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);
42 struct net init_net = {
43 .count = REFCOUNT_INIT(1),
44 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
46 EXPORT_SYMBOL(init_net);
48 static bool init_net_initialized;
50 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
51 * init_net_initialized and first_device pointer.
52 * This is internal net namespace object. Please, don't use it
55 DECLARE_RWSEM(pernet_ops_rwsem);
56 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
58 #define MIN_PERNET_OPS_ID \
59 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
61 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
63 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
65 static struct net_generic *net_alloc_generic(void)
67 struct net_generic *ng;
68 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
70 ng = kzalloc(generic_size, GFP_KERNEL);
72 ng->s.len = max_gen_ptrs;
77 static int net_assign_generic(struct net *net, unsigned int id, void *data)
79 struct net_generic *ng, *old_ng;
81 BUG_ON(id < MIN_PERNET_OPS_ID);
83 old_ng = rcu_dereference_protected(net->gen,
84 lockdep_is_held(&pernet_ops_rwsem));
85 if (old_ng->s.len > id) {
86 old_ng->ptr[id] = data;
90 ng = net_alloc_generic();
95 * Some synchronisation notes:
97 * The net_generic explores the net->gen array inside rcu
98 * read section. Besides once set the net->gen->ptr[x]
99 * pointer never changes (see rules in netns/generic.h).
101 * That said, we simply duplicate this array and schedule
102 * the old copy for kfree after a grace period.
105 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
106 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
109 rcu_assign_pointer(net->gen, ng);
110 kfree_rcu(old_ng, s.rcu);
114 static int ops_init(const struct pernet_operations *ops, struct net *net)
119 if (ops->id && ops->size) {
120 data = kzalloc(ops->size, GFP_KERNEL);
124 err = net_assign_generic(net, *ops->id, data);
130 err = ops->init(net);
141 static void ops_free(const struct pernet_operations *ops, struct net *net)
143 if (ops->id && ops->size) {
144 kfree(net_generic(net, *ops->id));
148 static void ops_pre_exit_list(const struct pernet_operations *ops,
149 struct list_head *net_exit_list)
154 list_for_each_entry(net, net_exit_list, exit_list)
159 static void ops_exit_list(const struct pernet_operations *ops,
160 struct list_head *net_exit_list)
164 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)
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 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
208 * is set to true, thus the caller knows that the new id must be notified via
211 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
213 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
214 bool alloc_it = *alloc;
218 /* Magic value for id 0. */
219 if (id == NET_ID_ZERO)
225 id = alloc_netid(net, peer, -1);
227 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
230 return NETNSA_NSID_NOT_ASSIGNED;
233 /* should be called with nsid_lock held */
234 static int __peernet2id(struct net *net, struct net *peer)
238 return __peernet2id_alloc(net, peer, &no);
241 static void rtnl_net_notifyid(struct net *net, int cmd, int id);
242 /* This function returns the id of a peer netns. If no id is assigned, one will
243 * be allocated and returned.
245 int peernet2id_alloc(struct net *net, struct net *peer)
247 bool alloc = false, alive = false;
250 if (refcount_read(&net->count) == 0)
251 return NETNSA_NSID_NOT_ASSIGNED;
252 spin_lock_bh(&net->nsid_lock);
254 * When peer is obtained from RCU lists, we may race with
255 * its cleanup. Check whether it's alive, and this guarantees
256 * we never hash a peer back to net->netns_ids, after it has
257 * just been idr_remove()'d from there in cleanup_net().
