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 static struct net_generic *net_alloc_generic(void)
77 struct net_generic *ng;
78 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
80 ng = kzalloc(generic_size, GFP_KERNEL);
82 ng->s.len = max_gen_ptrs;
87 static int net_assign_generic(struct net *net, unsigned int id, void *data)
89 struct net_generic *ng, *old_ng;
91 BUG_ON(id < MIN_PERNET_OPS_ID);
93 old_ng = rcu_dereference_protected(net->gen,
94 lockdep_is_held(&pernet_ops_rwsem));
95 if (old_ng->s.len > id) {
96 old_ng->ptr[id] = data;
100 ng = net_alloc_generic();
105 * Some synchronisation notes:
107 * The net_generic explores the net->gen array inside rcu
108 * read section. Besides once set the net->gen->ptr[x]
109 * pointer never changes (see rules in netns/generic.h).
111 * That said, we simply duplicate this array and schedule
112 * the old copy for kfree after a grace period.
115 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
116 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
119 rcu_assign_pointer(net->gen, ng);
120 kfree_rcu(old_ng, s.rcu);
124 static int ops_init(const struct pernet_operations *ops, struct net *net)
129 if (ops->id && ops->size) {
130 data = kzalloc(ops->size, GFP_KERNEL);
134 err = net_assign_generic(net, *ops->id, data);
140 err = ops->init(net);
151 static void ops_pre_exit_list(const struct pernet_operations *ops,
152 struct list_head *net_exit_list)
157 list_for_each_entry(net, net_exit_list, exit_list)
162 static void ops_exit_list(const struct pernet_operations *ops,
163 struct list_head *net_exit_list)
167 list_for_each_entry(net, net_exit_list, exit_list)
171 ops->exit_batch(net_exit_list);
174 static void ops_free_list(const struct pernet_operations *ops,
175 struct list_head *net_exit_list)
178 if (ops->size && ops->id) {
179 list_for_each_entry(net, net_exit_list, exit_list)
180 kfree(net_generic(net, *ops->id));
184 /* should be called with nsid_lock held */
185 static int alloc_netid(struct net *net, struct net *peer, int reqid)
187 int min = 0, max = 0;
194 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
197 /* This function is used by idr_for_each(). If net is equal to peer, the
198 * function returns the id so that idr_for_each() stops. Because we cannot
199 * returns the id 0 (idr_for_each() will not stop), we return the magic value
200 * NET_ID_ZERO (-1) for it.
202 #define NET_ID_ZERO -1
203 static int net_eq_idr(int id, void *net, void *peer)
205 if (net_eq(net, peer))
206 return id ? : NET_ID_ZERO;
210 /* Must be called from RCU-critical section or with nsid_lock held */
211 static int __peernet2id(const struct net *net, struct net *peer)
213 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
215 /* Magic value for id 0. */
216 if (id == NET_ID_ZERO)
221 return NETNSA_NSID_NOT_ASSIGNED;
224 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
225 struct nlmsghdr *nlh, gfp_t gfp);
226 /* This function returns the id of a peer netns. If no id is assigned, one will
227 * be allocated and returned.
229 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
233 if (refcount_read(&net->ns.count) == 0)
234 return NETNSA_NSID_NOT_ASSIGNED;
236 spin_lock_bh(&net->nsid_lock);
237 id = __peernet2id(net, peer);
239 spin_unlock_bh(&net->nsid_lock);
243 /* When peer is obtained from RCU lists, we may race with
244 * its cleanup. Check whether it's alive, and this guarantees
245 * we never hash a peer back to net->netns_ids, after it has
246 * just been idr_remove()'d from there in cleanup_net().
248 if (!maybe_get_net(peer)) {
249 spin_unlock_bh(&net->nsid_lock);
250 return NETNSA_NSID_NOT_ASSIGNED;
253 id = alloc_netid(net, peer, -1);
254 spin_unlock_bh(&net->nsid_lock);
258 return NETNSA_NSID_NOT_ASSIGNED;
260 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
264 EXPORT_SYMBOL_GPL(peernet2id_alloc);
266 /* This function returns, if assigned, the id of a peer netns. */
267 int peernet2id(const struct net *net, struct net *peer)
272 id = __peernet2id(net, peer);
277 EXPORT_SYMBOL(peernet2id);
279 /* This function returns true is the peer netns has an id assigned into the
282 bool peernet_has_id(const struct net *net, struct net *peer)
284 return peernet2id(net, peer) >= 0;
287 struct net *get_net_ns_by_id(const struct net *net, int id)
295 peer = idr_find(&net->netns_ids, id);
297 peer = maybe_get_net(peer);
304 * setup_net runs the initializers for the network namespace object.
