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) };
48 EXPORT_SYMBOL(init_net);
50 static bool init_net_initialized;
52 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
53 * init_net_initialized and first_device pointer.
54 * This is internal net namespace object. Please, don't use it
57 DECLARE_RWSEM(pernet_ops_rwsem);
58 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
60 #define MIN_PERNET_OPS_ID \
61 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
63 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
65 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
67 DEFINE_COOKIE(net_cookie);
69 static struct net_generic *net_alloc_generic(void)
71 struct net_generic *ng;
72 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
74 ng = kzalloc(generic_size, GFP_KERNEL);
76 ng->s.len = max_gen_ptrs;
81 static int net_assign_generic(struct net *net, unsigned int id, void *data)
83 struct net_generic *ng, *old_ng;
85 BUG_ON(id < MIN_PERNET_OPS_ID);
87 old_ng = rcu_dereference_protected(net->gen,
88 lockdep_is_held(&pernet_ops_rwsem));
89 if (old_ng->s.len > id) {
90 old_ng->ptr[id] = data;
94 ng = net_alloc_generic();
99 * Some synchronisation notes:
101 * The net_generic explores the net->gen array inside rcu
102 * read section. Besides once set the net->gen->ptr[x]
103 * pointer never changes (see rules in netns/generic.h).
105 * That said, we simply duplicate this array and schedule
106 * the old copy for kfree after a grace period.
109 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
110 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
113 rcu_assign_pointer(net->gen, ng);
114 kfree_rcu(old_ng, s.rcu);
118 static int ops_init(const struct pernet_operations *ops, struct net *net)
123 if (ops->id && ops->size) {
124 data = kzalloc(ops->size, GFP_KERNEL);
128 err = net_assign_generic(net, *ops->id, data);
134 err = ops->init(net);
145 static void ops_pre_exit_list(const struct pernet_operations *ops,
146 struct list_head *net_exit_list)
151 list_for_each_entry(net, net_exit_list, exit_list)
156 static void ops_exit_list(const struct pernet_operations *ops,
157 struct list_head *net_exit_list)
161 list_for_each_entry(net, net_exit_list, exit_list) {
167 ops->exit_batch(net_exit_list);
170 static void ops_free_list(const struct pernet_operations *ops,
171 struct list_head *net_exit_list)
174 if (ops->size && ops->id) {
175 list_for_each_entry(net, net_exit_list, exit_list)
176 kfree(net_generic(net, *ops->id));
180 /* should be called with nsid_lock held */
181 static int alloc_netid(struct net *net, struct net *peer, int reqid)
183 int min = 0, max = 0;
190 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
193 /* This function is used by idr_for_each(). If net is equal to peer, the
194 * function returns the id so that idr_for_each() stops. Because we cannot
195 * returns the id 0 (idr_for_each() will not stop), we return the magic value
196 * NET_ID_ZERO (-1) for it.
198 #define NET_ID_ZERO -1
199 static int net_eq_idr(int id, void *net, void *peer)
201 if (net_eq(net, peer))
202 return id ? : NET_ID_ZERO;
206 /* Must be called from RCU-critical section or with nsid_lock held */
207 static int __peernet2id(const struct net *net, struct net *peer)
209 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
211 /* Magic value for id 0. */
212 if (id == NET_ID_ZERO)
217 return NETNSA_NSID_NOT_ASSIGNED;
220 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
221 struct nlmsghdr *nlh, gfp_t gfp);
222 /* This function returns the id of a peer netns. If no id is assigned, one will
223 * be allocated and returned.
225 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
229 if (refcount_read(&net->ns.count) == 0)
230 return NETNSA_NSID_NOT_ASSIGNED;
232 spin_lock_bh(&net->nsid_lock);
233 id = __peernet2id(net, peer);
235 spin_unlock_bh(&net->nsid_lock);
239 /* When peer is obtained from RCU lists, we may race with
240 * its cleanup. Check whether it's alive, and this guarantees
241 * we never hash a peer back to net->netns_ids, after it has
242 * just been idr_remove()'d from there in cleanup_net().