259 if (maybe_get_net(peer))
260 alive = alloc = true;
261 id = __peernet2id_alloc(net, peer, &alloc);
262 spin_unlock_bh(&net->nsid_lock);
263 if (alloc && id >= 0)
264 rtnl_net_notifyid(net, RTM_NEWNSID, id);
269 EXPORT_SYMBOL_GPL(peernet2id_alloc);
271 /* This function returns, if assigned, the id of a peer netns. */
272 int peernet2id(struct net *net, struct net *peer)
276 spin_lock_bh(&net->nsid_lock);
277 id = __peernet2id(net, peer);
278 spin_unlock_bh(&net->nsid_lock);
281 EXPORT_SYMBOL(peernet2id);
283 /* This function returns true is the peer netns has an id assigned into the
286 bool peernet_has_id(struct net *net, struct net *peer)
288 return peernet2id(net, peer) >= 0;
291 struct net *get_net_ns_by_id(struct net *net, int id)
299 peer = idr_find(&net->netns_ids, id);
301 peer = maybe_get_net(peer);
308 * setup_net runs the initializers for the network namespace object.
310 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
312 /* Must be called with pernet_ops_rwsem held */
313 const struct pernet_operations *ops, *saved_ops;
315 LIST_HEAD(net_exit_list);
317 refcount_set(&net->count, 1);
318 refcount_set(&net->passive, 1);
319 get_random_bytes(&net->hash_mix, sizeof(u32));
320 net->dev_base_seq = 1;
321 net->user_ns = user_ns;
322 idr_init(&net->netns_ids);
323 spin_lock_init(&net->nsid_lock);
324 mutex_init(&net->ipv4.ra_mutex);
326 list_for_each_entry(ops, &pernet_list, list) {
327 error = ops_init(ops, net);
331 down_write(&net_rwsem);
332 list_add_tail_rcu(&net->list, &net_namespace_list);
333 up_write(&net_rwsem);
338 /* Walk through the list backwards calling the exit functions
339 * for the pernet modules whose init functions did not fail.
341 list_add(&net->exit_list, &net_exit_list);
343 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
344 ops_pre_exit_list(ops, &net_exit_list);
349 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
350 ops_exit_list(ops, &net_exit_list);
353 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
354 ops_free_list(ops, &net_exit_list);
360 static int __net_init net_defaults_init_net(struct net *net)
362 net->core.sysctl_somaxconn = SOMAXCONN;
366 static struct pernet_operations net_defaults_ops = {
367 .init = net_defaults_init_net,
370 static __init int net_defaults_init(void)
372 if (register_pernet_subsys(&net_defaults_ops))
373 panic("Cannot initialize net default settings");
378 core_initcall(net_defaults_init);
381 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
383 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
386 static void dec_net_namespaces(struct ucounts *ucounts)
388 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
391 static struct kmem_cache *net_cachep __ro_after_init;
392 static struct workqueue_struct *netns_wq;
394 static struct net *net_alloc(void)
396 struct net *net = NULL;
397 struct net_generic *ng;
399 ng = net_alloc_generic();
403 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
407 rcu_assign_pointer(net->gen, ng);
416 static void net_free(struct net *net)
418 kfree(rcu_access_pointer(net->gen));
419 kmem_cache_free(net_cachep, net);
422 void net_drop_ns(void *p)
425 if (ns && refcount_dec_and_test(&ns->passive))
429 struct net *copy_net_ns(unsigned long flags,
430 struct user_namespace *user_ns, struct net *old_net)
432 struct ucounts *ucounts;
436 if (!(flags & CLONE_NEWNET))
437 return get_net(old_net);
439 ucounts = inc_net_namespaces(user_ns);
441 return ERR_PTR(-ENOSPC);
448 refcount_set(&net->passive, 1);
449 net->ucounts = ucounts;
450 get_user_ns(user_ns);
452 rv = down_read_killable(&pernet_ops_rwsem);
456 rv = setup_net(net, user_ns);
458 up_read(&pernet_ops_rwsem);
462 put_user_ns(user_ns);
465 dec_net_namespaces(ucounts);
472 * net_ns_get_ownership - get sysfs ownership data for @net
473 * @net: network namespace in question (can be NULL)
474 * @uid: kernel user ID for sysfs objects
475 * @gid: kernel group ID for sysfs objects
477 * Returns the uid/gid pair of root in the user namespace associated with the
478 * given network namespace.
