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_exit_list(const struct pernet_operations
*ops
,
149 struct list_head
*net_exit_list
)
153 list_for_each_entry(net
, net_exit_list
, exit_list
)
157 ops
->exit_batch(net_exit_list
);
160 static void ops_free_list(const struct pernet_operations
*ops
,
161 struct list_head
*net_exit_list
)
164 if (ops
->size
&& ops
->id
) {
165 list_for_each_entry(net
, net_exit_list
, exit_list
)
170 /* should be called with nsid_lock held */
171 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
173 int min
= 0, max
= 0;
180 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
183 /* This function is used by idr_for_each(). If net is equal to peer, the
184 * function returns the id so that idr_for_each() stops. Because we cannot
185 * returns the id 0 (idr_for_each() will not stop), we return the magic value
186 * NET_ID_ZERO (-1) for it.
188 #define NET_ID_ZERO -1
189 static int net_eq_idr(int id
, void *net
, void *peer
)
191 if (net_eq(net
, peer
))
192 return id
? : NET_ID_ZERO
;
196 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
197 * is set to true, thus the caller knows that the new id must be notified via
200 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
202 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
203 bool alloc_it
= *alloc
;
207 /* Magic value for id 0. */
208 if (id
== NET_ID_ZERO
)
214 id
= alloc_netid(net
, peer
, -1);
216 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
219 return NETNSA_NSID_NOT_ASSIGNED
;
222 /* should be called with nsid_lock held */
223 static int __peernet2id(struct net
*net
, struct net
*peer
)
227 return __peernet2id_alloc(net
, peer
, &no
);
230 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
231 /* This function returns the id of a peer netns. If no id is assigned, one will
232 * be allocated and returned.
234 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
236 bool alloc
= false, alive
= false;
239 if (refcount_read(&net
->count
) == 0)
240 return NETNSA_NSID_NOT_ASSIGNED
;
241 spin_lock_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 alive
= alloc
= true;
250 id
= __peernet2id_alloc(net
, peer
, &alloc
);
251 spin_unlock_bh(&net
->nsid_lock
);
252 if (alloc
&& id
>= 0)
253 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
258 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
260 /* This function returns, if assigned, the id of a peer netns. */
261 int peernet2id(struct net
*net
, struct net
*peer
)
265 spin_lock_bh(&net
->nsid_lock
);
266 id
= __peernet2id(net
, peer
);
267 spin_unlock_bh(&net
->nsid_lock
);
270 EXPORT_SYMBOL(peernet2id
);
272 /* This function returns true is the peer netns has an id assigned into the
275 bool peernet_has_id(struct net
*net
, struct net
*peer
)
277 return peernet2id(net
, peer
) >= 0;
280 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
288 peer
= idr_find(&net
->netns_ids
, id
);
290 peer
= maybe_get_net(peer
);
297 * setup_net runs the initializers for the network namespace object.
299 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
301 /* Must be called with pernet_ops_rwsem held */
302 const struct pernet_operations
*ops
, *saved_ops
;
304 LIST_HEAD(net_exit_list
);
306 refcount_set(&net
->count
, 1);
307 refcount_set(&net
->passive
, 1);
308 get_random_bytes(&net
->hash_mix
, sizeof(u32
));
309 net
->dev_base_seq
= 1;
310 net
->user_ns
= user_ns
;
311 idr_init(&net
->netns_ids
);
312 spin_lock_init(&net
->nsid_lock
);
313 mutex_init(&net
->ipv4
.ra_mutex
);
315 list_for_each_entry(ops
, &pernet_list
, list
) {
316 error
= ops_init(ops
, net
);
320 down_write(&net_rwsem
);
321 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
322 up_write(&net_rwsem
);
327 /* Walk through the list backwards calling the exit functions
328 * for the pernet modules whose init functions did not fail.