244 if (!maybe_get_net(peer)) {
245 spin_unlock_bh(&net->nsid_lock);
246 return NETNSA_NSID_NOT_ASSIGNED;
249 id = alloc_netid(net, peer, -1);
250 spin_unlock_bh(&net->nsid_lock);
254 return NETNSA_NSID_NOT_ASSIGNED;
256 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
260 EXPORT_SYMBOL_GPL(peernet2id_alloc);
262 /* This function returns, if assigned, the id of a peer netns. */
263 int peernet2id(const struct net *net, struct net *peer)
268 id = __peernet2id(net, peer);
273 EXPORT_SYMBOL(peernet2id);
275 /* This function returns true is the peer netns has an id assigned into the
278 bool peernet_has_id(const struct net *net, struct net *peer)
280 return peernet2id(net, peer) >= 0;
283 struct net *get_net_ns_by_id(const struct net *net, int id)
291 peer = idr_find(&net->netns_ids, id);
293 peer = maybe_get_net(peer);
298 EXPORT_SYMBOL_GPL(get_net_ns_by_id);
301 * setup_net runs the initializers for the network namespace object.
303 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
305 /* Must be called with pernet_ops_rwsem held */
306 const struct pernet_operations *ops, *saved_ops;
308 LIST_HEAD(net_exit_list);
310 refcount_set(&net->ns.count, 1);
311 ref_tracker_dir_init(&net->refcnt_tracker, 128);
313 refcount_set(&net->passive, 1);
314 get_random_bytes(&net->hash_mix, sizeof(u32));
316 net->net_cookie = gen_cookie_next(&net_cookie);
318 net->dev_base_seq = 1;
319 net->user_ns = user_ns;
320 idr_init(&net->netns_ids);
321 spin_lock_init(&net->nsid_lock);
322 mutex_init(&net->ipv4.ra_mutex);
324 list_for_each_entry(ops, &pernet_list, list) {
325 error = ops_init(ops, net);
329 down_write(&net_rwsem);
330 list_add_tail_rcu(&net->list, &net_namespace_list);
331 up_write(&net_rwsem);
336 /* Walk through the list backwards calling the exit functions
337 * for the pernet modules whose init functions did not fail.
339 list_add(&net->exit_list, &net_exit_list);
341 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
342 ops_pre_exit_list(ops, &net_exit_list);
347 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
348 ops_exit_list(ops, &net_exit_list);
351 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
352 ops_free_list(ops, &net_exit_list);
358 static int __net_init net_defaults_init_net(struct net *net)
360 net->core.sysctl_somaxconn = SOMAXCONN;
361 net->core.sysctl_txrehash = SOCK_TXREHASH_ENABLED;
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);
408 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
409 if (!net->key_domain)
411 refcount_set(&net->key_domain->usage, 1);
414 rcu_assign_pointer(net->gen, ng);
420 kmem_cache_free(net_cachep, net);
428 static void net_free(struct net *net)
430 if (refcount_dec_and_test(&net->passive)) {
431 kfree(rcu_access_pointer(net->gen));
432 kmem_cache_free(net_cachep, net);
436 void net_drop_ns(void *p)
438 struct net *net = (struct net *)p;
444 struct net *copy_net_ns(unsigned long flags,
445 struct user_namespace *user_ns, struct net *old_net)
447 struct ucounts *ucounts;
451 if (!(flags & CLONE_NEWNET))
452 return get_net(old_net);
454 ucounts = inc_net_namespaces(user_ns);
456 return ERR_PTR(-ENOSPC);
463 refcount_set(&net->passive, 1);
464 net->ucounts = ucounts;
465 get_user_ns(user_ns);
467 rv = down_read_killable(&pernet_ops_rwsem);
471 rv = setup_net(net, user_ns);
473 up_read(&pernet_ops_rwsem);
478 key_remove_domain(net->key_domain);
480 put_user_ns(user_ns);
483 dec_net_namespaces(ucounts);
490 * net_ns_get_ownership - get sysfs ownership data for @net
491 * @net: network namespace in question (can be NULL)
492 * @uid: kernel user ID for sysfs objects
493 * @gid: kernel group ID for sysfs objects
495 * Returns the uid/gid pair of root in the user namespace associated with the
496 * given network namespace.
