1 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 #include <linux/workqueue.h>
4 #include <linux/rtnetlink.h>
5 #include <linux/cache.h>
6 #include <linux/slab.h>
7 #include <linux/list.h>
8 #include <linux/delay.h>
9 #include <linux/sched.h>
10 #include <linux/idr.h>
11 #include <linux/rculist.h>
12 #include <linux/nsproxy.h>
14 #include <linux/proc_ns.h>
15 #include <linux/file.h>
16 #include <linux/export.h>
17 #include <linux/user_namespace.h>
18 #include <linux/net_namespace.h>
19 #include <linux/sched/task.h>
20 #include <linux/uidgid.h>
23 #include <net/netlink.h>
24 #include <net/net_namespace.h>
25 #include <net/netns/generic.h>
28 * Our network namespace constructor/destructor lists
31 static LIST_HEAD(pernet_list
);
32 static struct list_head
*first_device
= &pernet_list
;
34 LIST_HEAD(net_namespace_list
);
35 EXPORT_SYMBOL_GPL(net_namespace_list
);
37 /* Protects net_namespace_list. Nests iside rtnl_lock() */
38 DECLARE_RWSEM(net_rwsem
);
39 EXPORT_SYMBOL_GPL(net_rwsem
);
41 struct net init_net
= {
42 .count
= REFCOUNT_INIT(1),
43 .dev_base_head
= LIST_HEAD_INIT(init_net
.dev_base_head
),
45 EXPORT_SYMBOL(init_net
);
47 static bool init_net_initialized
;
49 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
50 * init_net_initialized and first_device pointer.
51 * This is internal net namespace object. Please, don't use it
54 DECLARE_RWSEM(pernet_ops_rwsem
);
55 EXPORT_SYMBOL_GPL(pernet_ops_rwsem
);
57 #define MIN_PERNET_OPS_ID \
58 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
60 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
62 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
64 static struct net_generic
*net_alloc_generic(void)
66 struct net_generic
*ng
;
67 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
69 ng
= kzalloc(generic_size
, GFP_KERNEL
);
71 ng
->s
.len
= max_gen_ptrs
;
76 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
78 struct net_generic
*ng
, *old_ng
;
80 BUG_ON(id
< MIN_PERNET_OPS_ID
);
82 old_ng
= rcu_dereference_protected(net
->gen
,
83 lockdep_is_held(&pernet_ops_rwsem
));
84 if (old_ng
->s
.len
> id
) {
85 old_ng
->ptr
[id
] = data
;
89 ng
= net_alloc_generic();
94 * Some synchronisation notes:
96 * The net_generic explores the net->gen array inside rcu
97 * read section. Besides once set the net->gen->ptr[x]
98 * pointer never changes (see rules in netns/generic.h).
100 * That said, we simply duplicate this array and schedule
101 * the old copy for kfree after a grace period.
104 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
105 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
108 rcu_assign_pointer(net
->gen
, ng
);
109 kfree_rcu(old_ng
, s
.rcu
);
113 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
118 if (ops
->id
&& ops
->size
) {
119 data
= kzalloc(ops
->size
, GFP_KERNEL
);
123 err
= net_assign_generic(net
, *ops
->id
, data
);
129 err
= ops
->init(net
);
140 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
142 if (ops
->id
&& ops
->size
) {
143 kfree(net_generic(net
, *ops
->id
));
147 static void ops_exit_list(const struct pernet_operations
*ops
,
148 struct list_head
*net_exit_list
)
152 list_for_each_entry(net
, net_exit_list
, exit_list
)
156 ops
->exit_batch(net_exit_list
);
159 static void ops_free_list(const struct pernet_operations
*ops
,
160 struct list_head
*net_exit_list
)
163 if (ops
->size
&& ops
->id
) {
164 list_for_each_entry(net
, net_exit_list
, exit_list
)
169 /* should be called with nsid_lock held */
170 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
172 int min
= 0, max
= 0;
179 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
182 /* This function is used by idr_for_each(). If net is equal to peer, the
183 * function returns the id so that idr_for_each() stops. Because we cannot
184 * returns the id 0 (idr_for_each() will not stop), we return the magic value
185 * NET_ID_ZERO (-1) for it.