330 list_add(&net
->exit_list
, &net_exit_list
);
332 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
333 ops_exit_list(ops
, &net_exit_list
);
336 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
337 ops_free_list(ops
, &net_exit_list
);
343 static int __net_init
net_defaults_init_net(struct net
*net
)
345 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
349 static struct pernet_operations net_defaults_ops
= {
350 .init
= net_defaults_init_net
,
353 static __init
int net_defaults_init(void)
355 if (register_pernet_subsys(&net_defaults_ops
))
356 panic("Cannot initialize net default settings");
361 core_initcall(net_defaults_init
);
364 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
366 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
369 static void dec_net_namespaces(struct ucounts
*ucounts
)
371 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
374 static struct kmem_cache
*net_cachep __ro_after_init
;
375 static struct workqueue_struct
*netns_wq
;
377 static struct net
*net_alloc(void)
379 struct net
*net
= NULL
;
380 struct net_generic
*ng
;
382 ng
= net_alloc_generic();
386 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
390 rcu_assign_pointer(net
->gen
, ng
);
399 static void net_free(struct net
*net
)
401 kfree(rcu_access_pointer(net
->gen
));
402 kmem_cache_free(net_cachep
, net
);
405 void net_drop_ns(void *p
)
408 if (ns
&& refcount_dec_and_test(&ns
->passive
))
412 struct net
*copy_net_ns(unsigned long flags
,
413 struct user_namespace
*user_ns
, struct net
*old_net
)
415 struct ucounts
*ucounts
;
419 if (!(flags
& CLONE_NEWNET
))
420 return get_net(old_net
);
422 ucounts
= inc_net_namespaces(user_ns
);
424 return ERR_PTR(-ENOSPC
);
431 refcount_set(&net
->passive
, 1);
432 net
->ucounts
= ucounts
;
433 get_user_ns(user_ns
);
435 rv
= down_read_killable(&pernet_ops_rwsem
);
439 rv
= setup_net(net
, user_ns
);
441 up_read(&pernet_ops_rwsem
);
445 put_user_ns(user_ns
);
448 dec_net_namespaces(ucounts
);
455 * net_ns_get_ownership - get sysfs ownership data for @net
456 * @net: network namespace in question (can be NULL)
457 * @uid: kernel user ID for sysfs objects
458 * @gid: kernel group ID for sysfs objects
460 * Returns the uid/gid pair of root in the user namespace associated with the
461 * given network namespace.
463 void net_ns_get_ownership(const struct net
*net
, kuid_t
*uid
, kgid_t
*gid
)
466 kuid_t ns_root_uid
= make_kuid(net
->user_ns
, 0);
467 kgid_t ns_root_gid
= make_kgid(net
->user_ns
, 0);
469 if (uid_valid(ns_root_uid
))
472 if (gid_valid(ns_root_gid
))
475 *uid
= GLOBAL_ROOT_UID
;
476 *gid
= GLOBAL_ROOT_GID
;
479 EXPORT_SYMBOL_GPL(net_ns_get_ownership
);
481 static void unhash_nsid(struct net
*net
, struct net
*last
)
484 /* This function is only called from cleanup_net() work,
485 * and this work is the only process, that may delete
486 * a net from net_namespace_list. So, when the below
487 * is executing, the list may only grow. Thus, we do not
488 * use for_each_net_rcu() or net_rwsem.
493 spin_lock_bh(&tmp
->nsid_lock
);
494 id
= __peernet2id(tmp
, net
);
496 idr_remove(&tmp
->netns_ids
, id
);
497 spin_unlock_bh(&tmp
->nsid_lock
);
499 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
503 spin_lock_bh(&net
->nsid_lock
);
504 idr_destroy(&net
->netns_ids
);
505 spin_unlock_bh(&net
->nsid_lock
);
508 static LLIST_HEAD(cleanup_list
);
510 static void cleanup_net(struct work_struct
*work
)
512 const struct pernet_operations
*ops
;
513 struct net
*net
, *tmp
, *last
;
514 struct llist_node
*net_kill_list
;
515 LIST_HEAD(net_exit_list
);
517 /* Atomically snapshot the list of namespaces to cleanup */
518 net_kill_list
= llist_del_all(&cleanup_list
);
520 down_read(&pernet_ops_rwsem
);
522 /* Don't let anyone else find us. */
523 down_write(&net_rwsem
);
524 llist_for_each_entry(net
, net_kill_list
, cleanup_list
)
525 list_del_rcu(&net
->list
);
526 /* Cache last net. After we unlock rtnl, no one new net
527 * added to net_namespace_list can assign nsid pointer
528 * to a net from net_kill_list (see peernet2id_alloc()).
