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>
22 #include <linux/cookie.h>
25 #include <net/netlink.h>
26 #include <net/net_namespace.h>
27 #include <net/netns/generic.h>
30 * Our network namespace constructor/destructor lists
33 static LIST_HEAD(pernet_list);
34 static struct list_head *first_device = &pernet_list;
36 LIST_HEAD(net_namespace_list);
37 EXPORT_SYMBOL_GPL(net_namespace_list);
39 /* Protects net_namespace_list. Nests iside rtnl_lock() */
40 DECLARE_RWSEM(net_rwsem);
41 EXPORT_SYMBOL_GPL(net_rwsem);
44 static struct key_tag init_net_key_domain = { .usage = REFCOUNT_INIT(1) };
47 struct net init_net = {
48 .ns.count = REFCOUNT_INIT(1),
49 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
51 .key_domain = &init_net_key_domain,
54 EXPORT_SYMBOL(init_net);
56 static bool init_net_initialized;
58 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
59 * init_net_initialized and first_device pointer.
60 * This is internal net namespace object. Please, don't use it
63 DECLARE_RWSEM(pernet_ops_rwsem);
64 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
66 #define MIN_PERNET_OPS_ID \
67 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
69 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
71 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
73 DEFINE_COOKIE(net_cookie);
75 u64 __net_gen_cookie(struct net *net)
78 u64 res = atomic64_read(&net->net_cookie);
82 res = gen_cookie_next(&net_cookie);
83 atomic64_cmpxchg(&net->net_cookie, 0, res);
87 static struct net_generic *net_alloc_generic(void)
89 struct net_generic *ng;
90 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
92 ng = kzalloc(generic_size, GFP_KERNEL);
94 ng->s.len = max_gen_ptrs;
99 static int net_assign_generic(struct net *net, unsigned int id, void *data)
101 struct net_generic *ng, *old_ng;
103 BUG_ON(id < MIN_PERNET_OPS_ID);
105 old_ng = rcu_dereference_protected(net->gen,
106 lockdep_is_held(&pernet_ops_rwsem));
107 if (old_ng->s.len > id) {
108 old_ng->ptr[id] = data;
112 ng = net_alloc_generic();
117 * Some synchronisation notes:
119 * The net_generic explores the net->gen array inside rcu
120 * read section. Besides once set the net->gen->ptr[x]
121 * pointer never changes (see rules in netns/generic.h).
123 * That said, we simply duplicate this array and schedule
124 * the old copy for kfree after a grace period.
127 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
128 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
131 rcu_assign_pointer(net->gen, ng);
132 kfree_rcu(old_ng, s.rcu);
136 static int ops_init(const struct pernet_operations *ops, struct net *net)
141 if (ops->id && ops->size) {
142 data = kzalloc(ops->size, GFP_KERNEL);
146 err = net_assign_generic(net, *ops->id, data);
152 err = ops->init(net);
163 static void ops_free(const struct pernet_operations *ops, struct net *net)
165 if (ops->id && ops->size) {
166 kfree(net_generic(net, *ops->id));
170 static void ops_pre_exit_list(const struct pernet_operations *ops,
171 struct list_head *net_exit_list)
176 list_for_each_entry(net, net_exit_list, exit_list)
181 static void ops_exit_list(const struct pernet_operations *ops,
182 struct list_head *net_exit_list)
186 list_for_each_entry(net, net_exit_list, exit_list)
190 ops->exit_batch(net_exit_list);
193 static void ops_free_list(const struct pernet_operations *ops,
194 struct list_head *net_exit_list)
197 if (ops->size && ops->id) {
198 list_for_each_entry(net, net_exit_list, exit_list)
203 /* should be called with nsid_lock held */
204 static int alloc_netid(struct net *net, struct net *peer, int reqid)
206 int min = 0, max = 0;
213 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
216 /* This function is used by idr_for_each(). If net is equal to peer, the
217 * function returns the id so that idr_for_each() stops. Because we cannot
218 * returns the id 0 (idr_for_each() will not stop), we return the magic value
219 * NET_ID_ZERO (-1) for it.
