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
;
32 DEFINE_MUTEX(net_mutex
);
34 LIST_HEAD(net_namespace_list
);
35 EXPORT_SYMBOL_GPL(net_namespace_list
);
37 struct net init_net
= {
38 .count
= ATOMIC_INIT(1),
39 .dev_base_head
= LIST_HEAD_INIT(init_net
.dev_base_head
),
41 EXPORT_SYMBOL(init_net
);
43 static bool init_net_initialized
;
45 #define MIN_PERNET_OPS_ID \
46 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
48 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
50 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
52 static struct net_generic
*net_alloc_generic(void)
54 struct net_generic
*ng
;
55 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
57 ng
= kzalloc(generic_size
, GFP_KERNEL
);
59 ng
->s
.len
= max_gen_ptrs
;
64 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
66 struct net_generic
*ng
, *old_ng
;
68 BUG_ON(!mutex_is_locked(&net_mutex
));
69 BUG_ON(id
< MIN_PERNET_OPS_ID
);
71 old_ng
= rcu_dereference_protected(net
->gen
,
72 lockdep_is_held(&net_mutex
));
73 if (old_ng
->s
.len
> id
) {
74 old_ng
->ptr
[id
] = data
;
78 ng
= net_alloc_generic();
83 * Some synchronisation notes:
85 * The net_generic explores the net->gen array inside rcu
86 * read section. Besides once set the net->gen->ptr[x]
87 * pointer never changes (see rules in netns/generic.h).
89 * That said, we simply duplicate this array and schedule
90 * the old copy for kfree after a grace period.
93 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
94 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
97 rcu_assign_pointer(net
->gen
, ng
);
98 kfree_rcu(old_ng
, s
.rcu
);
102 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
107 if (ops
->id
&& ops
->size
) {
108 data
= kzalloc(ops
->size
, GFP_KERNEL
);
112 err
= net_assign_generic(net
, *ops
->id
, data
);
118 err
= ops
->init(net
);
129 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
131 if (ops
->id
&& ops
->size
) {
132 kfree(net_generic(net
, *ops
->id
));
136 static void ops_exit_list(const struct pernet_operations
*ops
,
137 struct list_head
*net_exit_list
)
141 list_for_each_entry(net
, net_exit_list
, exit_list
)
145 ops
->exit_batch(net_exit_list
);
148 static void ops_free_list(const struct pernet_operations
*ops
,
149 struct list_head
*net_exit_list
)
152 if (ops
->size
&& ops
->id
) {
153 list_for_each_entry(net
, net_exit_list
, exit_list
)
158 /* should be called with nsid_lock held */
159 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
161 int min
= 0, max
= 0;
168 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
171 /* This function is used by idr_for_each(). If net is equal to peer, the
172 * function returns the id so that idr_for_each() stops. Because we cannot
173 * returns the id 0 (idr_for_each() will not stop), we return the magic value
174 * NET_ID_ZERO (-1) for it.
176 #define NET_ID_ZERO -1
177 static int net_eq_idr(int id
, void *net
, void *peer
)
179 if (net_eq(net
, peer
))
180 return id
? : NET_ID_ZERO
;
184 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
185 * is set to true, thus the caller knows that the new id must be notified via
188 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
190 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
191 bool alloc_it
= *alloc
;
195 /* Magic value for id 0. */
196 if (id
== NET_ID_ZERO
)
202 id
= alloc_netid(net
, peer
, -1);
204 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
207 return NETNSA_NSID_NOT_ASSIGNED
;
210 /* should be called with nsid_lock held */
211 static int __peernet2id(struct net
*net
, struct net
*peer
)
215 return __peernet2id_alloc(net
, peer
, &no
);
218 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
219 /* This function returns the id of a peer netns. If no id is assigned, one will
220 * be allocated and returned.
222 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
227 if (atomic_read(&net
->count
) == 0)
228 return NETNSA_NSID_NOT_ASSIGNED
;
229 spin_lock_bh(&net
->nsid_lock
);
230 alloc
= atomic_read(&peer
->count
) == 0 ? false : true;
231 id
= __peernet2id_alloc(net
, peer
, &alloc
);
232 spin_unlock_bh(&net
->nsid_lock
);
233 if (alloc
&& id
>= 0)
234 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
237 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
239 /* This function returns, if assigned, the id of a peer netns. */
240 int peernet2id(struct net
*net
, struct net
*peer
)
244 spin_lock_bh(&net
->nsid_lock
);
245 id
= __peernet2id(net
, peer
);
246 spin_unlock_bh(&net
->nsid_lock
);
249 EXPORT_SYMBOL(peernet2id
);
251 /* This function returns true is the peer netns has an id assigned into the
254 bool peernet_has_id(struct net
*net
, struct net
*peer
)
256 return peernet2id(net
, peer
) >= 0;
259 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
267 spin_lock_bh(&net
->nsid_lock
);
268 peer
= idr_find(&net
->netns_ids
, id
);
271 spin_unlock_bh(&net
->nsid_lock
);
278 * setup_net runs the initializers for the network namespace object.