480 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
483 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
484 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
486 if (uid_valid(ns_root_uid))
489 if (gid_valid(ns_root_gid))
492 *uid = GLOBAL_ROOT_UID;
493 *gid = GLOBAL_ROOT_GID;
496 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
498 static void unhash_nsid(struct net *net, struct net *last)
501 /* This function is only called from cleanup_net() work,
502 * and this work is the only process, that may delete
503 * a net from net_namespace_list. So, when the below
504 * is executing, the list may only grow. Thus, we do not
505 * use for_each_net_rcu() or net_rwsem.
510 spin_lock_bh(&tmp->nsid_lock);
511 id = __peernet2id(tmp, net);
513 idr_remove(&tmp->netns_ids, id);
514 spin_unlock_bh(&tmp->nsid_lock);
516 rtnl_net_notifyid(tmp, RTM_DELNSID, id);
520 spin_lock_bh(&net->nsid_lock);
521 idr_destroy(&net->netns_ids);
522 spin_unlock_bh(&net->nsid_lock);
525 static LLIST_HEAD(cleanup_list);
527 static void cleanup_net(struct work_struct *work)
529 const struct pernet_operations *ops;
530 struct net *net, *tmp, *last;
531 struct llist_node *net_kill_list;
532 LIST_HEAD(net_exit_list);
534 /* Atomically snapshot the list of namespaces to cleanup */
535 net_kill_list = llist_del_all(&cleanup_list);
537 down_read(&pernet_ops_rwsem);
539 /* Don't let anyone else find us. */
540 down_write(&net_rwsem);
541 llist_for_each_entry(net, net_kill_list, cleanup_list)
542 list_del_rcu(&net->list);
543 /* Cache last net. After we unlock rtnl, no one new net
544 * added to net_namespace_list can assign nsid pointer
545 * to a net from net_kill_list (see peernet2id_alloc()).
546 * So, we skip them in unhash_nsid().
548 * Note, that unhash_nsid() does not delete nsid links
549 * between net_kill_list's nets, as they've already
550 * deleted from net_namespace_list. But, this would be
551 * useless anyway, as netns_ids are destroyed there.
553 last = list_last_entry(&net_namespace_list, struct net, list);
554 up_write(&net_rwsem);
556 llist_for_each_entry(net, net_kill_list, cleanup_list) {
557 unhash_nsid(net, last);
558 list_add_tail(&net->exit_list, &net_exit_list);
561 /* Run all of the network namespace pre_exit methods */
562 list_for_each_entry_reverse(ops, &pernet_list, list)
563 ops_pre_exit_list(ops, &net_exit_list);
566 * Another CPU might be rcu-iterating the list, wait for it.
567 * This needs to be before calling the exit() notifiers, so
568 * the rcu_barrier() below isn't sufficient alone.
569 * Also the pre_exit() and exit() methods need this barrier.
573 /* Run all of the network namespace exit methods */
574 list_for_each_entry_reverse(ops, &pernet_list, list)
575 ops_exit_list(ops, &net_exit_list);
577 /* Free the net generic variables */
578 list_for_each_entry_reverse(ops, &pernet_list, list)
579 ops_free_list(ops, &net_exit_list);
581 up_read(&pernet_ops_rwsem);
583 /* Ensure there are no outstanding rcu callbacks using this
588 /* Finally it is safe to free my network namespace structure */
589 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
590 list_del_init(&net->exit_list);
591 dec_net_namespaces(net->ucounts);
592 put_user_ns(net->user_ns);
598 * net_ns_barrier - wait until concurrent net_cleanup_work is done
600 * cleanup_net runs from work queue and will first remove namespaces
601 * from the global list, then run net exit functions.
603 * Call this in module exit path to make sure that all netns
604 * ->exit ops have been invoked before the function is removed.