221 #define NET_ID_ZERO -1
222 static int net_eq_idr(int id, void *net, void *peer)
224 if (net_eq(net, peer))
225 return id ? : NET_ID_ZERO;
229 /* Must be called from RCU-critical section or with nsid_lock held */
230 static int __peernet2id(const struct net *net, struct net *peer)
232 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
234 /* Magic value for id 0. */
235 if (id == NET_ID_ZERO)
240 return NETNSA_NSID_NOT_ASSIGNED;
243 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
244 struct nlmsghdr *nlh, gfp_t gfp);
245 /* This function returns the id of a peer netns. If no id is assigned, one will
246 * be allocated and returned.
248 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
252 if (refcount_read(&net->ns.count) == 0)
253 return NETNSA_NSID_NOT_ASSIGNED;
255 spin_lock_bh(&net->nsid_lock);
256 id = __peernet2id(net, peer);
258 spin_unlock_bh(&net->nsid_lock);
262 /* When peer is obtained from RCU lists, we may race with
263 * its cleanup. Check whether it's alive, and this guarantees
264 * we never hash a peer back to net->netns_ids, after it has
265 * just been idr_remove()'d from there in cleanup_net().
267 if (!maybe_get_net(peer)) {
268 spin_unlock_bh(&net->nsid_lock);
269 return NETNSA_NSID_NOT_ASSIGNED;
272 id = alloc_netid(net, peer, -1);
273 spin_unlock_bh(&net->nsid_lock);
277 return NETNSA_NSID_NOT_ASSIGNED;
279 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
283 EXPORT_SYMBOL_GPL(peernet2id_alloc);
285 /* This function returns, if assigned, the id of a peer netns. */
286 int peernet2id(const struct net *net, struct net *peer)
291 id = __peernet2id(net, peer);
296 EXPORT_SYMBOL(peernet2id);
298 /* This function returns true is the peer netns has an id assigned into the
301 bool peernet_has_id(const struct net *net, struct net *peer)
303 return peernet2id(net, peer) >= 0;
306 struct net *get_net_ns_by_id(const struct net *net, int id)
314 peer = idr_find(&net->netns_ids, id);
316 peer = maybe_get_net(peer);
323 * setup_net runs the initializers for the network namespace object.
325 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
327 /* Must be called with pernet_ops_rwsem held */
328 const struct pernet_operations *ops, *saved_ops;
330 LIST_HEAD(net_exit_list);
332 refcount_set(&net->ns.count, 1);
333 refcount_set(&net->passive, 1);
334 get_random_bytes(&net->hash_mix, sizeof(u32));
335 net->dev_base_seq = 1;
336 net->user_ns = user_ns;
337 idr_init(&net->netns_ids);
338 spin_lock_init(&net->nsid_lock);
339 mutex_init(&net->ipv4.ra_mutex);
341 list_for_each_entry(ops, &pernet_list, list) {
342 error = ops_init(ops, net);
346 down_write(&net_rwsem);
347 list_add_tail_rcu(&net->list, &net_namespace_list);
348 up_write(&net_rwsem);
353 /* Walk through the list backwards calling the exit functions
354 * for the pernet modules whose init functions did not fail.