280 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
282 /* Must be called with net_mutex held */
283 const struct pernet_operations
*ops
, *saved_ops
;
285 LIST_HEAD(net_exit_list
);
287 atomic_set(&net
->count
, 1);
288 refcount_set(&net
->passive
, 1);
289 net
->dev_base_seq
= 1;
290 net
->user_ns
= user_ns
;
291 idr_init(&net
->netns_ids
);
292 spin_lock_init(&net
->nsid_lock
);
294 list_for_each_entry(ops
, &pernet_list
, list
) {
295 error
= ops_init(ops
, net
);
303 /* Walk through the list backwards calling the exit functions
304 * for the pernet modules whose init functions did not fail.
306 list_add(&net
->exit_list
, &net_exit_list
);
308 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
309 ops_exit_list(ops
, &net_exit_list
);
312 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
313 ops_free_list(ops
, &net_exit_list
);
319 static int __net_init
net_defaults_init_net(struct net
*net
)
321 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
325 static struct pernet_operations net_defaults_ops
= {
326 .init
= net_defaults_init_net
,
329 static __init
int net_defaults_init(void)
331 if (register_pernet_subsys(&net_defaults_ops
))
332 panic("Cannot initialize net default settings");
337 core_initcall(net_defaults_init
);
340 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
342 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
345 static void dec_net_namespaces(struct ucounts
*ucounts
)
347 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
350 static struct kmem_cache
*net_cachep
;
351 static struct workqueue_struct
*netns_wq
;
353 static struct net
*net_alloc(void)
355 struct net
*net
= NULL
;
356 struct net_generic
*ng
;
358 ng
= net_alloc_generic();
362 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
366 rcu_assign_pointer(net
->gen
, ng
);
375 static void net_free(struct net
*net
)
377 kfree(rcu_access_pointer(net
->gen
));
378 kmem_cache_free(net_cachep
, net
);
381 void net_drop_ns(void *p
)
384 if (ns
&& refcount_dec_and_test(&ns
->passive
))
388 struct net
*copy_net_ns(unsigned long flags
,
389 struct user_namespace
*user_ns
, struct net
*old_net
)
391 struct ucounts
*ucounts
;
395 if (!(flags
& CLONE_NEWNET
))
396 return get_net(old_net
);
398 ucounts
= inc_net_namespaces(user_ns
);
400 return ERR_PTR(-ENOSPC
);
404 dec_net_namespaces(ucounts
);
405 return ERR_PTR(-ENOMEM
);
408 get_user_ns(user_ns
);
410 rv
= mutex_lock_killable(&net_mutex
);
413 dec_net_namespaces(ucounts
);
414 put_user_ns(user_ns
);
418 net
->ucounts
= ucounts
;
419 rv
= setup_net(net
, user_ns
);
422 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
425 mutex_unlock(&net_mutex
);
427 dec_net_namespaces(ucounts
);
428 put_user_ns(user_ns
);
435 static DEFINE_SPINLOCK(cleanup_list_lock
);
436 static LIST_HEAD(cleanup_list
); /* Must hold cleanup_list_lock to touch */
438 static void cleanup_net(struct work_struct
*work
)
440 const struct pernet_operations
*ops
;
441 struct net
*net
, *tmp
;
442 struct list_head net_kill_list
;
443 LIST_HEAD(net_exit_list
);
445 /* Atomically snapshot the list of namespaces to cleanup */
446 spin_lock_irq(&cleanup_list_lock
);
447 list_replace_init(&cleanup_list
, &net_kill_list
);
448 spin_unlock_irq(&cleanup_list_lock
);
450 mutex_lock(&net_mutex
);
452 /* Don't let anyone else find us. */
454 list_for_each_entry(net
, &net_kill_list
, cleanup_list
) {
455 list_del_rcu(&net
->list
);
456 list_add_tail(&net
->exit_list
, &net_exit_list
);
460 spin_lock_bh(&tmp
->nsid_lock
);
461 id
= __peernet2id(tmp
, net
);
463 idr_remove(&tmp
->netns_ids
, id
);
464 spin_unlock_bh(&tmp
->nsid_lock
);
466 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
468 spin_lock_bh(&net
->nsid_lock
);
469 idr_destroy(&net
->netns_ids
);
470 spin_unlock_bh(&net
->nsid_lock
);
476 * Another CPU might be rcu-iterating the list, wait for it.