187 #define NET_ID_ZERO -1
188 static int net_eq_idr(int id
, void *net
, void *peer
)
190 if (net_eq(net
, peer
))
191 return id
? : NET_ID_ZERO
;
195 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
196 * is set to true, thus the caller knows that the new id must be notified via
199 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
201 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
202 bool alloc_it
= *alloc
;
206 /* Magic value for id 0. */
207 if (id
== NET_ID_ZERO
)
213 id
= alloc_netid(net
, peer
, -1);
215 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
218 return NETNSA_NSID_NOT_ASSIGNED
;
221 /* should be called with nsid_lock held */
222 static int __peernet2id(struct net
*net
, struct net
*peer
)
226 return __peernet2id_alloc(net
, peer
, &no
);
229 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
230 /* This function returns the id of a peer netns. If no id is assigned, one will
231 * be allocated and returned.
233 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
235 bool alloc
= false, alive
= false;
238 if (refcount_read(&net
->count
) == 0)
239 return NETNSA_NSID_NOT_ASSIGNED
;
240 spin_lock_bh(&net
->nsid_lock
);
242 * When peer is obtained from RCU lists, we may race with
243 * its cleanup. Check whether it's alive, and this guarantees
244 * we never hash a peer back to net->netns_ids, after it has
245 * just been idr_remove()'d from there in cleanup_net().
247 if (maybe_get_net(peer
))
248 alive
= alloc
= true;
249 id
= __peernet2id_alloc(net
, peer
, &alloc
);
250 spin_unlock_bh(&net
->nsid_lock
);
251 if (alloc
&& id
>= 0)
252 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
257 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
259 /* This function returns, if assigned, the id of a peer netns. */
260 int peernet2id(struct net
*net
, struct net
*peer
)
264 spin_lock_bh(&net
->nsid_lock
);
265 id
= __peernet2id(net
, peer
);
266 spin_unlock_bh(&net
->nsid_lock
);
269 EXPORT_SYMBOL(peernet2id
);
271 /* This function returns true is the peer netns has an id assigned into the
274 bool peernet_has_id(struct net
*net
, struct net
*peer
)
276 return peernet2id(net
, peer
) >= 0;
279 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
287 peer
= idr_find(&net
->netns_ids
, id
);
289 peer
= maybe_get_net(peer
);
296 * setup_net runs the initializers for the network namespace object.
298 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
300 /* Must be called with pernet_ops_rwsem held */
301 const struct pernet_operations
*ops
, *saved_ops
;
303 LIST_HEAD(net_exit_list
);
305 refcount_set(&net
->count
, 1);
306 refcount_set(&net
->passive
, 1);
307 get_random_bytes(&net
->hash_mix
, sizeof(u32
));
308 net
->dev_base_seq
= 1;
309 net
->user_ns
= user_ns
;
310 idr_init(&net
->netns_ids
);
311 spin_lock_init(&net
->nsid_lock
);
312 mutex_init(&net
->ipv4
.ra_mutex
);
314 list_for_each_entry(ops
, &pernet_list
, list
) {
315 error
= ops_init(ops
, net
);
319 down_write(&net_rwsem
);
320 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
321 up_write(&net_rwsem
);
326 /* Walk through the list backwards calling the exit functions
327 * for the pernet modules whose init functions did not fail.
