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>
22 #include <net/netlink.h>
23 #include <net/net_namespace.h>
24 #include <net/netns/generic.h>
27 * Our network namespace constructor/destructor lists
30 static LIST_HEAD(pernet_list
);
31 static struct list_head
*first_device
= &pernet_list
;
33 LIST_HEAD(net_namespace_list
);
34 EXPORT_SYMBOL_GPL(net_namespace_list
);
36 /* Protects net_namespace_list. Nests iside rtnl_lock() */
37 DECLARE_RWSEM(net_rwsem
);
38 EXPORT_SYMBOL_GPL(net_rwsem
);
40 struct net init_net
= {
41 .count
= REFCOUNT_INIT(1),
42 .dev_base_head
= LIST_HEAD_INIT(init_net
.dev_base_head
),
44 EXPORT_SYMBOL(init_net
);
46 static bool init_net_initialized
;
48 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
49 * init_net_initialized and first_device pointer.
50 * This is internal net namespace object. Please, don't use it
53 DECLARE_RWSEM(pernet_ops_rwsem
);
55 #define MIN_PERNET_OPS_ID \
56 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
58 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
60 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
62 static struct net_generic
*net_alloc_generic(void)
64 struct net_generic
*ng
;
65 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
67 ng
= kzalloc(generic_size
, GFP_KERNEL
);
69 ng
->s
.len
= max_gen_ptrs
;
74 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
76 struct net_generic
*ng
, *old_ng
;
78 BUG_ON(id
< MIN_PERNET_OPS_ID
);
80 old_ng
= rcu_dereference_protected(net
->gen
,
81 lockdep_is_held(&pernet_ops_rwsem
));
82 if (old_ng
->s
.len
> id
) {
83 old_ng
->ptr
[id
] = data
;
87 ng
= net_alloc_generic();
92 * Some synchronisation notes:
94 * The net_generic explores the net->gen array inside rcu
95 * read section. Besides once set the net->gen->ptr[x]
96 * pointer never changes (see rules in netns/generic.h).
98 * That said, we simply duplicate this array and schedule
99 * the old copy for kfree after a grace period.
102 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
103 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
106 rcu_assign_pointer(net
->gen
, ng
);
107 kfree_rcu(old_ng
, s
.rcu
);
111 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
116 if (ops
->id
&& ops
->size
) {
117 data
= kzalloc(ops
->size
, GFP_KERNEL
);
121 err
= net_assign_generic(net
, *ops
->id
, data
);
127 err
= ops
->init(net
);
138 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
140 if (ops
->id
&& ops
->size
) {
141 kfree(net_generic(net
, *ops
->id
));
145 static void ops_exit_list(const struct pernet_operations
*ops
,
146 struct list_head
*net_exit_list
)
150 list_for_each_entry(net
, net_exit_list
, exit_list
)
154 ops
->exit_batch(net_exit_list
);
157 static void ops_free_list(const struct pernet_operations
*ops
,
158 struct list_head
*net_exit_list
)
161 if (ops
->size
&& ops
->id
) {
162 list_for_each_entry(net
, net_exit_list
, exit_list
)
167 /* should be called with nsid_lock held */
168 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
170 int min
= 0, max
= 0;
177 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
180 /* This function is used by idr_for_each(). If net is equal to peer, the
181 * function returns the id so that idr_for_each() stops. Because we cannot
182 * returns the id 0 (idr_for_each() will not stop), we return the magic value
183 * NET_ID_ZERO (-1) for it.
185 #define NET_ID_ZERO -1
186 static int net_eq_idr(int id
, void *net
, void *peer
)
188 if (net_eq(net
, peer
))
189 return id
? : NET_ID_ZERO
;
193 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
194 * is set to true, thus the caller knows that the new id must be notified via
197 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
199 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
200 bool alloc_it
= *alloc
;
204 /* Magic value for id 0. */
205 if (id
== NET_ID_ZERO
)
211 id
= alloc_netid(net
, peer
, -1);
213 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
216 return NETNSA_NSID_NOT_ASSIGNED
;
219 /* should be called with nsid_lock held */
220 static int __peernet2id(struct net
*net
, struct net
*peer
)
224 return __peernet2id_alloc(net
, peer
, &no
);
227 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
228 /* This function returns the id of a peer netns. If no id is assigned, one will
229 * be allocated and returned.