356 list_add(&net->exit_list, &net_exit_list);
358 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
359 ops_pre_exit_list(ops, &net_exit_list);
364 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
365 ops_exit_list(ops, &net_exit_list);
368 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
369 ops_free_list(ops, &net_exit_list);
375 static int __net_init net_defaults_init_net(struct net *net)
377 net->core.sysctl_somaxconn = SOMAXCONN;
381 static struct pernet_operations net_defaults_ops = {
382 .init = net_defaults_init_net,
385 static __init int net_defaults_init(void)
387 if (register_pernet_subsys(&net_defaults_ops))
388 panic("Cannot initialize net default settings");
393 core_initcall(net_defaults_init);
396 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
398 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
401 static void dec_net_namespaces(struct ucounts *ucounts)
403 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
406 static struct kmem_cache *net_cachep __ro_after_init;
407 static struct workqueue_struct *netns_wq;
409 static struct net *net_alloc(void)
411 struct net *net = NULL;
412 struct net_generic *ng;
414 ng = net_alloc_generic();
418 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
423 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
424 if (!net->key_domain)
426 refcount_set(&net->key_domain->usage, 1);
429 rcu_assign_pointer(net->gen, ng);
435 kmem_cache_free(net_cachep, net);
443 static void net_free(struct net *net)
445 kfree(rcu_access_pointer(net->gen));
446 kmem_cache_free(net_cachep, net);
449 void net_drop_ns(void *p)
452 if (ns && refcount_dec_and_test(&ns->passive))
456 struct net *copy_net_ns(unsigned long flags,
457 struct user_namespace *user_ns, struct net *old_net)
459 struct ucounts *ucounts;
463 if (!(flags & CLONE_NEWNET))
464 return get_net(old_net);
466 ucounts = inc_net_namespaces(user_ns);
468 return ERR_PTR(-ENOSPC);
475 refcount_set(&net->passive, 1);
476 net->ucounts = ucounts;
477 get_user_ns(user_ns);
479 rv = down_read_killable(&pernet_ops_rwsem);
483 rv = setup_net(net, user_ns);
485 up_read(&pernet_ops_rwsem);
489 key_remove_domain(net->key_domain);
490 put_user_ns(user_ns);
493 dec_net_namespaces(ucounts);
500 * net_ns_get_ownership - get sysfs ownership data for @net
501 * @net: network namespace in question (can be NULL)
502 * @uid: kernel user ID for sysfs objects
503 * @gid: kernel group ID for sysfs objects
505 * Returns the uid/gid pair of root in the user namespace associated with the
506 * given network namespace.
508 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
511 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
512 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
514 if (uid_valid(ns_root_uid))
517 if (gid_valid(ns_root_gid))
520 *uid = GLOBAL_ROOT_UID;
521 *gid = GLOBAL_ROOT_GID;
524 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
526 static void unhash_nsid(struct net *net, struct net *last)
529 /* This function is only called from cleanup_net() work,
530 * and this work is the only process, that may delete
531 * a net from net_namespace_list. So, when the below
532 * is executing, the list may only grow. Thus, we do not
533 * use for_each_net_rcu() or net_rwsem.
538 spin_lock_bh(&tmp->nsid_lock);
539 id = __peernet2id(tmp, net);
541 idr_remove(&tmp->netns_ids, id);
542 spin_unlock_bh(&tmp->nsid_lock);
544 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
549 spin_lock_bh(&net->nsid_lock);
550 idr_destroy(&net->netns_ids);
551 spin_unlock_bh(&net->nsid_lock);
554 static LLIST_HEAD(cleanup_list);
556 static void cleanup_net(struct work_struct *work)
558 const struct pernet_operations *ops;
559 struct net *net, *tmp, *last;
560 struct llist_node *net_kill_list;
561 LIST_HEAD(net_exit_list);
563 /* Atomically snapshot the list of namespaces to cleanup */
564 net_kill_list = llist_del_all(&cleanup_list);
566 down_read(&pernet_ops_rwsem);
568 /* Don't let anyone else find us. */
569 down_write(&net_rwsem);
570 llist_for_each_entry(net, net_kill_list, cleanup_list)
571 list_del_rcu(&net->list);
572 /* Cache last net. After we unlock rtnl, no one new net
573 * added to net_namespace_list can assign nsid pointer
574 * to a net from net_kill_list (see peernet2id_alloc()).
575 * So, we skip them in unhash_nsid().
577 * Note, that unhash_nsid() does not delete nsid links
578 * between net_kill_list's nets, as they've already
579 * deleted from net_namespace_list. But, this would be
580 * useless anyway, as netns_ids are destroyed there.
582 last = list_last_entry(&net_namespace_list, struct net, list);
583 up_write(&net_rwsem);
585 llist_for_each_entry(net, net_kill_list, cleanup_list) {
586 unhash_nsid(net, last);
587 list_add_tail(&net->exit_list, &net_exit_list);
590 /* Run all of the network namespace pre_exit methods */
591 list_for_each_entry_reverse(ops, &pernet_list, list)
592 ops_pre_exit_list(ops, &net_exit_list);
595 * Another CPU might be rcu-iterating the list, wait for it.
596 * This needs to be before calling the exit() notifiers, so
597 * the rcu_barrier() below isn't sufficient alone.