329 list_add(&net
->exit_list
, &net_exit_list
);
331 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
332 ops_exit_list(ops
, &net_exit_list
);
335 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
336 ops_free_list(ops
, &net_exit_list
);
342 static int __net_init
net_defaults_init_net(struct net
*net
)
344 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
348 static struct pernet_operations net_defaults_ops
= {
349 .init
= net_defaults_init_net
,
352 static __init
int net_defaults_init(void)
354 if (register_pernet_subsys(&net_defaults_ops
))
355 panic("Cannot initialize net default settings");
360 core_initcall(net_defaults_init
);
363 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
365 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
368 static void dec_net_namespaces(struct ucounts
*ucounts
)
370 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
373 static struct kmem_cache
*net_cachep __ro_after_init
;
374 static struct workqueue_struct
*netns_wq
;
376 static struct net
*net_alloc(void)
378 struct net
*net
= NULL
;
379 struct net_generic
*ng
;
381 ng
= net_alloc_generic();
385 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
389 rcu_assign_pointer(net
->gen
, ng
);
398 static void net_free(struct net
*net
)
400 kfree(rcu_access_pointer(net
->gen
));
401 kmem_cache_free(net_cachep
, net
);
404 void net_drop_ns(void *p
)
407 if (ns
&& refcount_dec_and_test(&ns
->passive
))
411 struct net
*copy_net_ns(unsigned long flags
,
412 struct user_namespace
*user_ns
, struct net
*old_net
)
414 struct ucounts
*ucounts
;
418 if (!(flags
& CLONE_NEWNET
))
419 return get_net(old_net
);
421 ucounts
= inc_net_namespaces(user_ns
);
423 return ERR_PTR(-ENOSPC
);
430 refcount_set(&net
->passive
, 1);
431 net
->ucounts
= ucounts
;
432 get_user_ns(user_ns
);
434 rv
= down_read_killable(&pernet_ops_rwsem
);
438 rv
= setup_net(net
, user_ns
);
440 up_read(&pernet_ops_rwsem
);
444 put_user_ns(user_ns
);
447 dec_net_namespaces(ucounts
);
454 * net_ns_get_ownership - get sysfs ownership data for @net
455 * @net: network namespace in question (can be NULL)
456 * @uid: kernel user ID for sysfs objects
457 * @gid: kernel group ID for sysfs objects
459 * Returns the uid/gid pair of root in the user namespace associated with the
460 * given network namespace.
462 void net_ns_get_ownership(const struct net
*net
, kuid_t
*uid
, kgid_t
*gid
)
465 kuid_t ns_root_uid
= make_kuid(net
->user_ns
, 0);
466 kgid_t ns_root_gid
= make_kgid(net
->user_ns
, 0);
468 if (uid_valid(ns_root_uid
))
471 if (gid_valid(ns_root_gid
))
474 *uid
= GLOBAL_ROOT_UID
;
475 *gid
= GLOBAL_ROOT_GID
;
478 EXPORT_SYMBOL_GPL(net_ns_get_ownership
);
480 static void unhash_nsid(struct net
*net
, struct net
*last
)
483 /* This function is only called from cleanup_net() work,
484 * and this work is the only process, that may delete
485 * a net from net_namespace_list. So, when the below
486 * is executing, the list may only grow. Thus, we do not
487 * use for_each_net_rcu() or net_rwsem.
492 spin_lock_bh(&tmp
->nsid_lock
);
493 id
= __peernet2id(tmp
, net
);
495 idr_remove(&tmp
->netns_ids
, id
);
496 spin_unlock_bh(&tmp
->nsid_lock
);
498 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
502 spin_lock_bh(&net
->nsid_lock
);
503 idr_destroy(&net
->netns_ids
);
504 spin_unlock_bh(&net
->nsid_lock
);
507 static LLIST_HEAD(cleanup_list
);
509 static void cleanup_net(struct work_struct
*work
)
511 const struct pernet_operations
*ops
;
512 struct net
*net
, *tmp
, *last
;
513 struct llist_node
*net_kill_list
;
514 LIST_HEAD(net_exit_list
);
516 /* Atomically snapshot the list of namespaces to cleanup */
517 net_kill_list
= llist_del_all(&cleanup_list
);
519 down_read(&pernet_ops_rwsem
);
521 /* Don't let anyone else find us. */
522 down_write(&net_rwsem
);
523 llist_for_each_entry(net
, net_kill_list
, cleanup_list
)
524 list_del_rcu(&net
->list
);
525 /* Cache last net. After we unlock rtnl, no one new net
526 * added to net_namespace_list can assign nsid pointer
527 * to a net from net_kill_list (see peernet2id_alloc()).