306 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
308 /* Must be called with pernet_ops_rwsem held */
309 const struct pernet_operations *ops, *saved_ops;
311 LIST_HEAD(net_exit_list);
313 refcount_set(&net->ns.count, 1);
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;
365 static struct pernet_operations net_defaults_ops = {
366 .init = net_defaults_init_net,
369 static __init int net_defaults_init(void)
371 if (register_pernet_subsys(&net_defaults_ops))
372 panic("Cannot initialize net default settings");
377 core_initcall(net_defaults_init);
380 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
382 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
385 static void dec_net_namespaces(struct ucounts *ucounts)
387 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
390 static struct kmem_cache *net_cachep __ro_after_init;
391 static struct workqueue_struct *netns_wq;
393 static struct net *net_alloc(void)
395 struct net *net = NULL;
396 struct net_generic *ng;
398 ng = net_alloc_generic();
402 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
407 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
408 if (!net->key_domain)
410 refcount_set(&net->key_domain->usage, 1);
413 rcu_assign_pointer(net->gen, ng);
419 kmem_cache_free(net_cachep, net);
427 static void net_free(struct net *net)
429 if (refcount_dec_and_test(&net->passive)) {
430 kfree(rcu_access_pointer(net->gen));
431 kmem_cache_free(net_cachep, net);
435 void net_drop_ns(void *p)
437 struct net *net = (struct net *)p;
443 struct net *copy_net_ns(unsigned long flags,
444 struct user_namespace *user_ns, struct net *old_net)
446 struct ucounts *ucounts;
450 if (!(flags & CLONE_NEWNET))
451 return get_net(old_net);
453 ucounts = inc_net_namespaces(user_ns);
455 return ERR_PTR(-ENOSPC);
462 refcount_set(&net->passive, 1);
463 net->ucounts = ucounts;
464 get_user_ns(user_ns);
466 rv = down_read_killable(&pernet_ops_rwsem);
470 rv = setup_net(net, user_ns);
472 up_read(&pernet_ops_rwsem);
476 key_remove_domain(net->key_domain);
477 put_user_ns(user_ns);
480 dec_net_namespaces(ucounts);
487 * net_ns_get_ownership - get sysfs ownership data for @net
488 * @net: network namespace in question (can be NULL)
489 * @uid: kernel user ID for sysfs objects
490 * @gid: kernel group ID for sysfs objects
492 * Returns the uid/gid pair of root in the user namespace associated with the
493 * given network namespace.
495 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
498 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
499 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
501 if (uid_valid(ns_root_uid))
504 if (gid_valid(ns_root_gid))
507 *uid = GLOBAL_ROOT_UID;
508 *gid = GLOBAL_ROOT_GID;
511 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
513 static void unhash_nsid(struct net *net, struct net *last)
516 /* This function is only called from cleanup_net() work,
517 * and this work is the only process, that may delete
518 * a net from net_namespace_list. So, when the below
519 * is executing, the list may only grow. Thus, we do not
520 * use for_each_net_rcu() or net_rwsem.
525 spin_lock_bh(&tmp->nsid_lock);
526 id = __peernet2id(tmp, net);
528 idr_remove(&tmp->netns_ids, id);
529 spin_unlock_bh(&tmp->nsid_lock);
531 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
536 spin_lock_bh(&net->nsid_lock);
537 idr_destroy(&net->netns_ids);
538 spin_unlock_bh(&net->nsid_lock);
541 static LLIST_HEAD(cleanup_list);
543 static void cleanup_net(struct work_struct *work)
545 const struct pernet_operations *ops;
546 struct net *net, *tmp, *last;
547 struct llist_node *net_kill_list;
548 LIST_HEAD(net_exit_list);
550 /* Atomically snapshot the list of namespaces to cleanup */
551 net_kill_list = llist_del_all(&cleanup_list);
553 down_read(&pernet_ops_rwsem);
555 /* Don't let anyone else find us. */
556 down_write(&net_rwsem);
557 llist_for_each_entry(net, net_kill_list, cleanup_list)
558 list_del_rcu(&net->list);
559 /* Cache last net. After we unlock rtnl, no one new net
560 * added to net_namespace_list can assign nsid pointer
561 * to a net from net_kill_list (see peernet2id_alloc()).