231 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
233 bool alloc
= false, alive
= false;
236 if (refcount_read(&net
->count
) == 0)
237 return NETNSA_NSID_NOT_ASSIGNED
;
238 spin_lock_bh(&net
->nsid_lock
);
240 * When peer is obtained from RCU lists, we may race with
241 * its cleanup. Check whether it's alive, and this guarantees
242 * we never hash a peer back to net->netns_ids, after it has
243 * just been idr_remove()'d from there in cleanup_net().
245 if (maybe_get_net(peer
))
246 alive
= alloc
= true;
247 id
= __peernet2id_alloc(net
, peer
, &alloc
);
248 spin_unlock_bh(&net
->nsid_lock
);
249 if (alloc
&& id
>= 0)
250 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
255 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
257 /* This function returns, if assigned, the id of a peer netns. */
258 int peernet2id(struct net
*net
, struct net
*peer
)
262 spin_lock_bh(&net
->nsid_lock
);
263 id
= __peernet2id(net
, peer
);
264 spin_unlock_bh(&net
->nsid_lock
);
267 EXPORT_SYMBOL(peernet2id
);
269 /* This function returns true is the peer netns has an id assigned into the
272 bool peernet_has_id(struct net
*net
, struct net
*peer
)
274 return peernet2id(net
, peer
) >= 0;
277 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
285 peer
= idr_find(&net
->netns_ids
, id
);
287 peer
= maybe_get_net(peer
);
294 * setup_net runs the initializers for the network namespace object.
296 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
298 /* Must be called with pernet_ops_rwsem held */
299 const struct pernet_operations
*ops
, *saved_ops
;
301 LIST_HEAD(net_exit_list
);
303 refcount_set(&net
->count
, 1);
304 refcount_set(&net
->passive
, 1);
305 net
->dev_base_seq
= 1;
306 net
->user_ns
= user_ns
;
307 idr_init(&net
->netns_ids
);
308 spin_lock_init(&net
->nsid_lock
);
309 mutex_init(&net
->ipv4
.ra_mutex
);
311 list_for_each_entry(ops
, &pernet_list
, list
) {
312 error
= ops_init(ops
, net
);
316 down_write(&net_rwsem
);
317 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
318 up_write(&net_rwsem
);
323 /* Walk through the list backwards calling the exit functions
324 * for the pernet modules whose init functions did not fail.
326 list_add(&net
->exit_list
, &net_exit_list
);
328 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
329 ops_exit_list(ops
, &net_exit_list
);
332 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
333 ops_free_list(ops
, &net_exit_list
);
339 static int __net_init
net_defaults_init_net(struct net
*net
)
341 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
345 static struct pernet_operations net_defaults_ops
= {
346 .init
= net_defaults_init_net
,
349 static __init
int net_defaults_init(void)
351 if (register_pernet_subsys(&net_defaults_ops
))
352 panic("Cannot initialize net default settings");
357 core_initcall(net_defaults_init
);
360 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
362 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
365 static void dec_net_namespaces(struct ucounts
*ucounts
)
367 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
370 static struct kmem_cache
*net_cachep __ro_after_init
;
371 static struct workqueue_struct
*netns_wq
;
373 static struct net
*net_alloc(void)
375 struct net
*net
= NULL
;
376 struct net_generic
*ng
;
378 ng
= net_alloc_generic();
382 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
386 rcu_assign_pointer(net
->gen
, ng
);
395 static void net_free(struct net
*net
)
397 kfree(rcu_access_pointer(net
->gen
));
398 kmem_cache_free(net_cachep
, net
);
401 void net_drop_ns(void *p
)
404 if (ns
&& refcount_dec_and_test(&ns
->passive
))
408 struct net
*copy_net_ns(unsigned long flags
,
409 struct user_namespace
*user_ns
, struct net
*old_net
)
411 struct ucounts
*ucounts
;
415 if (!(flags
& CLONE_NEWNET
))
416 return get_net(old_net
);
418 ucounts
= inc_net_namespaces(user_ns
);
420 return ERR_PTR(-ENOSPC
);
427 refcount_set(&net
->passive
, 1);
428 net
->ucounts
= ucounts
;
429 get_user_ns(user_ns
);
431 rv
= down_read_killable(&pernet_ops_rwsem
);
435 rv
= setup_net(net
, user_ns
);
437 up_read(&pernet_ops_rwsem
);
441 put_user_ns(user_ns
);
444 dec_net_namespaces(ucounts
);
450 static void unhash_nsid(struct net
*net
, struct net
*last
)
453 /* This function is only called from cleanup_net() work,
454 * and this work is the only process, that may delete
455 * a net from net_namespace_list. So, when the below
456 * is executing, the list may only grow. Thus, we do not
457 * use for_each_net_rcu() or net_rwsem.