528 * So, we skip them in unhash_nsid().
530 * Note, that unhash_nsid() does not delete nsid links
531 * between net_kill_list's nets, as they've already
532 * deleted from net_namespace_list. But, this would be
533 * useless anyway, as netns_ids are destroyed there.
535 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
536 up_write(&net_rwsem
);
538 llist_for_each_entry(net
, net_kill_list
, cleanup_list
) {
539 unhash_nsid(net
, last
);
540 list_add_tail(&net
->exit_list
, &net_exit_list
);
544 * Another CPU might be rcu-iterating the list, wait for it.
545 * This needs to be before calling the exit() notifiers, so
546 * the rcu_barrier() below isn't sufficient alone.
550 /* Run all of the network namespace exit methods */
551 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
552 ops_exit_list(ops
, &net_exit_list
);
554 /* Free the net generic variables */
555 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
556 ops_free_list(ops
, &net_exit_list
);
558 up_read(&pernet_ops_rwsem
);
560 /* Ensure there are no outstanding rcu callbacks using this
565 /* Finally it is safe to free my network namespace structure */
566 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
567 list_del_init(&net
->exit_list
);
568 dec_net_namespaces(net
->ucounts
);
569 put_user_ns(net
->user_ns
);
575 * net_ns_barrier - wait until concurrent net_cleanup_work is done
577 * cleanup_net runs from work queue and will first remove namespaces
578 * from the global list, then run net exit functions.
580 * Call this in module exit path to make sure that all netns
581 * ->exit ops have been invoked before the function is removed.
583 void net_ns_barrier(void)
585 down_write(&pernet_ops_rwsem
);
586 up_write(&pernet_ops_rwsem
);
588 EXPORT_SYMBOL(net_ns_barrier
);
590 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
592 void __put_net(struct net
*net
)
594 /* Cleanup the network namespace in process context */
595 if (llist_add(&net
->cleanup_list
, &cleanup_list
))
596 queue_work(netns_wq
, &net_cleanup_work
);
598 EXPORT_SYMBOL_GPL(__put_net
);
600 struct net
*get_net_ns_by_fd(int fd
)
603 struct ns_common
*ns
;
606 file
= proc_ns_fget(fd
);
608 return ERR_CAST(file
);
610 ns
= get_proc_ns(file_inode(file
));
611 if (ns
->ops
== &netns_operations
)
612 net
= get_net(container_of(ns
, struct net
, ns
));
614 net
= ERR_PTR(-EINVAL
);
621 struct net
*get_net_ns_by_fd(int fd
)
623 return ERR_PTR(-EINVAL
);
626 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
628 struct net
*get_net_ns_by_pid(pid_t pid
)
630 struct task_struct
*tsk
;
633 /* Lookup the network namespace */
634 net
= ERR_PTR(-ESRCH
);
636 tsk
= find_task_by_vpid(pid
);
638 struct nsproxy
*nsproxy
;
640 nsproxy
= tsk
->nsproxy
;
642 net
= get_net(nsproxy
->net_ns
);
648 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
650 static __net_init
int net_ns_net_init(struct net
*net
)
653 net
->ns
.ops
= &netns_operations
;
655 return ns_alloc_inum(&net
->ns
);
658 static __net_exit
void net_ns_net_exit(struct net
*net
)
660 ns_free_inum(&net
->ns
);
663 static struct pernet_operations __net_initdata net_ns_ops
= {
664 .init
= net_ns_net_init
,
665 .exit
= net_ns_net_exit
,
668 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
669 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
670 [NETNSA_NSID
] = { .type
= NLA_S32
},
671 [NETNSA_PID
] = { .type
= NLA_U32
},
672 [NETNSA_FD
] = { .type