562 * So, we skip them in unhash_nsid().
564 * Note, that unhash_nsid() does not delete nsid links
565 * between net_kill_list's nets, as they've already
566 * deleted from net_namespace_list. But, this would be
567 * useless anyway, as netns_ids are destroyed there.
569 last = list_last_entry(&net_namespace_list, struct net, list);
570 up_write(&net_rwsem);
572 llist_for_each_entry(net, net_kill_list, cleanup_list) {
573 unhash_nsid(net, last);
574 list_add_tail(&net->exit_list, &net_exit_list);
577 /* Run all of the network namespace pre_exit methods */
578 list_for_each_entry_reverse(ops, &pernet_list, list)
579 ops_pre_exit_list(ops, &net_exit_list);
582 * Another CPU might be rcu-iterating the list, wait for it.
583 * This needs to be before calling the exit() notifiers, so
584 * the rcu_barrier() below isn't sufficient alone.
585 * Also the pre_exit() and exit() methods need this barrier.
589 /* Run all of the network namespace exit methods */
590 list_for_each_entry_reverse(ops, &pernet_list, list)
591 ops_exit_list(ops, &net_exit_list);
593 /* Free the net generic variables */
594 list_for_each_entry_reverse(ops, &pernet_list, list)
595 ops_free_list(ops, &net_exit_list);
597 up_read(&pernet_ops_rwsem);
599 /* Ensure there are no outstanding rcu callbacks using this
604 /* Finally it is safe to free my network namespace structure */
605 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
606 list_del_init(&net->exit_list);
607 dec_net_namespaces(net->ucounts);
608 key_remove_domain(net->key_domain);
609 put_user_ns(net->user_ns);
615 * net_ns_barrier - wait until concurrent net_cleanup_work is done
617 * cleanup_net runs from work queue and will first remove namespaces
618 * from the global list, then run net exit functions.
620 * Call this in module exit path to make sure that all netns
621 * ->exit ops have been invoked before the function is removed.
623 void net_ns_barrier(void)
625 down_write(&pernet_ops_rwsem);
626 up_write(&pernet_ops_rwsem);
628 EXPORT_SYMBOL(net_ns_barrier);
630 static DECLARE_WORK(net_cleanup_work, cleanup_net);
632 void __put_net(struct net *net)
634 /* Cleanup the network namespace in process context */
635 if (llist_add(&net->cleanup_list, &cleanup_list))
636 queue_work(netns_wq, &net_cleanup_work);
638 EXPORT_SYMBOL_GPL(__put_net);
641 * get_net_ns - increment the refcount of the network namespace
642 * @ns: common namespace (net)
644 * Returns the net's common namespace.
646 struct ns_common *get_net_ns(struct ns_common *ns)
648 return &get_net(container_of(ns, struct net, ns))->ns;
650 EXPORT_SYMBOL_GPL(get_net_ns);
652 struct net *get_net_ns_by_fd(int fd)
655 struct ns_common *ns;
658 file = proc_ns_fget(fd);
660 return ERR_CAST(file);
662 ns = get_proc_ns(file_inode(file));
663 if (ns->ops == &netns_operations)
664 net = get_net(container_of(ns, struct net, ns));
666 net = ERR_PTR(-EINVAL);
671 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
674 struct net *get_net_ns_by_pid(pid_t pid)
676 struct task_struct *tsk;
679 /* Lookup the network namespace */
680 net = ERR_PTR(-ESRCH);
682 tsk = find_task_by_vpid(pid);
684 struct nsproxy *nsproxy;
686 nsproxy = tsk->nsproxy;
688 net = get_net(nsproxy->net_ns);
694 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
696 static __net_init int net_ns_net_init(struct net *net)
699 net->ns.ops = &netns_operations;
701 return ns_alloc_inum(&net->ns);
704 static __net_exit void net_ns_net_exit(struct net *net)
706 ns_free_inum(&net->ns);
709 static struct pernet_operations __net_initdata net_ns_ops = {
710 .init = net_ns_net_init,
711 .exit = net_ns_net_exit,
714 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
715 [NETNSA_NONE] = { .type = NLA_UNSPEC },
716 [NETNSA_NSID] = { .type = NLA_S32 },
717 [NETNSA_PID] = { .type = NLA_U32 },
718 [NETNSA_FD] = { .type = NLA_U32 },
719 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
722 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
723 struct netlink_ext_ack *extack)
725 struct net *net = sock_net(skb->sk);
726 struct nlattr *tb[NETNSA_MAX + 1];
731 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
732 NETNSA_MAX, rtnl_net_policy, extack);
735 if (!tb[NETNSA_NSID]) {
736 NL_SET_ERR_MSG(extack, "nsid is missing");
739 nsid = nla_get_s32(tb[NETNSA_NSID]);
741 if (tb[NETNSA_PID]) {
742 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
743 nla = tb[NETNSA_PID];
744 } else if (tb[NETNSA_FD]) {
745 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