598 * Also the pre_exit() and exit() methods need this barrier.
602 /* Run all of the network namespace exit methods */
603 list_for_each_entry_reverse(ops, &pernet_list, list)
604 ops_exit_list(ops, &net_exit_list);
606 /* Free the net generic variables */
607 list_for_each_entry_reverse(ops, &pernet_list, list)
608 ops_free_list(ops, &net_exit_list);
610 up_read(&pernet_ops_rwsem);
612 /* Ensure there are no outstanding rcu callbacks using this
617 /* Finally it is safe to free my network namespace structure */
618 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
619 list_del_init(&net->exit_list);
620 dec_net_namespaces(net->ucounts);
621 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);
653 struct net *get_net_ns_by_fd(int fd)
656 struct ns_common *ns;
659 file = proc_ns_fget(fd);
661 return ERR_CAST(file);
663 ns = get_proc_ns(file_inode(file));
664 if (ns->ops == &netns_operations)
665 net = get_net(container_of(ns, struct net, ns));
667 net = ERR_PTR(-EINVAL);
674 struct net *get_net_ns_by_fd(int fd)
676 return ERR_PTR(-EINVAL);
679 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 static int __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 __net_gen_cookie(&init_net);
1110 down_write(&pernet_ops_rwsem);
1111 if (setup_net(&init_net, &init_user_ns))
1112 panic("Could not setup the initial network namespace");
1114 init_net_initialized = true;
1115 up_write(&pernet_ops_rwsem);
1117 if (register_pernet_subsys(&net_ns_ops))
1118 panic("Could not register network namespace subsystems");
1120 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1121 RTNL_FLAG_DOIT_UNLOCKED);
1122 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1123 RTNL_FLAG_DOIT_UNLOCKED);
1128 pure_initcall(net_ns_init);
1130 #ifdef CONFIG_NET_NS
1131 static int __register_pernet_operations(struct list_head *list,
1132 struct pernet_operations *ops)
1136 LIST_HEAD(net_exit_list);
1138 list_add_tail(&ops->list, list);
1139 if (ops->init || (ops->id && ops->size)) {
1140 /* We held write locked pernet_ops_rwsem, and parallel
1141 * setup_net() and cleanup_net() are not possible.
1144 error = ops_init(ops, net);
1147 list_add_tail(&net->exit_list, &net_exit_list);
1153 /* If I have an error cleanup all namespaces I initialized */
1154 list_del(&ops->list);
1155 ops_pre_exit_list(ops, &net_exit_list);
1157 ops_exit_list(ops, &net_exit_list);
1158 ops_free_list(ops, &net_exit_list);
1162 static void __unregister_pernet_operations(struct pernet_operations *ops)
1165 LIST_HEAD(net_exit_list);
1167 list_del(&ops->list);
1168 /* See comment in __register_pernet_operations() */
1170 list_add_tail(&net->exit_list, &net_exit_list);
1171 ops_pre_exit_list(ops, &net_exit_list);
1173 ops_exit_list(ops, &net_exit_list);
1174 ops_free_list(ops, &net_exit_list);
1179 static int __register_pernet_operations(struct list_head *list,
1180 struct pernet_operations *ops)
1182 if (!init_net_initialized) {
1183 list_add_tail(&ops->list, list);
1187 return ops_init(ops, &init_net);
1190 static void __unregister_pernet_operations(struct pernet_operations *ops)
1192 if (!init_net_initialized) {
1193 list_del(&ops->list);
1195 LIST_HEAD(net_exit_list);
1196 list_add(&init_net.exit_list, &net_exit_list);
1197 ops_pre_exit_list(ops, &net_exit_list);
1199 ops_exit_list(ops, &net_exit_list);
1200 ops_free_list(ops, &net_exit_list);
1204 #endif /* CONFIG_NET_NS */
1206 static DEFINE_IDA(net_generic_ids);
1208 static int register_pernet_operations(struct list_head *list,
1209 struct pernet_operations *ops)
1214 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1219 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1221 error = __register_pernet_operations(list, ops);
1225 ida_free(&net_generic_ids, *ops->id);
1231 static void unregister_pernet_operations(struct pernet_operations *ops)
1233 __unregister_pernet_operations(ops);
1236 ida_free(&net_generic_ids, *ops->id);
1240 * register_pernet_subsys - register a network namespace subsystem
1241 * @ops: pernet operations structure for the subsystem
1243 * Register a subsystem which has init and exit functions
1244 * that are called when network namespaces are created and
1245 * destroyed respectively.