= NLA_U32
},
673 [NETNSA_TARGET_NSID
] = { .type
= NLA_S32
},
676 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
677 struct netlink_ext_ack
*extack
)
679 struct net
*net
= sock_net(skb
->sk
);
680 struct nlattr
*tb
[NETNSA_MAX
+ 1];
685 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
686 rtnl_net_policy
, extack
);
689 if (!tb
[NETNSA_NSID
]) {
690 NL_SET_ERR_MSG(extack
, "nsid is missing");
693 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
695 if (tb
[NETNSA_PID
]) {
696 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
697 nla
= tb
[NETNSA_PID
];
698 } else if (tb
[NETNSA_FD
]) {
699 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
702 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
706 NL_SET_BAD_ATTR(extack
, nla
);
707 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
708 return PTR_ERR(peer
);
711 spin_lock_bh(&net
->nsid_lock
);
712 if (__peernet2id(net
, peer
) >= 0) {
713 spin_unlock_bh(&net
->nsid_lock
);
715 NL_SET_BAD_ATTR(extack
, nla
);
716 NL_SET_ERR_MSG(extack
,
717 "Peer netns already has a nsid assigned");
721 err
= alloc_netid(net
, peer
, nsid
);
722 spin_unlock_bh(&net
->nsid_lock
);
724 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
726 } else if (err
== -ENOSPC
&& nsid
>= 0) {
728 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
729 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
736 static int rtnl_net_get_size(void)
738 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
739 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
740 + nla_total_size(sizeof(s32
)) /* NETNSA_CURRENT_NSID */
744 struct net_fill_args
{
754 static int rtnl_net_fill(struct sk_buff
*skb
, struct net_fill_args
*args
)
756 struct nlmsghdr
*nlh
;
757 struct rtgenmsg
*rth
;
759 nlh
= nlmsg_put(skb
, args
->portid
, args
->seq
, args
->cmd
, sizeof(*rth
),
764 rth
= nlmsg_data(nlh
);
765 rth
->rtgen_family
= AF_UNSPEC
;
767 if (nla_put_s32(skb
, NETNSA_NSID
, args
->nsid
))
768 goto nla_put_failure
;
771 nla_put_s32(skb
, NETNSA_CURRENT_NSID
, args
->ref_nsid
))
772 goto nla_put_failure
;
778 nlmsg_cancel(skb
, nlh
);
782 static int rtnl_net_valid_getid_req(struct sk_buff
*skb
,
783 const struct nlmsghdr
*nlh
,
785 struct netlink_ext_ack
*extack
)
789 if (!netlink_strict_get_check(skb
))
790 return nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
791 rtnl_net_policy
, extack
);
793 err
= nlmsg_parse_strict(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
794 rtnl_net_policy
, extack
);
798 for (i
= 0; i
<= NETNSA_MAX
; i
++) {
806 case NETNSA_TARGET_NSID
:
809 NL_SET_ERR_MSG(extack
, "Unsupported attribute in peer netns getid request");
817 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
818 struct netlink_ext_ack
*extack
)
820 struct net
*net
= sock_net(skb
->sk
);
821 struct nlattr
*tb
[NETNSA_MAX
+ 1];
822 struct net_fill_args fillargs
= {
823 .portid
= NETLINK_CB(skb
).portid
,
824 .seq
= nlh
->nlmsg_seq
,
827 struct net
*peer
, *target
= net
;
832 err
= rtnl_net_valid_getid_req(skb
, nlh
, tb
, extack
);
835 if (tb
[NETNSA_PID
]) {
836 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
837 nla
= tb
[NETNSA_PID
];
838 } else if (tb
[NETNSA_FD
]) {
839 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
841 } else if (tb
[NETNSA_NSID
]) {
842 peer
= get_net_ns_by_id(net
, nla_get_u32(tb
[NETNSA_NSID
]));
844 peer
= ERR_PTR(-ENOENT
);
845 nla
= tb
[NETNSA_NSID
];