748 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
752 NL_SET_BAD_ATTR(extack, nla);
753 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
754 return PTR_ERR(peer);
757 spin_lock_bh(&net->nsid_lock);
758 if (__peernet2id(net, peer) >= 0) {
759 spin_unlock_bh(&net->nsid_lock);
761 NL_SET_BAD_ATTR(extack, nla);
762 NL_SET_ERR_MSG(extack,
763 "Peer netns already has a nsid assigned");
767 err = alloc_netid(net, peer, nsid);
768 spin_unlock_bh(&net->nsid_lock);
770 rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
773 } else if (err == -ENOSPC && nsid >= 0) {
775 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
776 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
783 static int rtnl_net_get_size(void)
785 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
786 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
787 + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
791 struct net_fill_args {
801 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
803 struct nlmsghdr *nlh;
804 struct rtgenmsg *rth;
806 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
811 rth = nlmsg_data(nlh);
812 rth->rtgen_family = AF_UNSPEC;
814 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
815 goto nla_put_failure;
818 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
819 goto nla_put_failure;
825 nlmsg_cancel(skb, nlh);
829 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
830 const struct nlmsghdr *nlh,
832 struct netlink_ext_ack *extack)
836 if (!netlink_strict_get_check(skb))
837 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
838 tb, NETNSA_MAX, rtnl_net_policy,
841 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
842 NETNSA_MAX, rtnl_net_policy,
847 for (i = 0; i <= NETNSA_MAX; i++) {
855 case NETNSA_TARGET_NSID:
858 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
866 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
867 struct netlink_ext_ack *extack)
869 struct net *net = sock_net(skb->sk);
870 struct nlattr *tb[NETNSA_MAX + 1];
871 struct net_fill_args fillargs = {
872 .portid = NETLINK_CB(skb).portid,
873 .seq = nlh->nlmsg_seq,
876 struct net *peer, *target = net;
881 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
884 if (tb[NETNSA_PID]) {
885 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
886 nla = tb[NETNSA_PID];
887 } else if (tb[NETNSA_FD]) {
888 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
890 } else if (tb[NETNSA_NSID]) {
891 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
893 peer = ERR_PTR(-ENOENT);
894 nla = tb[NETNSA_NSID];
896 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
901 NL_SET_BAD_ATTR(extack, nla);
902 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
903 return PTR_ERR(peer);
906 if (tb[NETNSA_TARGET_NSID]) {
907 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
909 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
910 if (IS_ERR(target)) {
911 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
912 NL_SET_ERR_MSG(extack,
913 "Target netns reference is invalid");
914 err = PTR_ERR(target);
917 fillargs.add_ref = true;
918 fillargs.ref_nsid = peernet2id(net, peer);
921 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
927 fillargs.nsid = peernet2id(target, peer);
928 err = rtnl_net_fill(msg, &fillargs);
932 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
938 if (fillargs.add_ref)
944 struct rtnl_net_dump_cb {
948 struct net_fill_args fillargs;
953 /* Runs in RCU-critical section. */
954 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
956 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
959 if (net_cb->idx < net_cb->s_idx)
962 net_cb->fillargs.nsid = id;
963 if (net_cb->fillargs.add_ref)
964 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
965 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
974 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
975 struct rtnl_net_dump_cb *net_cb,
976 struct netlink_callback *cb)
978 struct netlink_ext_ack *extack = cb->extack;
979 struct nlattr *tb[NETNSA_MAX + 1];
982 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
983 NETNSA_MAX, rtnl_net_policy,
988 for (i = 0; i <= NETNSA_MAX; i++) {
992 if (i == NETNSA_TARGET_NSID) {
995 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
997 NL_SET_BAD_ATTR(extack, tb[i]);
998 NL_SET_ERR_MSG(extack,
999 "Invalid target network namespace id");
1000 return PTR_ERR(net);
1002 net_cb->fillargs.add_ref = true;
1003 net_cb->ref_net = net_cb->tgt_net;
1004 net_cb->tgt_net = net;
