GLOBAL_ROOT_GID;
515 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
517 static void unhash_nsid(struct net *net, struct net *last)
520 /* This function is only called from cleanup_net() work,
521 * and this work is the only process, that may delete
522 * a net from net_namespace_list. So, when the below
523 * is executing, the list may only grow. Thus, we do not
524 * use for_each_net_rcu() or net_rwsem.
529 spin_lock_bh(&tmp->nsid_lock);
530 id = __peernet2id(tmp, net);
532 idr_remove(&tmp->netns_ids, id);
533 spin_unlock_bh(&tmp->nsid_lock);
535 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
540 spin_lock_bh(&net->nsid_lock);
541 idr_destroy(&net->netns_ids);
542 spin_unlock_bh(&net->nsid_lock);
545 static LLIST_HEAD(cleanup_list);
547 static void cleanup_net(struct work_struct *work)
549 const struct pernet_operations *ops;
550 struct net *net, *tmp, *last;
551 struct llist_node *net_kill_list;
552 LIST_HEAD(net_exit_list);
554 /* Atomically snapshot the list of namespaces to cleanup */
555 net_kill_list = llist_del_all(&cleanup_list);
557 down_read(&pernet_ops_rwsem);
559 /* Don't let anyone else find us. */
560 down_write(&net_rwsem);
561 llist_for_each_entry(net, net_kill_list, cleanup_list)
562 list_del_rcu(&net->list);
563 /* Cache last net. After we unlock rtnl, no one new net
564 * added to net_namespace_list can assign nsid pointer
565 * to a net from net_kill_list (see peernet2id_alloc()).
566 * So, we skip them in unhash_nsid().
568 * Note, that unhash_nsid() does not delete nsid links
569 * between net_kill_list's nets, as they've already
570 * deleted from net_namespace_list. But, this would be
571 * useless anyway, as netns_ids are destroyed there.
573 last = list_last_entry(&net_namespace_list, struct net, list);
574 up_write(&net_rwsem);
576 llist_for_each_entry(net, net_kill_list, cleanup_list) {
577 unhash_nsid(net, last);
578 list_add_tail(&net->exit_list, &net_exit_list);
581 /* Run all of the network namespace pre_exit methods */
582 list_for_each_entry_reverse(ops, &pernet_list, list)
583 ops_pre_exit_list(ops, &net_exit_list);
586 * Another CPU might be rcu-iterating the list, wait for it.
587 * This needs to be before calling the exit() notifiers, so
588 * the rcu_barrier() below isn't sufficient alone.
589 * Also the pre_exit() and exit() methods need this barrier.
593 /* Run all of the network namespace exit methods */
594 list_for_each_entry_reverse(ops, &pernet_list, list)
595 ops_exit_list(ops, &net_exit_list);
597 /* Free the net generic variables */
598 list_for_each_entry_reverse(ops, &pernet_list, list)
599 ops_free_list(ops, &net_exit_list);
601 up_read(&pernet_ops_rwsem);
603 /* Ensure there are no outstanding rcu callbacks using this
608 /* Finally it is safe to free my network namespace structure */
609 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
610 list_del_init(&net->exit_list);
611 dec_net_namespaces(net->ucounts);
613 key_remove_domain(net->key_domain);
615 put_user_ns(net->user_ns);
621 * net_ns_barrier - wait until concurrent net_cleanup_work is done
623 * cleanup_net runs from work queue and will first remove namespaces
624 * from the global list, then run net exit functions.
626 * Call this in module exit path to make sure that all netns
627 * ->exit ops have been invoked before the function is removed.