847 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
852 NL_SET_BAD_ATTR(extack
, nla
);
853 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
854 return PTR_ERR(peer
);
857 if (tb
[NETNSA_TARGET_NSID
]) {
858 int id
= nla_get_s32(tb
[NETNSA_TARGET_NSID
]);
860 target
= rtnl_get_net_ns_capable(NETLINK_CB(skb
).sk
, id
);
861 if (IS_ERR(target
)) {
862 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_TARGET_NSID
]);
863 NL_SET_ERR_MSG(extack
,
864 "Target netns reference is invalid");
865 err
= PTR_ERR(target
);
868 fillargs
.add_ref
= true;
869 fillargs
.ref_nsid
= peernet2id(net
, peer
);
872 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
878 fillargs
.nsid
= peernet2id(target
, peer
);
879 err
= rtnl_net_fill(msg
, &fillargs
);
883 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
889 if (fillargs
.add_ref
)
895 struct rtnl_net_dump_cb
{
899 struct net_fill_args fillargs
;
904 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
906 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
909 if (net_cb
->idx
< net_cb
->s_idx
)
912 net_cb
->fillargs
.nsid
= id
;
913 if (net_cb
->fillargs
.add_ref
)
914 net_cb
->fillargs
.ref_nsid
= __peernet2id(net_cb
->ref_net
, peer
);
915 ret
= rtnl_net_fill(net_cb
->skb
, &net_cb
->fillargs
);
924 static int rtnl_valid_dump_net_req(const struct nlmsghdr
*nlh
, struct sock
*sk
,
925 struct rtnl_net_dump_cb
*net_cb
,
926 struct netlink_callback
*cb
)
928 struct netlink_ext_ack
*extack
= cb
->extack
;
929 struct nlattr
*tb
[NETNSA_MAX
+ 1];
932 err
= nlmsg_parse_strict(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
933 rtnl_net_policy
, extack
);
937 for (i
= 0; i
<= NETNSA_MAX
; i
++) {
941 if (i
== NETNSA_TARGET_NSID
) {
944 net
= rtnl_get_net_ns_capable(sk
, nla_get_s32(tb
[i
]));
946 NL_SET_BAD_ATTR(extack
, tb
[i
]);
947 NL_SET_ERR_MSG(extack
,
948 "Invalid target network namespace id");
951 net_cb
->fillargs
.add_ref
= true;
952 net_cb
->ref_net
= net_cb
->tgt_net
;
953 net_cb
->tgt_net
= net
;
955 NL_SET_BAD_ATTR(extack
, tb
[i
]);
956 NL_SET_ERR_MSG(extack
,
957 "Unsupported attribute in dump request");
965 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
967 struct rtnl_net_dump_cb net_cb
= {
968 .tgt_net
= sock_net(skb
->sk
),
971 .portid
= NETLINK_CB(cb
->skb
).portid
,
972 .seq
= cb
->nlh
->nlmsg_seq
,
973 .flags
= NLM_F_MULTI
,
977 .s_idx
= cb
->args
[0],
981 if (cb
->strict_check
) {
982 err
= rtnl_valid_dump_net_req(cb
->nlh
, skb
->sk
, &net_cb
, cb
);
987 spin_lock_bh(&net_cb
.tgt_net
->nsid_lock
);
988 if (net_cb
.fillargs
.add_ref
&&
989 !net_eq(net_cb
.ref_net
, net_cb
.tgt_net
) &&
990 !spin_trylock_bh(&net_cb
.ref_net
->nsid_lock
)) {
991 spin_unlock_bh(&net_cb
.tgt_net
->nsid_lock
);
995 idr_for_each(&net_cb
.tgt_net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
996 if (net_cb
.fillargs
.add_ref
&&
997 !net_eq(net_cb
.ref_net
, net_cb
.tgt_net
))
998 spin_unlock_bh(&net_cb
.ref_net
->nsid_lock
);
999 spin_unlock_bh(&net_cb
.tgt_net
->nsid_lock
);
1001 cb
->args
[0] = net_cb
.idx
;
1003 if (net_cb
.fillargs
.add_ref
)
1004 put_net(net_cb
.tgt_net
);
1005 return err
< 0 ? err
: skb
->len
;
1008 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
1010 struct net_fill_args fillargs
= {
1014 struct sk_buff
*msg
;
1017 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
1021 err
= rtnl_net_fill(msg
, &fillargs
);
1025 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