1006 NL_SET_BAD_ATTR(extack, tb[i]);
1007 NL_SET_ERR_MSG(extack,
1008 "Unsupported attribute in dump request");
1016 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
1018 struct rtnl_net_dump_cb net_cb = {
1019 .tgt_net = sock_net(skb->sk),
1022 .portid = NETLINK_CB(cb->skb).portid,
1023 .seq = cb->nlh->nlmsg_seq,
1024 .flags = NLM_F_MULTI,
1028 .s_idx = cb->args[0],
1032 if (cb->strict_check) {
1033 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1039 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1042 cb->args[0] = net_cb.idx;
1044 if (net_cb.fillargs.add_ref)
1045 put_net(net_cb.tgt_net);
1046 return err < 0 ? err : skb->len;
1049 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
1050 struct nlmsghdr *nlh, gfp_t gfp)
1052 struct net_fill_args fillargs = {
1054 .seq = nlh ? nlh->nlmsg_seq : 0,
1058 struct sk_buff *msg;
1061 msg = nlmsg_new(rtnl_net_get_size(), gfp);
1065 err = rtnl_net_fill(msg, &fillargs);
1069 rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, gfp);
1075 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1078 static int __init net_ns_init(void)
1080 struct net_generic *ng;
1082 #ifdef CONFIG_NET_NS
1083 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1085 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1087 /* Create workqueue for cleanup */
1088 netns_wq = create_singlethread_workqueue("netns");
1090 panic("Could not create netns workq");
1093 ng = net_alloc_generic();
1095 panic("Could not allocate generic netns");
1097 rcu_assign_pointer(init_net.gen, ng);
1099 down_write(&pernet_ops_rwsem);
1100 if (setup_net(&init_net, &init_user_ns))
1101 panic("Could not setup the initial network namespace");
1103 init_net_initialized = true;
1104 up_write(&pernet_ops_rwsem);
1106 if (register_pernet_subsys(&net_ns_ops))
1107 panic("Could not register network namespace subsystems");
1109 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1110 RTNL_FLAG_DOIT_UNLOCKED);
1111 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1112 RTNL_FLAG_DOIT_UNLOCKED);
1117 pure_initcall(net_ns_init);
1119 static void free_exit_list(struct pernet_operations *ops, struct list_head *net_exit_list)
1121 ops_pre_exit_list(ops, net_exit_list);
1123 ops_exit_list(ops, net_exit_list);
1124 ops_free_list(ops, net_exit_list);
1127 #ifdef CONFIG_NET_NS
1128 static int __register_pernet_operations(struct list_head *list,
1129 struct pernet_operations *ops)
1133 LIST_HEAD(net_exit_list);
1135 list_add_tail(&ops->list, list);
1136 if (ops->init || (ops->id && ops->size)) {
1137 /* We held write locked pernet_ops_rwsem, and parallel
1138 * setup_net() and cleanup_net() are not possible.
1141 error = ops_init(ops, net);
1144 list_add_tail(&net->exit_list, &net_exit_list);
1150 /* If I have an error cleanup all namespaces I initialized */
1151 list_del(&ops->list);
1152 free_exit_list(ops, &net_exit_list);
1156 static void __unregister_pernet_operations(struct pernet_operations *ops)
1159 LIST_HEAD(net_exit_list);
1161 list_del(&ops->list);
1162 /* See comment in __register_pernet_operations() */
1164 list_add_tail(&net->exit_list, &net_exit_list);
1166 free_exit_list(ops, &net_exit_list);
1171 static int __register_pernet_operations(struct list_head *list,
1172 struct pernet_operations *ops)
1174 if (!init_net_initialized) {
1175 list_add_tail(&ops->list, list);
1179 return ops_init(ops, &init_net);
1182 static void __unregister_pernet_operations(struct pernet_operations *ops)
1184 if (!init_net_initialized) {
1185 list_del(&ops->list);
1187 LIST_HEAD(net_exit_list);
1188 list_add(&init_net.exit_list, &net_exit_list);
1189 free_exit_list(ops, &net_exit_list);
1193 #endif /* CONFIG_NET_NS */
1195 static DEFINE_IDA(net_generic_ids);
1197 static int register_pernet_operations(struct list_head *list,
1198 struct pernet_operations *ops)
1203 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1208 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1210 error = __register_pernet_operations(list, ops);
1214 ida_free(&net_generic_ids, *ops->id);
1220 static void unregister_pernet_operations(struct pernet_operations *ops)
1222 __unregister_pernet_operations(ops);
1225 ida_free(&net_generic_ids, *ops->id);
1229 * register_pernet_subsys - register a network namespace subsystem
1230 * @ops: pernet operations structure for the subsystem
1232 * Register a subsystem which has init and exit functions
1233 * that are called when network namespaces are created and
1234 * destroyed respectively.