498 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
501 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
502 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
504 if (uid_valid(ns_root_uid))
507 if (gid_valid(ns_root_gid))
510 *uid = GLOBAL_ROOT_UID;
511 *gid = GLOBAL_ROOT_GID;
514 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
516 static void unhash_nsid(struct net *net, struct net *last)
519 /* This function is only called from cleanup_net() work,
520 * and this work is the only process, that may delete
521 * a net from net_namespace_list. So, when the below
522 * is executing, the list may only grow. Thus, we do not
523 * use for_each_net_rcu() or net_rwsem.
528 spin_lock_bh(&tmp->nsid_lock);
529 id = __peernet2id(tmp, net);
531 idr_remove(&tmp->netns_ids, id);
532 spin_unlock_bh(&tmp->nsid_lock);
534 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
539 spin_lock_bh(&net->nsid_lock);
540 idr_destroy(&net->netns_ids);
541 spin_unlock_bh(&net->nsid_lock);
544 static LLIST_HEAD(cleanup_list);
546 static void cleanup_net(struct work_struct *work)
548 const struct pernet_operations *ops;
549 struct net *net, *tmp, *last;
550 struct llist_node *net_kill_list;
551 LIST_HEAD(net_exit_list);
553 /* Atomically snapshot the list of namespaces to cleanup */
554 net_kill_list = llist_del_all(&cleanup_list);
556 down_read(&pernet_ops_rwsem);
558 /* Don't let anyone else find us. */
559 down_write(&net_rwsem);
560 llist_for_each_entry(net, net_kill_list, cleanup_list)
561 list_del_rcu(&net->list);
562 /* Cache last net. After we unlock rtnl, no one new net
563 * added to net_namespace_list can assign nsid pointer
564 * to a net from net_kill_list (see peernet2id_alloc()).
565 * So, we skip them in unhash_nsid().
567 * Note, that unhash_nsid() does not delete nsid links
568 * between net_kill_list's nets, as they've already
569 * deleted from net_namespace_list. But, this would be
570 * useless anyway, as netns_ids are destroyed there.
572 last = list_last_entry(&net_namespace_list, struct net, list);
573 up_write(&net_rwsem);
575 llist_for_each_entry(net, net_kill_list, cleanup_list) {
576 unhash_nsid(net, last);
577 list_add_tail(&net->exit_list, &net_exit_list);
580 /* Run all of the network namespace pre_exit methods */
581 list_for_each_entry_reverse(ops, &pernet_list, list)
582 ops_pre_exit_list(ops, &net_exit_list);
585 * Another CPU might be rcu-iterating the list, wait for it.
586 * This needs to be before calling the exit() notifiers, so
587 * the rcu_barrier() below isn't sufficient alone.
588 * Also the pre_exit() and exit() methods need this barrier.
592 /* Run all of the network namespace exit methods */
593 list_for_each_entry_reverse(ops, &pernet_list, list)
594 ops_exit_list(ops, &net_exit_list);
596 /* Free the net generic variables */
597 list_for_each_entry_reverse(ops, &pernet_list, list)
598 ops_free_list(ops, &net_exit_list);
600 up_read(&pernet_ops_rwsem);
602 /* Ensure there are no outstanding rcu callbacks using this
607 /* Finally it is safe to free my network namespace structure */
608 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
609 list_del_init(&net->exit_list);
610 dec_net_namespaces(net->ucounts);
612 key_remove_domain(net->key_domain);
614 put_user_ns(net->user_ns);
620 * net_ns_barrier - wait until concurrent net_cleanup_work is done
622 * cleanup_net runs from work queue and will first remove namespaces
623 * from the global list, then run net exit functions.