629 void net_ns_barrier(void)
631 down_write(&pernet_ops_rwsem);
632 up_write(&pernet_ops_rwsem);
634 EXPORT_SYMBOL(net_ns_barrier);
636 static DECLARE_WORK(net_cleanup_work, cleanup_net);
638 void __put_net(struct net *net)
640 ref_tracker_dir_exit(&net->refcnt_tracker);
641 /* Cleanup the network namespace in process context */
642 if (llist_add(&net->cleanup_list, &cleanup_list))
643 queue_work(netns_wq, &net_cleanup_work);
645 EXPORT_SYMBOL_GPL(__put_net);
648 * get_net_ns - increment the refcount of the network namespace
649 * @ns: common namespace (net)
651 * Returns the net's common namespace.
653 struct ns_common *get_net_ns(struct ns_common *ns)
655 return &get_net(container_of(ns, struct net, ns))->ns;
657 EXPORT_SYMBOL_GPL(get_net_ns);
659 struct net *get_net_ns_by_fd(int fd)
662 struct ns_common *ns;
665 file = proc_ns_fget(fd);
667 return ERR_CAST(file);
669 ns = get_proc_ns(file_inode(file));
670 if (ns->ops == &netns_operations)
671 net = get_net(container_of(ns, struct net, ns));
673 net = ERR_PTR(-EINVAL);
678 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
681 struct net *get_net_ns_by_pid(pid_t pid)
683 struct task_struct *tsk;
686 /* Lookup the network namespace */
687 net = ERR_PTR(-ESRCH);
689 tsk = find_task_by_vpid(pid);
691 struct nsproxy *nsproxy;
693 nsproxy = tsk->nsproxy;
695 net = get_net(nsproxy->net_ns);
701 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
703 static __net_init int net_ns_net_init(struct net *net)
706 net->ns.ops = &netns_operations;
708 return ns_alloc_inum(&net->ns);
711 static __net_exit void net_ns_net_exit(struct net *net)
713 ns_free_inum(&net->ns);
716 static struct pernet_operations __net_initdata net_ns_ops = {
717 .init = net_ns_net_init,
718 .exit = net_ns_net_exit,
721 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
722 [NETNSA_NONE] = { .type = NLA_UNSPEC },
723 [NETNSA_NSID] = { .type = NLA_S32 },
724 [NETNSA_PID] = { .type = NLA_U32 },
725 [NETNSA_FD] = { .type = NLA_U32 },
726 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
729 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
730 struct netlink_ext_ack *extack)
732 struct net *net = sock_net(skb->sk);
733 struct nlattr *tb[NETNSA_MAX + 1];
738 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
739 NETNSA_MAX, rtnl_net_policy, extack);
742 if (!tb[NETNSA_NSID]) {
743 NL_SET_ERR_MSG(extack, "nsid is missing");
746 nsid = nla_get_s32(tb[NETNSA_NSID]);
748 if (tb[NETNSA_PID]) {
749 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
750 nla = tb[NETNSA_PID];
751 } else if (tb[NETNSA_FD]) {
752 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
755 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
759 NL_SET_BAD_ATTR(extack, nla);
760 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
761 return PTR_ERR(peer);
764 spin_lock_bh(&net->nsid_lock);
765 if (__peernet2id(net, peer) >= 0) {
766 spin_unlock_bh(&net->nsid_lock);
768 NL_SET_BAD_ATTR(extack, nla);
769 NL_SET_ERR_MSG(extack,
770 "Peer netns already has a nsid assigned");
774 err = alloc_netid(net, peer, nsid);
775 spin_unlock_bh(&net->nsid_lock);
777 rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
780 } else if (err == -ENOSPC && nsid >= 0) {
782 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
783 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
790 static int rtnl_net_get_size(void)
792 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
793 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
794 + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
798 struct net_fill_args {
808 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
810 struct nlmsghdr *nlh;
811 struct rtgenmsg *rth;
813 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
818 rth = nlmsg_data(nlh);
819 rth->rtgen_family = AF_UNSPEC;
821 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
822 goto nla_put_failure;
825 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