1236 * When registered all network namespace init functions are
1237 * called for every existing network namespace. Allowing kernel
1238 * modules to have a race free view of the set of network namespaces.
1240 * When a new network namespace is created all of the init
1241 * methods are called in the order in which they were registered.
1243 * When a network namespace is destroyed all of the exit methods
1244 * are called in the reverse of the order with which they were
1247 int register_pernet_subsys(struct pernet_operations *ops)
1250 down_write(&pernet_ops_rwsem);
1251 error = register_pernet_operations(first_device, ops);
1252 up_write(&pernet_ops_rwsem);
1255 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1258 * unregister_pernet_subsys - unregister a network namespace subsystem
1259 * @ops: pernet operations structure to manipulate
1261 * Remove the pernet operations structure from the list to be
1262 * used when network namespaces are created or destroyed. In
1263 * addition run the exit method for all existing network
1266 void unregister_pernet_subsys(struct pernet_operations *ops)
1268 down_write(&pernet_ops_rwsem);
1269 unregister_pernet_operations(ops);
1270 up_write(&pernet_ops_rwsem);
1272 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1275 * register_pernet_device - register a network namespace device
1276 * @ops: pernet operations structure for the subsystem
1278 * Register a device which has init and exit functions
1279 * that are called when network namespaces are created and
1280 * destroyed respectively.
1282 * When registered all network namespace init functions are
1283 * called for every existing network namespace. Allowing kernel
1284 * modules to have a race free view of the set of network namespaces.
1286 * When a new network namespace is created all of the init
1287 * methods are called in the order in which they were registered.
1289 * When a network namespace is destroyed all of the exit methods
1290 * are called in the reverse of the order with which they were
1293 int register_pernet_device(struct pernet_operations *ops)
1296 down_write(&pernet_ops_rwsem);
1297 error = register_pernet_operations(&pernet_list, ops);
1298 if (!error && (first_device == &pernet_list))
1299 first_device = &ops->list;
1300 up_write(&pernet_ops_rwsem);
1303 EXPORT_SYMBOL_GPL(register_pernet_device);
1306 * unregister_pernet_device - unregister a network namespace netdevice
1307 * @ops: pernet operations structure to manipulate
1309 * Remove the pernet operations structure from the list to be
1310 * used when network namespaces are created or destroyed. In
1311 * addition run the exit method for all existing network
1314 void unregister_pernet_device(struct pernet_operations *ops)
1316 down_write(&pernet_ops_rwsem);
1317 if (&ops->list == first_device)
1318 first_device = first_device->next;
1319 unregister_pernet_operations(ops);
1320 up_write(&pernet_ops_rwsem);
1322 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1324 #ifdef CONFIG_NET_NS
1325 static struct ns_common *netns_get(struct task_struct *task)
1327 struct net *net = NULL;
1328 struct nsproxy *nsproxy;
1331 nsproxy = task->nsproxy;
1333 net = get_net(nsproxy->net_ns);
1336 return net ? &net->ns : NULL;
1339 static inline struct net *to_net_ns(struct ns_common *ns)
1341 return container_of(ns, struct net, ns);
1344 static void netns_put(struct ns_common *ns)
1346 put_net(to_net_ns(ns));
1349 static int netns_install(struct nsset *nsset, struct ns_common *ns)
1351 struct nsproxy *nsproxy = nsset->nsproxy;
1352 struct net *net = to_net_ns(ns);
1354 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1355 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
1358 put_net(nsproxy->net_ns);
1359 nsproxy->net_ns = get_net(net);
1363 static struct user_namespace *netns_owner(struct ns_common *ns)
1365 return to_net_ns(ns)->user_ns;
1368 const struct proc_ns_operations netns_operations = {
1370 .type = CLONE_NEWNET,
1373 .install = netns_install,
1374 .owner = netns_owner,