625 * Call this in module exit path to make sure that all netns
626 * ->exit ops have been invoked before the function is removed.
628 void net_ns_barrier(void)
630 down_write(&pernet_ops_rwsem);
631 up_write(&pernet_ops_rwsem);
633 EXPORT_SYMBOL(net_ns_barrier);
635 static DECLARE_WORK(net_cleanup_work, cleanup_net);
637 void __put_net(struct net *net)
639 ref_tracker_dir_exit(&net->refcnt_tracker);
640 /* Cleanup the network namespace in process context */
641 if (llist_add(&net->cleanup_list, &cleanup_list))
642 queue_work(netns_wq, &net_cleanup_work);
644 EXPORT_SYMBOL_GPL(__put_net);
647 * get_net_ns - increment the refcount of the network namespace
648 * @ns: common namespace (net)
650 * Returns the net's common namespace.
652 struct ns_common *get_net_ns(struct ns_common *ns)
654 return &get_net(container_of(ns, struct net, ns))->ns;
656 EXPORT_SYMBOL_GPL(get_net_ns);
658 struct net *get_net_ns_by_fd(int fd)
661 struct ns_common *ns;
664 file = proc_ns_fget(fd);
666 return ERR_CAST(file);
668 ns = get_proc_ns(file_inode(file));
669 if (ns->ops == &netns_operations)
670 net = get_net(container_of(ns, struct net, ns));
672 net = ERR_PTR(-EINVAL);
677 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
680 struct net *get_net_ns_by_pid(pid_t pid)
682 struct task_struct *tsk;
685 /* Lookup the network namespace */
686 net = ERR_PTR(-ESRCH);
688 tsk = find_task_by_vpid(pid);
690 struct nsproxy *nsproxy;
692 nsproxy = tsk->nsproxy;
694 net = get_net(nsproxy->net_ns);
700 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
702 static __net_init int net_ns_net_init(struct net *net)
705 net->ns.ops = &netns_operations;
707 return ns_alloc_inum(&net->ns);
710 static __net_exit void net_ns_net_exit(struct net *net)
712 ns_free_inum(&net->ns);
715 static struct pernet_operations __net_initdata net_ns_ops = {
716 .init = net_ns_net_init,
717 .exit = net_ns_net_exit,
720 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
721 [NETNSA_NONE] = { .type = NLA_UNSPEC },
722 [NETNSA_NSID] = { .type = NLA_S32 },
723 [NETNSA_PID] = { .type = NLA_U32 },
724 [NETNSA_FD] = { .type = NLA_U32 },
725 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
728 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
729 struct netlink_ext_ack *extack)
731 struct net *net = sock_net(skb->sk);
732 struct nlattr *tb[NETNSA_MAX + 1];
737 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
738 NETNSA_MAX, rtnl_net_policy, extack);
741 if (!tb[NETNSA_NSID]) {
742 NL_SET_ERR_MSG(extack, "nsid is missing");
745 nsid = nla_get_s32(tb[NETNSA_NSID]);
747 if (tb[NETNSA_PID]) {
748 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
749 nla = tb[NETNSA_PID];
750 } else if (tb[NETNSA_FD]) {
751 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
754 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
758 NL_SET_BAD_ATTR(extack, nla);
759 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
760 return PTR_ERR(peer);
763 spin_lock_bh(&net->nsid_lock);
764 if (__peernet2id(net, peer) >= 0) {
765 spin_unlock_bh(&net->nsid_lock);
767 NL_SET_BAD_ATTR(extack, nla);
768 NL_SET_ERR_MSG(extack,
769 "Peer netns already has a nsid assigned");
773 err = alloc_netid(net, peer, nsid);
774 spin_unlock_bh(&net->nsid_lock);
776 rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
779 } else if (err == -ENOSPC && nsid >= 0) {
781 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
782 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
789 static int rtnl_net_get_size(void)
791 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
792 + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
793 + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
797 struct net_fill_args {
807 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
809 struct nlmsghdr *nlh;
810 struct rtgenmsg *rth;
812 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
817 rth = nlmsg_data(nlh);