462 spin_lock_bh(&tmp
->nsid_lock
);
463 id
= __peernet2id(tmp
, net
);
465 idr_remove(&tmp
->netns_ids
, id
);
466 spin_unlock_bh(&tmp
->nsid_lock
);
468 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
472 spin_lock_bh(&net
->nsid_lock
);
473 idr_destroy(&net
->netns_ids
);
474 spin_unlock_bh(&net
->nsid_lock
);
477 static LLIST_HEAD(cleanup_list
);
479 static void cleanup_net(struct work_struct
*work
)
481 const struct pernet_operations
*ops
;
482 struct net
*net
, *tmp
, *last
;
483 struct llist_node
*net_kill_list
;
484 LIST_HEAD(net_exit_list
);
486 /* Atomically snapshot the list of namespaces to cleanup */
487 net_kill_list
= llist_del_all(&cleanup_list
);
489 down_read(&pernet_ops_rwsem
);
491 /* Don't let anyone else find us. */
492 down_write(&net_rwsem
);
493 llist_for_each_entry(net
, net_kill_list
, cleanup_list
)
494 list_del_rcu(&net
->list
);
495 /* Cache last net. After we unlock rtnl, no one new net
496 * added to net_namespace_list can assign nsid pointer
497 * to a net from net_kill_list (see peernet2id_alloc()).
498 * So, we skip them in unhash_nsid().
500 * Note, that unhash_nsid() does not delete nsid links
501 * between net_kill_list's nets, as they've already
502 * deleted from net_namespace_list. But, this would be
503 * useless anyway, as netns_ids are destroyed there.
505 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
506 up_write(&net_rwsem
);
508 llist_for_each_entry(net
, net_kill_list
, cleanup_list
) {
509 unhash_nsid(net
, last
);
510 list_add_tail(&net
->exit_list
, &net_exit_list
);
514 * Another CPU might be rcu-iterating the list, wait for it.
515 * This needs to be before calling the exit() notifiers, so
516 * the rcu_barrier() below isn't sufficient alone.
520 /* Run all of the network namespace exit methods */
521 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
522 ops_exit_list(ops
, &net_exit_list
);
524 /* Free the net generic variables */
525 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
526 ops_free_list(ops
, &net_exit_list
);
528 up_read(&pernet_ops_rwsem
);
530 /* Ensure there are no outstanding rcu callbacks using this
535 /* Finally it is safe to free my network namespace structure */
536 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
537 list_del_init(&net
->exit_list
);
538 dec_net_namespaces(net
->ucounts
);
539 put_user_ns(net
->user_ns
);
545 * net_ns_barrier - wait until concurrent net_cleanup_work is done
547 * cleanup_net runs from work queue and will first remove namespaces
548 * from the global list, then run net exit functions.
550 * Call this in module exit path to make sure that all netns
551 * ->exit ops have been invoked before the function is removed.