1031 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
1034 static int __init
net_ns_init(void)
1036 struct net_generic
*ng
;
1038 #ifdef CONFIG_NET_NS
1039 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
1041 SLAB_PANIC
|SLAB_ACCOUNT
, NULL
);
1043 /* Create workqueue for cleanup */
1044 netns_wq
= create_singlethread_workqueue("netns");
1046 panic("Could not create netns workq");
1049 ng
= net_alloc_generic();
1051 panic("Could not allocate generic netns");
1053 rcu_assign_pointer(init_net
.gen
, ng
);
1055 down_write(&pernet_ops_rwsem
);
1056 if (setup_net(&init_net
, &init_user_ns
))
1057 panic("Could not setup the initial network namespace");
1059 init_net_initialized
= true;
1060 up_write(&pernet_ops_rwsem
);
1062 if (register_pernet_subsys(&net_ns_ops
))
1063 panic("Could not register network namespace subsystems");
1065 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
1066 RTNL_FLAG_DOIT_UNLOCKED
);
1067 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
1068 RTNL_FLAG_DOIT_UNLOCKED
);
1073 pure_initcall(net_ns_init
);
1075 #ifdef CONFIG_NET_NS
1076 static int __register_pernet_operations(struct list_head
*list
,
1077 struct pernet_operations
*ops
)
1081 LIST_HEAD(net_exit_list
);
1083 list_add_tail(&ops
->list
, list
);
1084 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
1085 /* We held write locked pernet_ops_rwsem, and parallel
1086 * setup_net() and cleanup_net() are not possible.
1089 error
= ops_init(ops
, net
);
1092 list_add_tail(&net
->exit_list
, &net_exit_list
);
1098 /* If I have an error cleanup all namespaces I initialized */
1099 list_del(&ops
->list
);
1100 ops_exit_list(ops
, &net_exit_list
);
1101 ops_free_list(ops
, &net_exit_list
);
1105 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
1108 LIST_HEAD(net_exit_list
);
1110 list_del(&ops
->list
);
1111 /* See comment in __register_pernet_operations() */
1113 list_add_tail(&net
->exit_list
, &net_exit_list
);
1114 ops_exit_list(ops
, &net_exit_list
);
1115 ops_free_list(ops
, &net_exit_list
);
1120 static int __register_pernet_operations(struct list_head
*list
,
1121 struct pernet_operations
*ops
)
1123 if (!init_net_initialized
) {
1124 list_add_tail(&ops
->list
, list
);
1128 return ops_init(ops
, &init_net
);
1131 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
1133 if (!init_net_initialized
) {
1134 list_del(&ops
->list
);
1136 LIST_HEAD(net_exit_list
);
1137 list_add(&init_net
.exit_list
, &net_exit_list
);
1138 ops_exit_list(ops
, &net_exit_list
);
1139 ops_free_list(ops
, &net_exit_list
);
1143 #endif /* CONFIG_NET_NS */
1145 static DEFINE_IDA(net_generic_ids
);
1147 static int register_pernet_operations(struct list_head
*list
,
1148 struct pernet_operations
*ops
)
1153 error
= ida_alloc_min(&net_generic_ids
, MIN_PERNET_OPS_ID
,
1158 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
1160 error
= __register_pernet_operations(list
, ops
);
1164 ida_free(&net_generic_ids
, *ops
->id
);
1170 static void unregister_pernet_operations(struct pernet_operations
*ops
)
1172 __unregister_pernet_operations(ops
);
1175 ida_free(&net_generic_ids
, *ops
->id
);
1179 * register_pernet_subsys - register a network namespace subsystem
1180 * @ops: pernet operations structure for the subsystem
1182 * Register a subsystem which has init and exit functions
1183 * that are called when network namespaces are created and
1184 * destroyed respectively.