606 void net_ns_barrier(void)
608 down_write(&pernet_ops_rwsem);
609 up_write(&pernet_ops_rwsem);
611 EXPORT_SYMBOL(net_ns_barrier);
613 static DECLARE_WORK(net_cleanup_work, cleanup_net);
615 void __put_net(struct net *net)
617 /* Cleanup the network namespace in process context */
618 if (llist_add(&net->cleanup_list, &cleanup_list))
619 queue_work(netns_wq, &net_cleanup_work);
621 EXPORT_SYMBOL_GPL(__put_net);
623 struct net *get_net_ns_by_fd(int fd)
626 struct ns_common *ns;
629 file = proc_ns_fget(fd);
631 return ERR_CAST(file);
633 ns = get_proc_ns(file_inode(file));
634 if (ns->ops == &netns_operations)
635 net = get_net(container_of(ns, struct net, ns));
637 net = ERR_PTR(-EINVAL);
644 struct net *get_net_ns_by_fd(int fd)
646 return ERR_PTR(-EINVAL);
649 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
651 struct net *get_net_ns_by_pid(pid_t pid)
653 struct task_struct *tsk;
656 /* Lookup the network namespace */
657 net = ERR_PTR(-ESRCH);
659 tsk = find_task_by_vpid(pid);
661 struct nsproxy *nsproxy;
663 nsproxy = tsk->nsproxy;
665 net = get_net(nsproxy->net_ns);
671 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
673 static __net_init int net_ns_net_init(struct net *net)
676 net->ns.ops = &netns_operations;
678 return ns_alloc_inum(&net->ns);
681 static __net_exit void net_ns_net_exit(struct net *net)
683 ns_free_inum(&net->ns);
686 static struct pernet_operations __net_initdata net_ns_ops = {
687 .init = net_ns_net_init,
688 .exit = net_ns_net_exit,
691 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
692 [NETNSA_NONE] = { .type = NLA_UNSPEC },
693 [NETNSA_NSID] = { .type = NLA_S32 },
694 [NETNSA_PID] = { .type = NLA_U32 },
695 [NETNSA_FD] = { .type = NLA_U32 },
696 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
699 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
700 struct netlink_ext_ack *extack)
702 struct net *net = sock_net(skb->sk);
703 struct nlattr *tb[NETNSA_MAX + 1];
708 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
709 NETNSA_MAX, rtnl_net_policy, extack);
712 if (!tb[NETNSA_NSID]) {
713 NL_SET_ERR_MSG(extack, "nsid is missing");
716 nsid = nla_get_s32(tb[NETNSA_NSID]);
718 if (tb[NETNSA_PID]) {
719 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
720 nla = tb[NETNSA_PID];
721 } else if (tb[NETNSA_FD]) {
722 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
725 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
729 NL_SET_BAD_ATTR(extack, nla);
730 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
731 return PTR_ERR(peer);
734 spin_lock_bh(&net->nsid_lock);
735 if (__peernet2id(net, peer) >= 0) {
736 spin_unlock_bh(&net->nsid_lock);
738 NL_SET_BAD_ATTR(extack, nla);
739 NL_SET_ERR_MSG(extack,
740 "Peer netns already has a nsid assigned");
744 err = alloc_netid(net, peer, nsid);
745 spin_unlock_bh(&net->nsid_lock);
747 rtnl_net_notifyid(net, RTM_NEWNSID, err);
749 } else if (err == -ENOSPC && nsid >= 0) {
751 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
752 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
759 static int rtnl_net_get_size(void)
761 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
762 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
763 + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
767 struct net_fill_args {
777 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
779 struct nlmsghdr *nlh;
780 struct rtgenmsg *rth;
782 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
787 rth = nlmsg_data(nlh);
788 rth->rtgen_family = AF_UNSPEC;
790 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
791 goto nla_put_failure;
794 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
795 goto nla_put_failure;
801 nlmsg_cancel(skb, nlh);
805 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
806 const struct nlmsghdr *nlh,
808 struct netlink_ext_ack *extack)
812 if (!netlink_strict_get_check(skb))