553 void net_ns_barrier(void)
555 down_write(&pernet_ops_rwsem
);
556 up_write(&pernet_ops_rwsem
);
558 EXPORT_SYMBOL(net_ns_barrier
);
560 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
562 void __put_net(struct net
*net
)
564 /* Cleanup the network namespace in process context */
565 if (llist_add(&net
->cleanup_list
, &cleanup_list
))
566 queue_work(netns_wq
, &net_cleanup_work
);
568 EXPORT_SYMBOL_GPL(__put_net
);
570 struct net
*get_net_ns_by_fd(int fd
)
573 struct ns_common
*ns
;
576 file
= proc_ns_fget(fd
);
578 return ERR_CAST(file
);
580 ns
= get_proc_ns(file_inode(file
));
581 if (ns
->ops
== &netns_operations
)
582 net
= get_net(container_of(ns
, struct net
, ns
));
584 net
= ERR_PTR(-EINVAL
);
591 struct net
*get_net_ns_by_fd(int fd
)
593 return ERR_PTR(-EINVAL
);
596 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
598 struct net
*get_net_ns_by_pid(pid_t pid
)
600 struct task_struct
*tsk
;
603 /* Lookup the network namespace */
604 net
= ERR_PTR(-ESRCH
);
606 tsk
= find_task_by_vpid(pid
);
608 struct nsproxy
*nsproxy
;
610 nsproxy
= tsk
->nsproxy
;
612 net
= get_net(nsproxy
->net_ns
);
618 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
620 static __net_init
int net_ns_net_init(struct net
*net
)
623 net
->ns
.ops
= &netns_operations
;
625 return ns_alloc_inum(&net
->ns
);
628 static __net_exit
void net_ns_net_exit(struct net
*net
)
630 ns_free_inum(&net
->ns
);
633 static struct pernet_operations __net_initdata net_ns_ops
= {
634 .init
= net_ns_net_init
,
635 .exit
= net_ns_net_exit
,
638 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
639 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
640 [NETNSA_NSID
] = { .type
= NLA_S32
},
641 [NETNSA_PID
] = { .type
= NLA_U32
},
642 [NETNSA_FD
] = { .type
= NLA_U32
},
645 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
646 struct netlink_ext_ack
*extack
)
648 struct net
*net
= sock_net(skb
->sk
);
649 struct nlattr
*tb
[NETNSA_MAX
+ 1];
654 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
655 rtnl_net_policy
, extack
);
658 if (!tb
[NETNSA_NSID
]) {
659 NL_SET_ERR_MSG(extack
, "nsid is missing");
662 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
664 if (tb
[NETNSA_PID
]) {
665 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
666 nla
= tb
[NETNSA_PID
];
667 } else if (tb
[NETNSA_FD
]) {
668 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
671 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
675 NL_SET_BAD_ATTR(extack
, nla
);
676 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
677 return PTR_ERR(peer
);
680 spin_lock_bh(&net
->nsid_lock
);
681 if (__peernet2id(net
, peer
) >= 0) {
682 spin_unlock_bh(&net
->nsid_lock
);
684 NL_SET_BAD_ATTR(extack
, nla
);
685 NL_SET_ERR_MSG(extack
,
686 "Peer netns already has a nsid assigned");
690 err
= alloc_netid(net
, peer
, nsid
);
691 spin_unlock_bh(&net
->nsid_lock
);
693 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
695 } else if (err
== -ENOSPC
&& nsid
>= 0) {
697 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
698 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
705 static int rtnl_net_get_size(void)
707 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
708 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
712 static int rtnl_net_fill(struct sk_buff
*skb
, u32 portid
, u32 seq
, int flags
,
713 int cmd
, struct net
*net
, int nsid
)
715 struct nlmsghdr
*nlh
;
716 struct rtgenmsg
*rth
;
718 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rth
), flags
);
722 rth
= nlmsg_data(nlh
);
723 rth
->rtgen_family
= AF_UNSPEC
;
725 if (nla_put_s32(skb
, NETNSA_NSID
, nsid
))
726 goto nla_put_failure
;
732 nlmsg_cancel(skb
, nlh
);
736 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
737 struct netlink_ext_ack
*extack
)
739 struct net
*net
= sock_net(skb
->sk
);
740 struct nlattr
*tb
[NETNSA_MAX
+ 1];
746 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
747 rtnl_net_policy
, extack
);
750 if (tb
[NETNSA_PID
]) {
751 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
752 nla
= tb
[NETNSA_PID
];
753 } else if (tb
[NETNSA_FD
]) {
754 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
757 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
762 NL_SET_BAD_ATTR(extack
, nla
);
763 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
764 return PTR_ERR(peer
);
767 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
773 id
= peernet2id(net
, peer
);
774 err
= rtnl_net_fill(msg