1186 * When registered all network namespace init functions are
1187 * called for every existing network namespace. Allowing kernel
1188 * modules to have a race free view of the set of network namespaces.
1190 * When a new network namespace is created all of the init
1191 * methods are called in the order in which they were registered.
1193 * When a network namespace is destroyed all of the exit methods
1194 * are called in the reverse of the order with which they were
1197 int register_pernet_subsys(struct pernet_operations
*ops
)
1200 down_write(&pernet_ops_rwsem
);
1201 error
= register_pernet_operations(first_device
, ops
);
1202 up_write(&pernet_ops_rwsem
);
1205 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1208 * unregister_pernet_subsys - unregister a network namespace subsystem
1209 * @ops: pernet operations structure to manipulate
1211 * Remove the pernet operations structure from the list to be
1212 * used when network namespaces are created or destroyed. In
1213 * addition run the exit method for all existing network
1216 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1218 down_write(&pernet_ops_rwsem
);
1219 unregister_pernet_operations(ops
);
1220 up_write(&pernet_ops_rwsem
);
1222 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1225 * register_pernet_device - register a network namespace device
1226 * @ops: pernet operations structure for the subsystem
1228 * Register a device which has init and exit functions
1229 * that are called when network namespaces are created and
1230 * destroyed respectively.
1232 * When registered all network namespace init functions are
1233 * called for every existing network namespace. Allowing kernel
1234 * modules to have a race free view of the set of network namespaces.
1236 * When a new network namespace is created all of the init
1237 * methods are called in the order in which they were registered.
1239 * When a network namespace is destroyed all of the exit methods
1240 * are called in the reverse of the order with which they were
1243 int register_pernet_device(struct pernet_operations
*ops
)
1246 down_write(&pernet_ops_rwsem
);
1247 error
= register_pernet_operations(&pernet_list
, ops
);
1248 if (!error
&& (first_device
== &pernet_list
))
1249 first_device
= &ops
->list
;
1250 up_write(&pernet_ops_rwsem
);
1253 EXPORT_SYMBOL_GPL(register_pernet_device
);
1256 * unregister_pernet_device - unregister a network namespace netdevice
1257 * @ops: pernet operations structure to manipulate
1259 * Remove the pernet operations structure from the list to be
1260 * used when network namespaces are created or destroyed. In
1261 * addition run the exit method for all existing network
1264 void unregister_pernet_device(struct pernet_operations
*ops
)
1266 down_write(&pernet_ops_rwsem
);
1267 if (&ops
->list
== first_device
)
1268 first_device
= first_device
->next
;
1269 unregister_pernet_operations(ops
);
1270 up_write(&pernet_ops_rwsem
);
1272 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1274 #ifdef CONFIG_NET_NS
1275 static struct ns_common
*netns_get(struct task_struct
*task
)
1277 struct net
*net
= NULL
;
1278 struct nsproxy
*nsproxy
;
1281 nsproxy
= task
->nsproxy
;
1283 net
= get_net(nsproxy
->net_ns
);
1286 return net
? &net
->ns
: NULL
;
1289 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1291 return container_of(ns
, struct net
, ns
);
1294 static void netns_put(struct ns_common
*ns
)
1296 put_net(to_net_ns(ns
));
1299 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1301 struct net
*net
= to_net_ns(ns
);
1303 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1304 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1307 put_net(nsproxy
->net_ns
);
1308 nsproxy
->net_ns
= get_net(net
);
1312 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1314 return to_net_ns(ns
)->user_ns
;
1317 const struct proc_ns_operations netns_operations
= {
1319 .type
= CLONE_NEWNET
,
1322 .install
= netns_install
,
1323 .owner
= netns_owner
,