813 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
814 tb, NETNSA_MAX, rtnl_net_policy,
817 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
818 NETNSA_MAX, rtnl_net_policy,
823 for (i = 0; i <= NETNSA_MAX; i++) {
831 case NETNSA_TARGET_NSID:
834 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
842 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
843 struct netlink_ext_ack *extack)
845 struct net *net = sock_net(skb->sk);
846 struct nlattr *tb[NETNSA_MAX + 1];
847 struct net_fill_args fillargs = {
848 .portid = NETLINK_CB(skb).portid,
849 .seq = nlh->nlmsg_seq,
852 struct net *peer, *target = net;
857 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
860 if (tb[NETNSA_PID]) {
861 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
862 nla = tb[NETNSA_PID];
863 } else if (tb[NETNSA_FD]) {
864 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
866 } else if (tb[NETNSA_NSID]) {
867 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
869 peer = ERR_PTR(-ENOENT);
870 nla = tb[NETNSA_NSID];
872 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
877 NL_SET_BAD_ATTR(extack, nla);
878 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
879 return PTR_ERR(peer);
882 if (tb[NETNSA_TARGET_NSID]) {
883 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
885 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
886 if (IS_ERR(target)) {
887 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
888 NL_SET_ERR_MSG(extack,
889 "Target netns reference is invalid");
890 err = PTR_ERR(target);
893 fillargs.add_ref = true;
894 fillargs.ref_nsid = peernet2id(net, peer);
897 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
903 fillargs.nsid = peernet2id(target, peer);
904 err = rtnl_net_fill(msg, &fillargs);
908 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
914 if (fillargs.add_ref)
920 struct rtnl_net_dump_cb {
924 struct net_fill_args fillargs;
929 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
931 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
934 if (net_cb->idx < net_cb->s_idx)
937 net_cb->fillargs.nsid = id;
938 if (net_cb->fillargs.add_ref)
939 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
940 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
949 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
950 struct rtnl_net_dump_cb *net_cb,
951 struct netlink_callback *cb)
953 struct netlink_ext_ack *extack = cb->extack;
954 struct nlattr *tb[NETNSA_MAX + 1];
957 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
958 NETNSA_MAX, rtnl_net_policy,
963 for (i = 0; i <= NETNSA_MAX; i++) {
967 if (i == NETNSA_TARGET_NSID) {
970 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
972 NL_SET_BAD_ATTR(extack, tb[i]);
973 NL_SET_ERR_MSG(extack,
974 "Invalid target network namespace id");
977 net_cb->fillargs.add_ref = true;
978 net_cb->ref_net = net_cb->tgt_net;
979 net_cb->tgt_net = net;
981 NL_SET_BAD_ATTR(extack, tb[i]);
982 NL_SET_ERR_MSG(extack,
983 "Unsupported attribute in dump request");
991 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
993 struct rtnl_net_dump_cb net_cb = {
994 .tgt_net = sock_net(skb->sk),
997 .portid = NETLINK_CB(cb->skb).portid,
998 .seq = cb->nlh->nlmsg_seq,
999 .flags = NLM_F_MULTI,
1003 .s_idx = cb->args[0],
1007 if (cb->strict_check) {
1008 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1013 spin_lock_bh(&net_cb.tgt_net->nsid_lock);
1014 if (net_cb.fillargs.add_ref &&
1015 !net_eq(net_cb.ref_net, net_cb.tgt_net) &&
1016 !spin_trylock_bh(&net_cb.ref_net->nsid_lock)) {
1017 spin_unlock_bh(&net_cb.tgt_net->nsid_lock);
1021 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1022 if (net_cb.fillargs.add_ref &&
1023 !net_eq(net_cb.ref_net, net_cb.tgt_net))
1024 spin_unlock_bh(&net_cb.ref_net->nsid_lock);
1025 spin_unlock_bh(&net_cb.tgt_net->nsid_lock);
1027 cb->args[0] = net_cb.idx;
1029 if (net_cb.fillargs.add_ref)