826 goto nla_put_failure;
832 nlmsg_cancel(skb, nlh);
836 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
837 const struct nlmsghdr *nlh,
839 struct netlink_ext_ack *extack)
843 if (!netlink_strict_get_check(skb))
844 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
845 tb, NETNSA_MAX, rtnl_net_policy,
848 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
849 NETNSA_MAX, rtnl_net_policy,
854 for (i = 0; i <= NETNSA_MAX; i++) {
862 case NETNSA_TARGET_NSID:
865 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
873 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
874 struct netlink_ext_ack *extack)
876 struct net *net = sock_net(skb->sk);
877 struct nlattr *tb[NETNSA_MAX + 1];
878 struct net_fill_args fillargs = {
879 .portid = NETLINK_CB(skb).portid,
880 .seq = nlh->nlmsg_seq,
883 struct net *peer, *target = net;
888 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
891 if (tb[NETNSA_PID]) {
892 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
893 nla = tb[NETNSA_PID];
894 } else if (tb[NETNSA_FD]) {
895 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
897 } else if (tb[NETNSA_NSID]) {
898 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
900 peer = ERR_PTR(-ENOENT);
901 nla = tb[NETNSA_NSID];
903 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
908 NL_SET_BAD_ATTR(extack, nla);
909 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
910 return PTR_ERR(peer);
913 if (tb[NETNSA_TARGET_NSID]) {
914 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
916 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
917 if (IS_ERR(target)) {
918 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
919 NL_SET_ERR_MSG(extack,
920 "Target netns reference is invalid");
921 err = PTR_ERR(target);
924 fillargs.add_ref = true;
925 fillargs.ref_nsid = peernet2id(net, peer);
928 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
934 fillargs.nsid = peernet2id(target, peer);
935 err = rtnl_net_fill(msg, &fillargs);
939 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
945 if (fillargs.add_ref)
951 struct rtnl_net_dump_cb {
955 struct net_fill_args fillargs;
960 /* Runs in RCU-critical section. */
961 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
963 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
966 if (net_cb->idx < net_cb->s_idx)
969 net_cb->fillargs.nsid = id;
970 if (net_cb->fillargs.add_ref)
971 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
972 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
981 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
982 struct rtnl_net_dump_cb *net_cb,
983 struct netlink_callback *cb)
985 struct netlink_ext_ack *extack = cb->extack;
986 struct nlattr *tb[NETNSA_MAX + 1];
989 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
990 NETNSA_MAX, rtnl_net_policy,
995 for (i = 0; i <= NETNSA_MAX; i++) {
999 if (i == NETNSA_TARGET_NSID) {
1002 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
1004 NL_SET_BAD_ATTR(extack, tb[i]);
1005 NL_SET_ERR_MSG(extack,
1006 "Invalid target network namespace id");
1007 return PTR_ERR(net);
1009 net_cb->fillargs.add_ref = true;
1010 net_cb->ref_net = net_cb->tgt_net;
1011 net_cb->tgt_net = net;
1013 NL_SET_BAD_ATTR(extack, tb[i]);
1014 NL_SET_ERR_MSG(extack,
1015 "Unsupported attribute in dump request");
1023 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
1025 struct rtnl_net_dump_cb net_cb = {
1026 .tgt_net = sock_net(skb->sk),
1029 .portid = NETLINK_CB(cb->skb).portid,
1030 .seq = cb->nlh->nlmsg_seq,
1031 .flags = NLM_F_MULTI,
1035 .s_idx = cb->args[0],
1039 if (cb->strict_check) {
1040 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1046 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1049 cb->args[0] = net_cb.idx;
1051 if (net_cb.fillargs.add_ref)
1052 put_net(net_cb.tgt_net);
1053 return err < 0 ? err : skb->len;
1056 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
1057 struct nlmsghdr *nlh, gfp_t gfp)