529 * So, we skip them in unhash_nsid().
531 * Note, that unhash_nsid() does not delete nsid links
532 * between net_kill_list's nets, as they've already
533 * deleted from net_namespace_list. But, this would be
534 * useless anyway, as netns_ids are destroyed there.
536 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
537 up_write(&net_rwsem
);
539 llist_for_each_entry(net
, net_kill_list
, cleanup_list
) {
540 unhash_nsid(net
, last
);
541 list_add_tail(&net
->exit_list
, &net_exit_list
);
545 * Another CPU might be rcu-iterating the list, wait for it.
546 * This needs to be before calling the exit() notifiers, so
547 * the rcu_barrier() below isn't sufficient alone.
551 /* Run all of the network namespace exit methods */
552 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
553 ops_exit_list(ops
, &net_exit_list
);
555 /* Free the net generic variables */
556 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
557 ops_free_list(ops
, &net_exit_list
);
559 up_read(&pernet_ops_rwsem
);
561 /* Ensure there are no outstanding rcu callbacks using this
566 /* Finally it is safe to free my network namespace structure */
567 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
568 list_del_init(&net
->exit_list
);
569 dec_net_namespaces(net
->ucounts
);
570 put_user_ns(net
->user_ns
);
576 * net_ns_barrier - wait until concurrent net_cleanup_work is done
578 * cleanup_net runs from work queue and will first remove namespaces
579 * from the global list, then run net exit functions.
581 * Call this in module exit path to make sure that all netns
582 * ->exit ops have been invoked before the function is removed.
584 void net_ns_barrier(void)
586 down_write(&pernet_ops_rwsem
);
587 up_write(&pernet_ops_rwsem
);
589 EXPORT_SYMBOL(net_ns_barrier
);
591 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
593 void __put_net(struct net
*net
)
595 /* Cleanup the network namespace in process context */
596 if (llist_add(&net
->cleanup_list
, &cleanup_list
))
597 queue_work(netns_wq
, &net_cleanup_work
);
599 EXPORT_SYMBOL_GPL(__put_net
);
601 struct net
*get_net_ns_by_fd(int fd
)
604 struct ns_common
*ns
;
607 file
= proc_ns_fget(fd
);
609 return ERR_CAST(file
);
611 ns
= get_proc_ns(file_inode(file
));
612 if (ns
->ops
== &netns_operations
)
613 net
= get_net(container_of(ns
, struct net
, ns
));
615 net
= ERR_PTR(-EINVAL
);
622 struct net
*get_net_ns_by_fd(int fd
)
624 return ERR_PTR(-EINVAL
);
627 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
629 struct net
*get_net_ns_by_pid(pid_t pid
)
631 struct task_struct
*tsk
;
634 /* Lookup the network namespace */
635 net
= ERR_PTR(-ESRCH
);
637 tsk
= find_task_by_vpid(pid
);
639 struct nsproxy
*nsproxy
;
641 nsproxy
= tsk
->nsproxy
;
643 net
= get_net(nsproxy
->net_ns
);
649 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
651 static __net_init
int net_ns_net_init(struct net
*net
)
654 net
->ns
.ops
= &netns_operations
;
656 return ns_alloc_inum(&net
->ns
);
659 static __net_exit
void net_ns_net_exit(struct net
*net
)
661 ns_free_inum(&net
->ns
);
664 static struct pernet_operations __net_initdata net_ns_ops
= {
665 .init
= net_ns_net_init
,
666 .exit
= net_ns_net_exit
,
669 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
670 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
671 [NETNSA_NSID
] = { .type
= NLA_S32
},
672 [NETNSA_PID
] = { .type
= NLA_U32
},
673 [NETNSA_FD
] = { .type
= NLA_U32