477 * This needs to be before calling the exit() notifiers, so
478 * the rcu_barrier() below isn't sufficient alone.
482 /* Run all of the network namespace exit methods */
483 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
484 ops_exit_list(ops
, &net_exit_list
);
486 /* Free the net generic variables */
487 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
488 ops_free_list(ops
, &net_exit_list
);
490 mutex_unlock(&net_mutex
);
492 /* Ensure there are no outstanding rcu callbacks using this
497 /* Finally it is safe to free my network namespace structure */
498 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
499 list_del_init(&net
->exit_list
);
500 dec_net_namespaces(net
->ucounts
);
501 put_user_ns(net
->user_ns
);
507 * net_ns_barrier - wait until concurrent net_cleanup_work is done
509 * cleanup_net runs from work queue and will first remove namespaces
510 * from the global list, then run net exit functions.
512 * Call this in module exit path to make sure that all netns
513 * ->exit ops have been invoked before the function is removed.
515 void net_ns_barrier(void)
517 mutex_lock(&net_mutex
);
518 mutex_unlock(&net_mutex
);
520 EXPORT_SYMBOL(net_ns_barrier
);
522 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
524 void __put_net(struct net
*net
)
526 /* Cleanup the network namespace in process context */
529 spin_lock_irqsave(&cleanup_list_lock
, flags
);
530 list_add(&net
->cleanup_list
, &cleanup_list
);
531 spin_unlock_irqrestore(&cleanup_list_lock
, flags
);
533 queue_work(netns_wq
, &net_cleanup_work
);
535 EXPORT_SYMBOL_GPL(__put_net
);
537 struct net
*get_net_ns_by_fd(int fd
)
540 struct ns_common
*ns
;
543 file
= proc_ns_fget(fd
);
545 return ERR_CAST(file
);
547 ns
= get_proc_ns(file_inode(file
));
548 if (ns
->ops
== &netns_operations
)
549 net
= get_net(container_of(ns
, struct net
, ns
));
551 net
= ERR_PTR(-EINVAL
);
558 struct net
*get_net_ns_by_fd(int fd
)
560 return ERR_PTR(-EINVAL
);
563 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
565 struct net
*get_net_ns_by_pid(pid_t pid
)
567 struct task_struct
*tsk
;
570 /* Lookup the network namespace */
571 net
= ERR_PTR(-ESRCH
);
573 tsk
= find_task_by_vpid(pid
);
575 struct nsproxy
*nsproxy
;
577 nsproxy
= tsk
->nsproxy
;
579 net
= get_net(nsproxy
->net_ns
);
585 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
587 static __net_init
int net_ns_net_init(struct net
*net
)
590 net
->ns
.ops
= &netns_operations
;
592 return ns_alloc_inum(&net
->ns
);
595 static __net_exit
void net_ns_net_exit(struct net
*net
)
597 ns_free_inum(&net
->ns
);
600 static struct pernet_operations __net_initdata net_ns_ops
= {
601 .init
= net_ns_net_init
,
602 .exit
= net_ns_net_exit
,
605 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
606 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
607 [NETNSA_NSID
] = { .type
= NLA_S32
},
608 [NETNSA_PID
] = { .type
= NLA_U32
},
609 [NETNSA_FD
] = { .type
= NLA_U32
},
612 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
613 struct netlink_ext_ack
*extack
)
615 struct net
*net
= sock_net(skb
->sk
);
616 struct nlattr
*tb
[NETNSA_MAX
+ 1];
621 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
622 rtnl_net_policy
, extack
);
625 if (!tb
[NETNSA_NSID
]) {
626 NL_SET_ERR_MSG(extack
, "nsid is missing");
629 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
631 if (tb
[NETNSA_PID
]) {
632 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
633 nla
= tb
[NETNSA_PID
];
634 } else if (tb
[NETNSA_FD
]) {
635 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
638 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
642 NL_SET_BAD_ATTR(extack
, nla
);
643 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
644 return PTR_ERR(peer
);
647 spin_lock_bh(&net
->nsid_lock
);
648 if (__peernet2id(net
, peer
) >= 0) {
649 spin_unlock_bh(&net
->nsid_lock
);
651 NL_SET_BAD_ATTR(extack
, nla
);
652 NL_SET_ERR_MSG(extack
,
653 "Peer netns already has a nsid assigned");
657 err
= alloc_netid(net
, peer
, nsid
);
658 spin_unlock_bh(&net
->nsid_lock
);
660 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
662 } else if (err
== -ENOSPC
&& nsid
>= 0) {
664 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