1059 struct net_fill_args fillargs = {
1061 .seq = nlh ? nlh->nlmsg_seq : 0,
1065 struct sk_buff *msg;
1068 msg = nlmsg_new(rtnl_net_get_size(), gfp);
1072 err = rtnl_net_fill(msg, &fillargs);
1076 rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, gfp);
1082 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1085 void __init net_ns_init(void)
1087 struct net_generic *ng;
1089 #ifdef CONFIG_NET_NS
1090 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1092 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1094 /* Create workqueue for cleanup */
1095 netns_wq = create_singlethread_workqueue("netns");
1097 panic("Could not create netns workq");
1100 ng = net_alloc_generic();
1102 panic("Could not allocate generic netns");
1104 rcu_assign_pointer(init_net.gen, ng);
1107 init_net.key_domain = &init_net_key_domain;
1109 down_write(&pernet_ops_rwsem);
1110 if (setup_net(&init_net, &init_user_ns))
1111 panic("Could not setup the initial network namespace");
1113 init_net_initialized = true;
1114 up_write(&pernet_ops_rwsem);
1116 if (register_pernet_subsys(&net_ns_ops))
1117 panic("Could not register network namespace subsystems");
1119 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1120 RTNL_FLAG_DOIT_UNLOCKED);
1121 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1122 RTNL_FLAG_DOIT_UNLOCKED);
1125 static void free_exit_list(struct pernet_operations *ops, struct list_head *net_exit_list)
1127 ops_pre_exit_list(ops, net_exit_list);
1129 ops_exit_list(ops, net_exit_list);
1130 ops_free_list(ops, net_exit_list);
1133 #ifdef CONFIG_NET_NS
1134 static int __register_pernet_operations(struct list_head *list,
1135 struct pernet_operations *ops)
1139 LIST_HEAD(net_exit_list);
1141 list_add_tail(&ops->list, list);
1142 if (ops->init || (ops->id && ops->size)) {
1143 /* We held write locked pernet_ops_rwsem, and parallel
1144 * setup_net() and cleanup_net() are not possible.
1147 struct mem_cgroup *old, *memcg;
1149 memcg = mem_cgroup_or_root(get_mem_cgroup_from_obj(net));
1150 old = set_active_memcg(memcg);
1151 error = ops_init(ops, net);
1152 set_active_memcg(old);
1153 mem_cgroup_put(memcg);
1156 list_add_tail(&net->exit_list, &net_exit_list);
1162 /* If I have an error cleanup all namespaces I initialized */
1163 list_del(&ops->list);
1164 free_exit_list(ops, &net_exit_list);
1168 static void __unregister_pernet_operations(struct pernet_operations *ops)
1171 LIST_HEAD(net_exit_list);
1173 list_del(&ops->list);
1174 /* See comment in __register_pernet_operations() */
1176 list_add_tail(&net->exit_list, &net_exit_list);
1178 free_exit_list(ops, &net_exit_list);
1183 static int __register_pernet_operations(struct list_head *list,
1184 struct pernet_operations *ops)
1186 if (!init_net_initialized) {
1187 list_add_tail(&ops->list, list);
1191 return ops_init(ops, &init_net);
1194 static void __unregister_pernet_operations(struct pernet_operations *ops)
1196 if (!init_net_initialized) {
1197 list_del(&ops->list);
1199 LIST_HEAD(net_exit_list);
1200 list_add(&init_net.exit_list, &net_exit_list);
1201 free_exit_list(ops, &net_exit_list);
1205 #endif /* CONFIG_NET_NS */
1207 static DEFINE_IDA(net_generic_ids);
1209 static int register_pernet_operations(struct list_head *list,
1210 struct pernet_operations *ops)
1215 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1220 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1222 error = __register_pernet_operations(list, ops);
1226 ida_free(&net_generic_ids, *ops->id);
1232 static void unregister_pernet_operations(struct pernet_operations *ops)
1234 __unregister_pernet_operations(ops);
1237 ida_free(&net_generic_ids, *ops->id);
1241 * register_pernet_subsys - register a network namespace subsystem
1242 * @ops: pernet operations structure for the subsystem
1244 * Register a subsystem which has init and exit functions
1245 * that are called when network namespaces are created and
1246 * destroyed respectively.