},
674 [NETNSA_TARGET_NSID
] = { .type
= NLA_S32
},
677 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
678 struct netlink_ext_ack
*extack
)
680 struct net
*net
= sock_net(skb
->sk
);
681 struct nlattr
*tb
[NETNSA_MAX
+ 1];
686 err
= nlmsg_parse_deprecated(nlh
, sizeof(struct rtgenmsg
), tb
,
687 NETNSA_MAX
, rtnl_net_policy
, extack
);
690 if (!tb
[NETNSA_NSID
]) {
691 NL_SET_ERR_MSG(extack
, "nsid is missing");
694 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
696 if (tb
[NETNSA_PID
]) {
697 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
698 nla
= tb
[NETNSA_PID
];
699 } else if (tb
[NETNSA_FD
]) {
700 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
703 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
707 NL_SET_BAD_ATTR(extack
, nla
);
708 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
709 return PTR_ERR(peer
);
712 spin_lock_bh(&net
->nsid_lock
);
713 if (__peernet2id(net
, peer
) >= 0) {
714 spin_unlock_bh(&net
->nsid_lock
);
716 NL_SET_BAD_ATTR(extack
, nla
);
717 NL_SET_ERR_MSG(extack
,
718 "Peer netns already has a nsid assigned");
722 err
= alloc_netid(net
, peer
, nsid
);
723 spin_unlock_bh(&net
->nsid_lock
);
725 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
727 } else if (err
== -ENOSPC
&& nsid
>= 0) {
729 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
730 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
737 static int rtnl_net_get_size(void)
739 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
740 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
741 + nla_total_size(sizeof(s32
)) /* NETNSA_CURRENT_NSID */
745 struct net_fill_args
{
755 static int rtnl_net_fill(struct sk_buff
*skb
, struct net_fill_args
*args
)
757 struct nlmsghdr
*nlh
;
758 struct rtgenmsg
*rth
;
760 nlh
= nlmsg_put(skb
, args
->portid
, args
->seq
, args
->cmd
, sizeof(*rth
),
765 rth
= nlmsg_data(nlh
);
766 rth
->rtgen_family
= AF_UNSPEC
;
768 if (nla_put_s32(skb
, NETNSA_NSID
, args
->nsid
))
769 goto nla_put_failure
;
772 nla_put_s32(skb
, NETNSA_CURRENT_NSID
, args
->ref_nsid
))
773 goto nla_put_failure
;
779 nlmsg_cancel(skb
, nlh
);
783 static int rtnl_net_valid_getid_req(struct sk_buff
*skb
,
784 const struct nlmsghdr
*nlh
,
786 struct netlink_ext_ack
*extack
)
790 if (!netlink_strict_get_check(skb
))
791 return nlmsg_parse_deprecated(nlh
, sizeof(struct rtgenmsg
),
792 tb
, NETNSA_MAX
, rtnl_net_policy
,
795 err
= nlmsg_parse_deprecated_strict(nlh
, sizeof(struct rtgenmsg
), tb
,
796 NETNSA_MAX
, rtnl_net_policy
,
801 for (i
= 0; i
<= NETNSA_MAX
; i
++) {
809 case NETNSA_TARGET_NSID
:
812 NL_SET_ERR_MSG(extack
, "Unsupported attribute in peer netns getid request");
820 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
821 struct netlink_ext_ack
*extack
)
823 struct net
*net
= sock_net(skb
->sk
);
824 struct nlattr
*tb
[NETNSA_MAX
+ 1];
825 struct net_fill_args fillargs
= {
826 .portid
= NETLINK_CB(skb
).portid
,
827 .seq
= nlh
->nlmsg_seq
,
830 struct net
*peer
, *target
= net
;
835 err
= rtnl_net_valid_getid_req(skb
, nlh
, tb
, extack
);
838 if (tb
[NETNSA_PID
]) {
839 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
840 nla
= tb
[NETNSA_PID
];
841 } else if (tb
[NETNSA_FD
]) {
842 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
844 } else if (tb
[NETNSA_NSID
]) {
845 peer
= get_net_ns_by_id(net
, nla_get_s32(tb
[NETNSA_NSID
]));
847 peer
= ERR_PTR(-ENOENT
);
848 nla
= tb
[NETNSA_NSID
];