1247 * When registered all network namespace init functions are
1248 * called for every existing network namespace. Allowing kernel
1249 * modules to have a race free view of the set of network namespaces.
1251 * When a new network namespace is created all of the init
1252 * methods are called in the order in which they were registered.
1254 * When a network namespace is destroyed all of the exit methods
1255 * are called in the reverse of the order with which they were
1258 int register_pernet_subsys(struct pernet_operations *ops)
1261 down_write(&pernet_ops_rwsem);
1262 error = register_pernet_operations(first_device, ops);
1263 up_write(&pernet_ops_rwsem);
1266 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1269 * unregister_pernet_subsys - unregister a network namespace subsystem
1270 * @ops: pernet operations structure to manipulate
1272 * Remove the pernet operations structure from the list to be
1273 * used when network namespaces are created or destroyed. In
1274 * addition run the exit method for all existing network
1277 void unregister_pernet_subsys(struct pernet_operations *ops)
1279 down_write(&pernet_ops_rwsem);
1280 unregister_pernet_operations(ops);
1281 up_write(&pernet_ops_rwsem);
1283 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1286 * register_pernet_device - register a network namespace device
1287 * @ops: pernet operations structure for the subsystem
1289 * Register a device which has init and exit functions
1290 * that are called when network namespaces are created and
1291 * destroyed respectively.
1293 * When registered all network namespace init functions are
1294 * called for every existing network namespace. Allowing kernel
1295 * modules to have a race free view of the set of network namespaces.
1297 * When a new network namespace is created all of the init
1298 * methods are called in the order in which they were registered.
1300 * When a network namespace is destroyed all of the exit methods
1301 * are called in the reverse of the order with which they were
1304 int register_pernet_device(struct pernet_operations *ops)
1307 down_write(&pernet_ops_rwsem);
1308 error = register_pernet_operations(&pernet_list, ops);
1309 if (!error && (first_device == &pernet_list))
1310 first_device = &ops->list;
1311 up_write(&pernet_ops_rwsem);
1314 EXPORT_SYMBOL_GPL(register_pernet_device);
1317 * unregister_pernet_device - unregister a network namespace netdevice
1318 * @ops: pernet operations structure to manipulate
1320 * Remove the pernet operations structure from the list to be
1321 * used when network namespaces are created or destroyed. In
1322 * addition run the exit method for all existing network
1325 void unregister_pernet_device(struct pernet_operations *ops)
1327 down_write(&pernet_ops_rwsem);
1328 if (&ops->list == first_device)
1329 first_device = first_device->next;
1330 unregister_pernet_operations(ops);
1331 up_write(&pernet_ops_rwsem);
1333 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1335 #ifdef CONFIG_NET_NS
1336 static struct ns_common *netns_get(struct task_struct *task)
1338 struct net *net = NULL;
1339 struct nsproxy *nsproxy;
1342 nsproxy = task->nsproxy;
1344 net = get_net(nsproxy->net_ns);
1347 return net ? &net->ns : NULL;
1350 static inline struct net *to_net_ns(struct ns_common *ns)
1352 return container_of(ns, struct net, ns);
1355 static void netns_put(struct ns_common *ns)
1357 put_net(to_net_ns(ns));
1360 static int netns_install(struct nsset *nsset, struct ns_common *ns)
1362 struct nsproxy *nsproxy = nsset->nsproxy;
1363 struct net *net = to_net_ns(ns);
1365 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1366 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
1369 put_net(nsproxy->net_ns);
1370 nsproxy->net_ns = get_net(net);
1374 static struct user_namespace *netns_owner(struct ns_common *ns)
1376 return to_net_ns(ns)->user_ns;
1379 const struct proc_ns_operations netns_operations = {
1381 .type = CLONE_NEWNET,
1384 .install = netns_install,
1385 .owner = netns_owner,