818 rth->rtgen_family = AF_UNSPEC;
820 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
821 goto nla_put_failure;
824 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
825 goto nla_put_failure;
831 nlmsg_cancel(skb, nlh);
835 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
836 const struct nlmsghdr *nlh,
838 struct netlink_ext_ack *extack)
842 if (!netlink_strict_get_check(skb))
843 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
844 tb, NETNSA_MAX, rtnl_net_policy,
847 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
848 NETNSA_MAX, rtnl_net_policy,
853 for (i = 0; i <= NETNSA_MAX; i++) {
861 case NETNSA_TARGET_NSID:
864 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
872 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
873 struct netlink_ext_ack *extack)
875 struct net *net = sock_net(skb->sk);
876 struct nlattr *tb[NETNSA_MAX + 1];
877 struct net_fill_args fillargs = {
878 .portid = NETLINK_CB(skb).portid,
879 .seq = nlh->nlmsg_seq,
882 struct net *peer, *target = net;
887 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
890 if (tb[NETNSA_PID]) {
891 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
892 nla = tb[NETNSA_PID];
893 } else if (tb[NETNSA_FD]) {
894 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
896 } else if (tb[NETNSA_NSID]) {
897 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
899 peer = ERR_PTR(-ENOENT);
900 nla = tb[NETNSA_NSID];
902 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
907 NL_SET_BAD_ATTR(extack, nla);
908 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
909 return PTR_ERR(peer);
912 if (tb[NETNSA_TARGET_NSID]) {
913 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
915 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
916 if (IS_ERR(target)) {
917 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
918 NL_SET_ERR_MSG(extack,
919 "Target netns reference is invalid");
920 err = PTR_ERR(target);
923 fillargs.add_ref = true;
924 fillargs.ref_nsid = peernet2id(net, peer);
927 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
933 fillargs.nsid = peernet2id(target, peer);
934 err = rtnl_net_fill(msg, &fillargs);
938 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
944 if (fillargs.add_ref)
950 struct rtnl_net_dump_cb {
954 struct net_fill_args fillargs;
959 /* Runs in RCU-critical section. */
960 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
962 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
965 if (net_cb->idx < net_cb->s_idx)
968 net_cb->fillargs.nsid = id;
969 if (net_cb->fillargs.add_ref)
970 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
971 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
980 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
981 struct rtnl_net_dump_cb *net_cb,
982 struct netlink_callback *cb)
984 struct netlink_ext_ack *extack = cb->extack;
985 struct nlattr *tb[NETNSA_MAX + 1];
988 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
989 NETNSA_MAX, rtnl_net_policy,
994 for (i = 0; i <= NETNSA_MAX; i++) {
998 if (i == NETNSA_TARGET_NSID) {
1001 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
1003 NL_SET_BAD_ATTR(extack, tb[i]);
1004 NL_SET_ERR_MSG(extack,
1005 "Invalid target network namespace id");
1006 return PTR_ERR(net);
1008 net_cb->fillargs.add_ref = true;
1009 net_cb->ref_net = net_cb->tgt_net;
1010 net_cb->tgt_net = net;
1012 NL_SET_BAD_ATTR(extack, tb[i]);
1013 NL_SET_ERR_MSG(extack,
1014 "Unsupported attribute in dump request");
1022 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
1024 struct rtnl_net_dump_cb net_cb = {
1025 .tgt_net = sock_net(skb->sk),
1028 .portid = NETLINK_CB(cb->skb).portid,
1029 .seq = cb->nlh->nlmsg_seq,
1030 .flags = NLM_F_MULTI,
1034 .s_idx = cb->args[0],
1038 if (cb->strict_check) {
1039 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1045 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1048 cb->args[0] = net_cb.idx;