850 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
855 NL_SET_BAD_ATTR(extack
, nla
);
856 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
857 return PTR_ERR(peer
);
860 if (tb
[NETNSA_TARGET_NSID
]) {
861 int id
= nla_get_s32(tb
[NETNSA_TARGET_NSID
]);
863 target
= rtnl_get_net_ns_capable(NETLINK_CB(skb
).sk
, id
);
864 if (IS_ERR(target
)) {
865 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_TARGET_NSID
]);
866 NL_SET_ERR_MSG(extack
,
867 "Target netns reference is invalid");
868 err
= PTR_ERR(target
);
871 fillargs
.add_ref
= true;
872 fillargs
.ref_nsid
= peernet2id(net
, peer
);
875 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
881 fillargs
.nsid
= peernet2id(target
, peer
);
882 err
= rtnl_net_fill(msg
, &fillargs
);
886 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
892 if (fillargs
.add_ref
)
898 struct rtnl_net_dump_cb
{
902 struct net_fill_args fillargs
;
907 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
909 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
912 if (net_cb
->idx
< net_cb
->s_idx
)
915 net_cb
->fillargs
.nsid
= id
;
916 if (net_cb
->fillargs
.add_ref
)
917 net_cb
->fillargs
.ref_nsid
= __peernet2id(net_cb
->ref_net
, peer
);
918 ret
= rtnl_net_fill(net_cb
->skb
, &net_cb
->fillargs
);
927 static int rtnl_valid_dump_net_req(const struct nlmsghdr
*nlh
, struct sock
*sk
,
928 struct rtnl_net_dump_cb
*net_cb
,
929 struct netlink_callback
*cb
)
931 struct netlink_ext_ack
*extack
= cb
->extack
;
932 struct nlattr
*tb
[NETNSA_MAX
+ 1];
935 err
= nlmsg_parse_deprecated_strict(nlh
, sizeof(struct rtgenmsg
), tb
,
936 NETNSA_MAX
, rtnl_net_policy
,
941 for (i
= 0; i
<= NETNSA_MAX
; i
++) {
945 if (i
== NETNSA_TARGET_NSID
) {
948 net
= rtnl_get_net_ns_capable(sk
, nla_get_s32(tb
[i
]));
950 NL_SET_BAD_ATTR(extack
, tb
[i
]);
951 NL_SET_ERR_MSG(extack
,
952 "Invalid target network namespace id");
955 net_cb
->fillargs
.add_ref
= true;
956 net_cb
->ref_net
= net_cb
->tgt_net
;
957 net_cb
->tgt_net
= net
;
959 NL_SET_BAD_ATTR(extack
, tb
[i
]);
960 NL_SET_ERR_MSG(extack
,
961 "Unsupported attribute in dump request");
969 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
971 struct rtnl_net_dump_cb net_cb
= {
972 .tgt_net
= sock_net(skb
->sk
),
975 .portid
= NETLINK_CB(cb
->skb
).portid
,
976 .seq
= cb
->nlh
->nlmsg_seq
,
977 .flags
= NLM_F_MULTI
,
981 .s_idx
= cb
->args
[0],
985 if (cb
->strict_check
) {
986 err
= rtnl_valid_dump_net_req(cb
->nlh
, skb
->sk
, &net_cb
, cb
);
991 spin_lock_bh(&net_cb
.tgt_net
->nsid_lock
);
992 if (net_cb
.fillargs
.add_ref
&&
993 !net_eq(net_cb
.ref_net
, net_cb
.tgt_net
) &&
994 !spin_trylock_bh(&net_cb
.ref_net
->nsid_lock
)) {
995 spin_unlock_bh(&net_cb
.tgt_net
->nsid_lock
);
999 idr_for_each(&net_cb
.tgt_net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
1000 if (net_cb
.fillargs
.add_ref
&&
1001 !net_eq(net_cb
.ref_net
, net_cb
.tgt_net
))
1002 spin_unlock_bh(&net_cb
.ref_net
->nsid_lock
);
1003 spin_unlock_bh(&net_cb
.tgt_net
->nsid_lock
);
1005 cb
->args
[0] = net_cb
.idx
;
1007 if (net_cb
.fillargs
.add_ref
)
1008 put_net(net_cb
.tgt_net
);
1009 return err
< 0 ? err
: skb
->len
;
1012 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
1014 struct net_fill_args fillargs
= {
1018 struct sk_buff
*msg
;
1021 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
1025 err
= rtnl_net_fill(msg
, &fillargs
);
1029 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