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
775 RTM_NEWNSID
, net
, id
);
779 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
789 struct rtnl_net_dump_cb
{
792 struct netlink_callback
*cb
;
797 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
799 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
802 if (net_cb
->idx
< net_cb
->s_idx
)
805 ret
= rtnl_net_fill(net_cb
->skb
, NETLINK_CB(net_cb
->cb
->skb
).portid
,
806 net_cb
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
807 RTM_NEWNSID
, net_cb
->net
, id
);
816 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
818 struct net
*net
= sock_net(skb
->sk
);
819 struct rtnl_net_dump_cb net_cb
= {
824 .s_idx
= cb
->args
[0],
827 spin_lock_bh(&net
->nsid_lock
);
828 idr_for_each(&net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
829 spin_unlock_bh(&net
->nsid_lock
);
831 cb
->args
[0] = net_cb
.idx
;
835 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
840 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
844 err
= rtnl_net_fill(msg
, 0, 0, 0, cmd
, net
, id
);
848 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
854 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
857 static int __init
net_ns_init(void)
859 struct net_generic
*ng
;
862 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
864 SLAB_PANIC
|SLAB_ACCOUNT
, NULL
);
866 /* Create workqueue for cleanup */
867 netns_wq
= create_singlethread_workqueue("netns");
869 panic("Could not create netns workq");
872 ng
= net_alloc_generic();
874 panic("Could not allocate generic netns");
876 rcu_assign_pointer(init_net
.gen
, ng
);
878 down_write(&pernet_ops_rwsem
);
879 if (setup_net(&init_net
, &init_user_ns
))
880 panic("Could not setup the initial network namespace");
882 init_net_initialized
= true;
883 up_write(&pernet_ops_rwsem
);
885 register_pernet_subsys(&net_ns_ops
);
887 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
888 RTNL_FLAG_DOIT_UNLOCKED
);
889 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
890 RTNL_FLAG_DOIT_UNLOCKED
);
895 pure_initcall(net_ns_init
);
898 static int __register_pernet_operations(struct list_head
*list
,
899 struct pernet_operations
*ops
)
903 LIST_HEAD(net_exit_list
);
905 list_add_tail(&ops
->list
, list
);
906 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
907 /* We held write locked pernet_ops_rwsem, and parallel
908 * setup_net() and cleanup_net() are not possible.
911 error
= ops_init(ops
, net
);
914 list_add_tail(&net
->exit_list
, &net_exit_list
);
920 /* If I have an error cleanup all namespaces I initialized */
921 list_del(&ops
->list
);
922 ops_exit_list(ops
, &net_exit_list
);
923 ops_free_list(ops
, &net_exit_list
);
927 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
930 LIST_HEAD(net_exit_list
);
932 list_del(&ops
->list
);
933 /* See comment in __register_pernet_operations() */
935 list_add_tail(&net
->exit_list
, &net_exit_list
);
936 ops_exit_list(ops
, &net_exit_list
);
937 ops_free_list(ops
, &net_exit_list
);
942 static int __register_pernet_operations(struct list_head
*list
,
943 struct pernet_operations
*ops
)
945 if (!init_net_initialized
) {
946 list_add_tail(&ops
->list
, list
);
950 return ops_init(ops
, &init_net
);
953 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
955 if (!init_net_initialized
) {
956 list_del(&ops
->list
);
958 LIST_HEAD(net_exit_list
);
959 list_add(&init_net
.exit_list
, &net_exit_list
);
960 ops_exit_list(ops
, &net_exit_list
);
961 ops_free_list(ops
, &net_exit_list
);
965 #endif /* CONFIG_NET_NS */
967 static DEFINE_IDA(net_generic_ids
);
969 static int register_pernet_operations(struct list_head
*list
,
970 struct pernet_operations
*ops
)
976 error
= ida_get_new_above(&net_generic_ids
, MIN_PERNET_OPS_ID
, ops
->id
);
978 if (error
== -EAGAIN
) {
979 ida_pre_get(&net_generic_ids
, GFP_KERNEL
);
984 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
986 error
= __register_pernet_operations(list
, ops
);
990 ida_remove(&net_generic_ids
, *ops
->id
);
996 static void unregister_pernet_operations(struct pernet_operations
*ops
)
998 __unregister_pernet_operations(ops
);
1001 ida_remove(&net_generic_ids
, *ops
->id
);
1005 * register_pernet_subsys - register a network namespace subsystem
1006 * @ops: pernet operations structure for the subsystem
1008 * Register a subsystem which has init and exit functions
1009 * that are called when network namespaces are created and
1010 * destroyed respectively.