1050 if (net_cb.fillargs.add_ref)
1051 put_net(net_cb.tgt_net);
1052 return err < 0 ? err : skb->len;
1055 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
1056 struct nlmsghdr *nlh, gfp_t gfp)
1058 struct net_fill_args fillargs = {
1060 .seq = nlh ? nlh->nlmsg_seq : 0,
1064 struct sk_buff *msg;
1067 msg = nlmsg_new(rtnl_net_get_size(), gfp);
1071 err = rtnl_net_fill(msg, &fillargs);
1075 rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, gfp);
1081 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1084 void __init net_ns_init(void)
1086 struct net_generic *ng;
1088 #ifdef CONFIG_NET_NS
1089 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1091 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1093 /* Create workqueue for cleanup */
1094 netns_wq = create_singlethread_workqueue("netns");
1096 panic("Could not create netns workq");
1099 ng = net_alloc_generic();
1101 panic("Could not allocate generic netns");
1103 rcu_assign_pointer(init_net.gen, ng);
1106 init_net.key_domain = &init_net_key_domain;
1108 down_write(&pernet_ops_rwsem);
1109 if (setup_net(&init_net, &init_user_ns))
1110 panic("Could not setup the initial network namespace");
1112 init_net_initialized = true;
1113 up_write(&pernet_ops_rwsem);
1115 if (register_pernet_subsys(&net_ns_ops))
1116 panic("Could not register network namespace subsystems");
1118 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1119 RTNL_FLAG_DOIT_UNLOCKED);
1120 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1121 RTNL_FLAG_DOIT_UNLOCKED);
1124 static void free_exit_list(struct pernet_operations *ops, struct list_head *net_exit_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);
1132 #ifdef CONFIG_NET_NS
1133 static int __register_pernet_operations(struct list_head *list,
1134 struct pernet_operations *ops)
1138 LIST_HEAD(net_exit_list);
1140 list_add_tail(&ops->list, list);
1141 if (ops->init || (ops->id && ops->size)) {
1142 /* We held write locked pernet_ops_rwsem, and parallel
1143 * setup_net() and cleanup_net() are not possible.
1146 error = ops_init(ops, net);
1149 list_add_tail(&net->exit_list, &net_exit_list);
1155 /* If I have an error cleanup all namespaces I initialized */
1156 list_del(&ops->list);
1157 free_exit_list(ops, &net_exit_list);
1161 static void __unregister_pernet_operations(struct pernet_operations *ops)
1164 LIST_HEAD(net_exit_list);
1166 list_del(&ops->list);
1167 /* See comment in __register_pernet_operations() */
1169 list_add_tail(&net->exit_list, &net_exit_list);
1171 free_exit_list(ops, &net_exit_list);
1176 static int __register_pernet_operations(struct list_head *list,
1177 struct pernet_operations *ops)
1179 if (!init_net_initialized) {
1180 list_add_tail(&ops->list, list);
1184 return ops_init(ops, &init_net);
1187 static void __unregister_pernet_operations(struct pernet_operations *ops)
1189 if (!init_net_initialized) {
1190 list_del(&ops->list);
1192 LIST_HEAD(net_exit_list);
1193 list_add(&init_net.exit_list, &net_exit_list);
1194 free_exit_list(ops, &net_exit_list);
1198 #endif /* CONFIG_NET_NS */
1200 static DEFINE_IDA(net_generic_ids);
1202 static int register_pernet_operations(struct list_head *list,
1203 struct pernet_operations *ops)
1208 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1213 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1215 error = __register_pernet_operations(list, ops);
1219 ida_free(&net_generic_ids, *ops->id);
1225 static void unregister_pernet_operations(struct pernet_operations *ops)
1227 __unregister_pernet_operations(ops);
1230 ida_free(&net_generic_ids, *ops->id);
1234 * register_pernet_subsys - register a network namespace subsystem
1235 * @ops: pernet operations structure for the subsystem
1237 * Register a subsystem which has init and exit functions
1238 * that are called when network namespaces are created and
1239 * destroyed respectively.