1030 put_net(net_cb.tgt_net);
1031 return err < 0 ? err : skb->len;
1034 static void rtnl_net_notifyid(struct net *net, int cmd, int id)
1036 struct net_fill_args fillargs = {
1040 struct sk_buff *msg;
1043 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
1047 err = rtnl_net_fill(msg, &fillargs);
1051 rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
1057 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1060 static int __init net_ns_init(void)
1062 struct net_generic *ng;
1064 #ifdef CONFIG_NET_NS
1065 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1067 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1069 /* Create workqueue for cleanup */
1070 netns_wq = create_singlethread_workqueue("netns");
1072 panic("Could not create netns workq");
1075 ng = net_alloc_generic();
1077 panic("Could not allocate generic netns");
1079 rcu_assign_pointer(init_net.gen, ng);
1081 down_write(&pernet_ops_rwsem);
1082 if (setup_net(&init_net, &init_user_ns))
1083 panic("Could not setup the initial network namespace");
1085 init_net_initialized = true;
1086 up_write(&pernet_ops_rwsem);
1088 if (register_pernet_subsys(&net_ns_ops))
1089 panic("Could not register network namespace subsystems");
1091 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1092 RTNL_FLAG_DOIT_UNLOCKED);
1093 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1094 RTNL_FLAG_DOIT_UNLOCKED);
1099 pure_initcall(net_ns_init);
1101 #ifdef CONFIG_NET_NS
1102 static int __register_pernet_operations(struct list_head *list,
1103 struct pernet_operations *ops)
1107 LIST_HEAD(net_exit_list);
1109 list_add_tail(&ops->list, list);
1110 if (ops->init || (ops->id && ops->size)) {
1111 /* We held write locked pernet_ops_rwsem, and parallel
1112 * setup_net() and cleanup_net() are not possible.
1115 error = ops_init(ops, net);
1118 list_add_tail(&net->exit_list, &net_exit_list);
1124 /* If I have an error cleanup all namespaces I initialized */
1125 list_del(&ops->list);
1126 ops_pre_exit_list(ops, &net_exit_list);
1128 ops_exit_list(ops, &net_exit_list);
1129 ops_free_list(ops, &net_exit_list);
1133 static void __unregister_pernet_operations(struct pernet_operations *ops)
1136 LIST_HEAD(net_exit_list);
1138 list_del(&ops->list);
1139 /* See comment in __register_pernet_operations() */
1141 list_add_tail(&net->exit_list, &net_exit_list);
1142 ops_pre_exit_list(ops, &net_exit_list);
1144 ops_exit_list(ops, &net_exit_list);
1145 ops_free_list(ops, &net_exit_list);
1150 static int __register_pernet_operations(struct list_head *list,
1151 struct pernet_operations *ops)
1153 if (!init_net_initialized) {
1154 list_add_tail(&ops->list, list);
1158 return ops_init(ops, &init_net);
1161 static void __unregister_pernet_operations(struct pernet_operations *ops)
1163 if (!init_net_initialized) {
1164 list_del(&ops->list);
1166 LIST_HEAD(net_exit_list);
1167 list_add(&init_net.exit_list, &net_exit_list);
1168 ops_pre_exit_list(ops, &net_exit_list);
1170 ops_exit_list(ops, &net_exit_list);
1171 ops_free_list(ops, &net_exit_list);
1175 #endif /* CONFIG_NET_NS */
1177 static DEFINE_IDA(net_generic_ids);
1179 static int register_pernet_operations(struct list_head *list,
1180 struct pernet_operations *ops)
1185 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1190 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1192 error = __register_pernet_operations(list, ops);
1196 ida_free(&net_generic_ids, *ops->id);
1202 static void unregister_pernet_operations(struct pernet_operations *ops)
1204 __unregister_pernet_operations(ops);
1207 ida_free(&net_generic_ids, *ops->id);
1211 * register_pernet_subsys - register a network namespace subsystem
1212 * @ops: pernet operations structure for the subsystem
1214 * Register a subsystem which has init and exit functions
1215 * that are called when network namespaces are created and
1216 * destroyed respectively.