1248 * When registered all network namespace init functions are
1249 * called for every existing network namespace. Allowing kernel
1250 * modules to have a race free view of the set of network namespaces.
1252 * When a new network namespace is created all of the init
1253 * methods are called in the order in which they were registered.
1255 * When a network namespace is destroyed all of the exit methods
1256 * are called in the reverse of the order with which they were
1259 int register_pernet_subsys(struct pernet_operations *ops)
1262 down_write(&pernet_ops_rwsem);
1263 error = register_pernet_operations(first_device, ops);
1264 up_write(&pernet_ops_rwsem);
1267 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1270 * unregister_pernet_subsys - unregister a network namespace subsystem
1271 * @ops: pernet operations structure to manipulate
1273 * Remove the pernet operations structure from the list to be
1274 * used when network namespaces are created or destroyed. In
1275 * addition run the exit method for all existing network
1278 void unregister_pernet_subsys(struct pernet_operations *ops)
1280 down_write(&pernet_ops_rwsem);
1281 unregister_pernet_operations(ops);
1282 up_write(&pernet_ops_rwsem);
1284 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1287 * register_pernet_device - register a network namespace device
1288 * @ops: pernet operations structure for the subsystem
1290 * Register a device which has init and exit functions
1291 * that are called when network namespaces are created and
1292 * destroyed respectively.
1294 * When registered all network namespace init functions are
1295 * called for every existing network namespace. Allowing kernel
1296 * modules to have a race free view of the set of network namespaces.
1298 * When a new network namespace is created all of the init
1299 * methods are called in the order in which they were registered.
1301 * When a network namespace is destroyed all of the exit methods
1302 * are called in the reverse of the order with which they were
1305 int register_pernet_device(struct pernet_operations *ops)
1308 down_write(&pernet_ops_rwsem);
1309 error = register_pernet_operations(&pernet_list, ops);
1310 if (!error && (first_device == &pernet_list))
1311 first_device = &ops->list;
1312 up_write(&pernet_ops_rwsem);
1315 EXPORT_SYMBOL_GPL(register_pernet_device);
1318 * unregister_pernet_device - unregister a network namespace netdevice
1319 * @ops: pernet operations structure to manipulate
1321 * Remove the pernet operations structure from the list to be
1322 * used when network namespaces are created or destroyed. In
1323 * addition run the exit method for all existing network
1326 void unregister_pernet_device(struct pernet_operations *ops)
1328 down_write(&pernet_ops_rwsem);
1329 if (&ops->list == first_device)
1330 first_device = first_device->next;
1331 unregister_pernet_operations(ops);
1332 up_write(&pernet_ops_rwsem);
1334 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1336 #ifdef CONFIG_NET_NS
1337 static struct ns_common *netns_get(struct task_struct *task)
1339 struct net *net = NULL;
1340 struct nsproxy *nsproxy;
1343 nsproxy = task->nsproxy;
1345 net = get_net(nsproxy->net_ns);
1348 return net ? &net->ns : NULL;
1351 static inline struct net *to_net_ns(struct ns_common *ns)
1353 return container_of(ns, struct net, ns);
1356 static void netns_put(struct ns_common *ns)
1358 put_net(to_net_ns(ns));
1361 static int netns_install(struct nsset *nsset, struct ns_common *ns)
1363 struct nsproxy *nsproxy = nsset->nsproxy;
1364 struct net *net = to_net_ns(ns);
1366 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1367 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
1370 put_net(nsproxy->net_ns);
1371 nsproxy->net_ns = get_net(net);
1375 static struct user_namespace *netns_owner(struct ns_common *ns)
1377 return to_net_ns(ns)->user_ns;
1380 const struct proc_ns_operations netns_operations = {
1382 .type = CLONE_NEWNET,
1385 .install = netns_install,
1386 .owner = netns_owner,