1241 * When registered all network namespace init functions are
1242 * called for every existing network namespace. Allowing kernel
1243 * modules to have a race free view of the set of network namespaces.
1245 * When a new network namespace is created all of the init
1246 * methods are called in the order in which they were registered.
1248 * When a network namespace is destroyed all of the exit methods
1249 * are called in the reverse of the order with which they were
1252 int register_pernet_subsys(struct pernet_operations *ops)
1255 down_write(&pernet_ops_rwsem);
1256 error = register_pernet_operations(first_device, ops);
1257 up_write(&pernet_ops_rwsem);
1260 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1263 * unregister_pernet_subsys - unregister a network namespace subsystem
1264 * @ops: pernet operations structure to manipulate
1266 * Remove the pernet operations structure from the list to be
1267 * used when network namespaces are created or destroyed. In
1268 * addition run the exit method for all existing network
1271 void unregister_pernet_subsys(struct pernet_operations *ops)
1273 down_write(&pernet_ops_rwsem);
1274 unregister_pernet_operations(ops);
1275 up_write(&pernet_ops_rwsem);
1277 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1280 * register_pernet_device - register a network namespace device
1281 * @ops: pernet operations structure for the subsystem
1283 * Register a device which has init and exit functions
1284 * that are called when network namespaces are created and
1285 * destroyed respectively.
1287 * When registered all network namespace init functions are
1288 * called for every existing network namespace. Allowing kernel
1289 * modules to have a race free view of the set of network namespaces.
1291 * When a new network namespace is created all of the init
1292 * methods are called in the order in which they were registered.
1294 * When a network namespace is destroyed all of the exit methods
1295 * are called in the reverse of the order with which they were
1298 int register_pernet_device(struct pernet_operations *ops)
1301 down_write(&pernet_ops_rwsem);
1302 error = register_pernet_operations(&pernet_list, ops);
1303 if (!error && (first_device == &pernet_list))
1304 first_device = &ops->list;
1305 up_write(&pernet_ops_rwsem);
1308 EXPORT_SYMBOL_GPL(register_pernet_device);
1311 * unregister_pernet_device - unregister a network namespace netdevice
1312 * @ops: pernet operations structure to manipulate
1314 * Remove the pernet operations structure from the list to be
1315 * used when network namespaces are created or destroyed. In
1316 * addition run the exit method for all existing network
1319 void unregister_pernet_device(struct pernet_operations *ops)
1321 down_write(&pernet_ops_rwsem);
1322 if (&ops->list == first_device)
1323 first_device = first_device->next;
1324 unregister_pernet_operations(ops);
1325 up_write(&pernet_ops_rwsem);
1327 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1329 #ifdef CONFIG_NET_NS
1330 static struct ns_common *netns_get(struct task_struct *task)
1332 struct net *net = NULL;
1333 struct nsproxy *nsproxy;
1336 nsproxy = task->nsproxy;
1338 net = get_net(nsproxy->net_ns);
1341 return net ? &net->ns : NULL;
1344 static inline struct net *to_net_ns(struct ns_common *ns)
1346 return container_of(ns, struct net, ns);
1349 static void netns_put(struct ns_common *ns)
1351 put_net(to_net_ns(ns));
1354 static int netns_install(struct nsset *nsset, struct ns_common *ns)
1356 struct nsproxy *nsproxy = nsset->nsproxy;
1357 struct net *net = to_net_ns(ns);
1359 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1360 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
1363 put_net(nsproxy->net_ns);
1364 nsproxy->net_ns = get_net(net);
1368 static struct user_namespace *netns_owner(struct ns_common *ns)
1370 return to_net_ns(ns)->user_ns;
1373 const struct proc_ns_operations netns_operations = {
1375 .type = CLONE_NEWNET,
1378 .install = netns_install,
1379 .owner = netns_owner,