1218 * When registered all network namespace init functions are
1219 * called for every existing network namespace. Allowing kernel
1220 * modules to have a race free view of the set of network namespaces.
1222 * When a new network namespace is created all of the init
1223 * methods are called in the order in which they were registered.
1225 * When a network namespace is destroyed all of the exit methods
1226 * are called in the reverse of the order with which they were
1229 int register_pernet_subsys(struct pernet_operations *ops)
1232 down_write(&pernet_ops_rwsem);
1233 error = register_pernet_operations(first_device, ops);
1234 up_write(&pernet_ops_rwsem);
1237 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1240 * unregister_pernet_subsys - unregister a network namespace subsystem
1241 * @ops: pernet operations structure to manipulate
1243 * Remove the pernet operations structure from the list to be
1244 * used when network namespaces are created or destroyed. In
1245 * addition run the exit method for all existing network
1248 void unregister_pernet_subsys(struct pernet_operations *ops)
1250 down_write(&pernet_ops_rwsem);
1251 unregister_pernet_operations(ops);
1252 up_write(&pernet_ops_rwsem);
1254 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1257 * register_pernet_device - register a network namespace device
1258 * @ops: pernet operations structure for the subsystem
1260 * Register a device which has init and exit functions
1261 * that are called when network namespaces are created and
1262 * destroyed respectively.
1264 * When registered all network namespace init functions are
1265 * called for every existing network namespace. Allowing kernel
1266 * modules to have a race free view of the set of network namespaces.
1268 * When a new network namespace is created all of the init
1269 * methods are called in the order in which they were registered.
1271 * When a network namespace is destroyed all of the exit methods
1272 * are called in the reverse of the order with which they were
1275 int register_pernet_device(struct pernet_operations *ops)
1278 down_write(&pernet_ops_rwsem);
1279 error = register_pernet_operations(&pernet_list, ops);
1280 if (!error && (first_device == &pernet_list))
1281 first_device = &ops->list;
1282 up_write(&pernet_ops_rwsem);
1285 EXPORT_SYMBOL_GPL(register_pernet_device);
1288 * unregister_pernet_device - unregister a network namespace netdevice
1289 * @ops: pernet operations structure to manipulate
1291 * Remove the pernet operations structure from the list to be
1292 * used when network namespaces are created or destroyed. In
1293 * addition run the exit method for all existing network
1296 void unregister_pernet_device(struct pernet_operations *ops)
1298 down_write(&pernet_ops_rwsem);
1299 if (&ops->list == first_device)
1300 first_device = first_device->next;
1301 unregister_pernet_operations(ops);
1302 up_write(&pernet_ops_rwsem);
1304 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1306 #ifdef CONFIG_NET_NS
1307 static struct ns_common *netns_get(struct task_struct *task)
1309 struct net *net = NULL;
1310 struct nsproxy *nsproxy;
1313 nsproxy = task->nsproxy;
1315 net = get_net(nsproxy->net_ns);
1318 return net ? &net->ns : NULL;
1321 static inline struct net *to_net_ns(struct ns_common *ns)
1323 return container_of(ns, struct net, ns);
1326 static void netns_put(struct ns_common *ns)
1328 put_net(to_net_ns(ns));
1331 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1333 struct net *net = to_net_ns(ns);
1335 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1336 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1339 put_net(nsproxy->net_ns);
1340 nsproxy->net_ns = get_net(net);
1344 static struct user_namespace *netns_owner(struct ns_common *ns)
1346 return to_net_ns(ns)->user_ns;
1349 const struct proc_ns_operations netns_operations = {
1351 .type = CLONE_NEWNET,
1354 .install = netns_install,
1355 .owner = netns_owner,