1012 * When registered all network namespace init functions are
1013 * called for every existing network namespace. Allowing kernel
1014 * modules to have a race free view of the set of network namespaces.
1016 * When a new network namespace is created all of the init
1017 * methods are called in the order in which they were registered.
1019 * When a network namespace is destroyed all of the exit methods
1020 * are called in the reverse of the order with which they were
1023 int register_pernet_subsys(struct pernet_operations
*ops
)
1026 down_write(&pernet_ops_rwsem
);
1027 error
= register_pernet_operations(first_device
, ops
);
1028 up_write(&pernet_ops_rwsem
);
1031 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1034 * unregister_pernet_subsys - unregister a network namespace subsystem
1035 * @ops: pernet operations structure to manipulate
1037 * Remove the pernet operations structure from the list to be
1038 * used when network namespaces are created or destroyed. In
1039 * addition run the exit method for all existing network
1042 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1044 down_write(&pernet_ops_rwsem
);
1045 unregister_pernet_operations(ops
);
1046 up_write(&pernet_ops_rwsem
);
1048 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1051 * register_pernet_device - register a network namespace device
1052 * @ops: pernet operations structure for the subsystem
1054 * Register a device which has init and exit functions
1055 * that are called when network namespaces are created and
1056 * destroyed respectively.
1058 * When registered all network namespace init functions are
1059 * called for every existing network namespace. Allowing kernel
1060 * modules to have a race free view of the set of network namespaces.
1062 * When a new network namespace is created all of the init
1063 * methods are called in the order in which they were registered.
1065 * When a network namespace is destroyed all of the exit methods
1066 * are called in the reverse of the order with which they were
1069 int register_pernet_device(struct pernet_operations
*ops
)
1072 down_write(&pernet_ops_rwsem
);
1073 error
= register_pernet_operations(&pernet_list
, ops
);
1074 if (!error
&& (first_device
== &pernet_list
))
1075 first_device
= &ops
->list
;
1076 up_write(&pernet_ops_rwsem
);
1079 EXPORT_SYMBOL_GPL(register_pernet_device
);
1082 * unregister_pernet_device - unregister a network namespace netdevice
1083 * @ops: pernet operations structure to manipulate
1085 * Remove the pernet operations structure from the list to be
1086 * used when network namespaces are created or destroyed. In
1087 * addition run the exit method for all existing network
1090 void unregister_pernet_device(struct pernet_operations
*ops
)
1092 down_write(&pernet_ops_rwsem
);
1093 if (&ops
->list
== first_device
)
1094 first_device
= first_device
->next
;
1095 unregister_pernet_operations(ops
);
1096 up_write(&pernet_ops_rwsem
);
1098 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1100 #ifdef CONFIG_NET_NS
1101 static struct ns_common
*netns_get(struct task_struct
*task
)
1103 struct net
*net
= NULL
;
1104 struct nsproxy
*nsproxy
;
1107 nsproxy
= task
->nsproxy
;
1109 net
= get_net(nsproxy
->net_ns
);
1112 return net
? &net
->ns
: NULL
;
1115 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1117 return container_of(ns
, struct net
, ns
);
1120 static void netns_put(struct ns_common
*ns
)
1122 put_net(to_net_ns(ns
));
1125 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1127 struct net
*net
= to_net_ns(ns
);
1129 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1130 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1133 put_net(nsproxy
->net_ns
);
1134 nsproxy
->net_ns
= get_net(net
);
1138 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1140 return to_net_ns(ns
)->user_ns
;
1143 const struct proc_ns_operations netns_operations
= {
1145 .type
= CLONE_NEWNET
,
1148 .install
= netns_install
,
1149 .owner
= netns_owner
,