1035 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
1038 static int __init
net_ns_init(void)
1040 struct net_generic
*ng
;
1042 #ifdef CONFIG_NET_NS
1043 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
1045 SLAB_PANIC
|SLAB_ACCOUNT
, NULL
);
1047 /* Create workqueue for cleanup */
1048 netns_wq
= create_singlethread_workqueue("netns");
1050 panic("Could not create netns workq");
1053 ng
= net_alloc_generic();
1055 panic("Could not allocate generic netns");
1057 rcu_assign_pointer(init_net
.gen
, ng
);
1059 down_write(&pernet_ops_rwsem
);
1060 if (setup_net(&init_net
, &init_user_ns
))
1061 panic("Could not setup the initial network namespace");
1063 init_net_initialized
= true;
1064 up_write(&pernet_ops_rwsem
);
1066 if (register_pernet_subsys(&net_ns_ops
))
1067 panic("Could not register network namespace subsystems");
1069 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
1070 RTNL_FLAG_DOIT_UNLOCKED
);
1071 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
1072 RTNL_FLAG_DOIT_UNLOCKED
);
1077 pure_initcall(net_ns_init
);
1079 #ifdef CONFIG_NET_NS
1080 static int __register_pernet_operations(struct list_head
*list
,
1081 struct pernet_operations
*ops
)
1085 LIST_HEAD(net_exit_list
);
1087 list_add_tail(&ops
->list
, list
);
1088 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
1089 /* We held write locked pernet_ops_rwsem, and parallel
1090 * setup_net() and cleanup_net() are not possible.
1093 error
= ops_init(ops
, net
);
1096 list_add_tail(&net
->exit_list
, &net_exit_list
);
1102 /* If I have an error cleanup all namespaces I initialized */
1103 list_del(&ops
->list
);
1104 ops_exit_list(ops
, &net_exit_list
);
1105 ops_free_list(ops
, &net_exit_list
);
1109 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
1112 LIST_HEAD(net_exit_list
);
1114 list_del(&ops
->list
);
1115 /* See comment in __register_pernet_operations() */
1117 list_add_tail(&net
->exit_list
, &net_exit_list
);
1118 ops_exit_list(ops
, &net_exit_list
);
1119 ops_free_list(ops
, &net_exit_list
);
1124 static int __register_pernet_operations(struct list_head
*list
,
1125 struct pernet_operations
*ops
)
1127 if (!init_net_initialized
) {
1128 list_add_tail(&ops
->list
, list
);
1132 return ops_init(ops
, &init_net
);
1135 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
1137 if (!init_net_initialized
) {
1138 list_del(&ops
->list
);
1140 LIST_HEAD(net_exit_list
);
1141 list_add(&init_net
.exit_list
, &net_exit_list
);
1142 ops_exit_list(ops
, &net_exit_list
);
1143 ops_free_list(ops
, &net_exit_list
);
1147 #endif /* CONFIG_NET_NS */
1149 static DEFINE_IDA(net_generic_ids
);
1151 static int register_pernet_operations(struct list_head
*list
,
1152 struct pernet_operations
*ops
)
1157 error
= ida_alloc_min(&net_generic_ids
, MIN_PERNET_OPS_ID
,
1162 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
1164 error
= __register_pernet_operations(list
, ops
);
1168 ida_free(&net_generic_ids
, *ops
->id
);
1174 static void unregister_pernet_operations(struct pernet_operations
*ops
)
1176 __unregister_pernet_operations(ops
);
1179 ida_free(&net_generic_ids
, *ops
->id
);
1183 * register_pernet_subsys - register a network namespace subsystem
1184 * @ops: pernet operations structure for the subsystem
1186 * Register a subsystem which has init and exit functions
1187 * that are called when network namespaces are created and
1188 * destroyed respectively.
1190 * When registered all network namespace init functions are
1191 * called for every existing network namespace. Allowing kernel
1192 * modules to have a race free view of the set of network namespaces.
1194 * When a new network namespace is created all of the init
1195 * methods are called in the order in which they were registered.
1197 * When a network namespace is destroyed all of the exit methods
1198 * are called in the reverse of the order with which they were
1201 int register_pernet_subsys(struct pernet_operations
*ops
)
1204 down_write(&pernet_ops_rwsem
);
1205 error
= register_pernet_operations(first_device
, ops
);
1206 up_write(&pernet_ops_rwsem
);
1209 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1212 * unregister_pernet_subsys - unregister a network namespace subsystem
1213 * @ops: pernet operations structure to manipulate
1215 * Remove the pernet operations structure from the list to be
1216 * used when network namespaces are created or destroyed. In
1217 * addition run the exit method for all existing network
1220 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1222 down_write(&pernet_ops_rwsem
);
1223 unregister_pernet_operations(ops
);
1224 up_write(&pernet_ops_rwsem
);
1226 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1229 * register_pernet_device - register a network namespace device
1230 * @ops: pernet operations structure for the subsystem
1232 * Register a device 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_device(struct pernet_operations
*ops
)
1250 down_write(&pernet_ops_rwsem
);
1251 error
= register_pernet_operations(&pernet_list
, ops
);
1252 if (!error
&& (first_device
== &pernet_list
))
1253 first_device
= &ops
->list
;
1254 up_write(&pernet_ops_rwsem
);
1257 EXPORT_SYMBOL_GPL(register_pernet_device
);
1260 * unregister_pernet_device - unregister a network namespace netdevice
1261 * @ops: pernet operations structure to manipulate
1263 * Remove the pernet operations structure from the list to be
1264 * used when network namespaces are created or destroyed. In
1265 * addition run the exit method for all existing network
1268 void unregister_pernet_device(struct pernet_operations
*ops
)
1270 down_write(&pernet_ops_rwsem
);
1271 if (&ops
->list
== first_device
)
1272 first_device
= first_device
->next
;
1273 unregister_pernet_operations(ops
);
1274 up_write(&pernet_ops_rwsem
);
1276 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1278 #ifdef CONFIG_NET_NS
1279 static struct ns_common
*netns_get(struct task_struct
*task
)
1281 struct net
*net
= NULL
;
1282 struct nsproxy
*nsproxy
;
1285 nsproxy
= task
->nsproxy
;
1287 net
= get_net(nsproxy
->net_ns
);
1290 return net
? &net
->ns
: NULL
;
1293 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1295 return container_of(ns
, struct net
, ns
);
1298 static void netns_put(struct ns_common
*ns
)
1300 put_net(to_net_ns(ns
));
1303 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1305 struct net
*net
= to_net_ns(ns
);
1307 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1308 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1311 put_net(nsproxy
->net_ns
);
1312 nsproxy
->net_ns
= get_net(net
);
1316 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1318 return to_net_ns(ns
)->user_ns
;
1321 const struct proc_ns_operations netns_operations
= {
1323 .type
= CLONE_NEWNET
,
1326 .install
= netns_install
,
1327 .owner
= netns_owner
,