665 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
672 static int rtnl_net_get_size(void)
674 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
675 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
679 static int rtnl_net_fill(struct sk_buff
*skb
, u32 portid
, u32 seq
, int flags
,
680 int cmd
, struct net
*net
, int nsid
)
682 struct nlmsghdr
*nlh
;
683 struct rtgenmsg
*rth
;
685 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rth
), flags
);
689 rth
= nlmsg_data(nlh
);
690 rth
->rtgen_family
= AF_UNSPEC
;
692 if (nla_put_s32(skb
, NETNSA_NSID
, nsid
))
693 goto nla_put_failure
;
699 nlmsg_cancel(skb
, nlh
);
703 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
704 struct netlink_ext_ack
*extack
)
706 struct net
*net
= sock_net(skb
->sk
);
707 struct nlattr
*tb
[NETNSA_MAX
+ 1];
713 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
714 rtnl_net_policy
, extack
);
717 if (tb
[NETNSA_PID
]) {
718 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
719 nla
= tb
[NETNSA_PID
];
720 } else if (tb
[NETNSA_FD
]) {
721 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
724 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
729 NL_SET_BAD_ATTR(extack
, nla
);
730 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
731 return PTR_ERR(peer
);
734 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
740 id
= peernet2id(net
, peer
);
741 err
= rtnl_net_fill(msg
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
742 RTM_NEWNSID
, net
, id
);
746 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
756 struct rtnl_net_dump_cb
{
759 struct netlink_callback
*cb
;
764 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
766 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
769 if (net_cb
->idx
< net_cb
->s_idx
)
772 ret
= rtnl_net_fill(net_cb
->skb
, NETLINK_CB(net_cb
->cb
->skb
).portid
,
773 net_cb
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
774 RTM_NEWNSID
, net_cb
->net
, id
);
783 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
785 struct net
*net
= sock_net(skb
->sk
);
786 struct rtnl_net_dump_cb net_cb
= {
791 .s_idx
= cb
->args
[0],
794 spin_lock_bh(&net
->nsid_lock
);
795 idr_for_each(&net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
796 spin_unlock_bh(&net
->nsid_lock
);
798 cb
->args
[0] = net_cb
.idx
;
802 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
807 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
811 err
= rtnl_net_fill(msg
, 0, 0, 0, cmd
, net
, id
);
815 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
821 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
824 static int __init
net_ns_init(void)
826 struct net_generic
*ng
;
829 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
833 /* Create workqueue for cleanup */
834 netns_wq
= create_singlethread_workqueue("netns");
836 panic("Could not create netns workq");
839 ng
= net_alloc_generic();
841 panic("Could not allocate generic netns");
843 rcu_assign_pointer(init_net
.gen
, ng
);
845 mutex_lock(&net_mutex
);
846 if (setup_net(&init_net
, &init_user_ns
))
847 panic("Could not setup the initial network namespace");
849 init_net_initialized
= true;
852 list_add_tail_rcu(&init_net
.list
, &net_namespace_list
);
855 mutex_unlock(&net_mutex
);
857 register_pernet_subsys(&net_ns_ops
);
859 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
860 RTNL_FLAG_DOIT_UNLOCKED
);
861 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
862 RTNL_FLAG_DOIT_UNLOCKED
);
867 pure_initcall(net_ns_init
);
870 static int __register_pernet_operations(struct list_head
*list
,
871 struct pernet_operations
*ops
)
875 LIST_HEAD(net_exit_list
);
877 list_add_tail(&ops
->list
, list
);
878 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
880 error
= ops_init(ops
, net
);
883 list_add_tail(&net
->exit_list
, &net_exit_list
);
889 /* If I have an error cleanup all namespaces I initialized */
890 list_del(&ops
->list
);
891 ops_exit_list(ops
, &net_exit_list
);
892 ops_free_list(ops
, &net_exit_list
);
896 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
899 LIST_HEAD(net_exit_list
);
901 list_del(&ops
->list
);
903 list_add_tail(&net
->exit_list
, &net_exit_list
);
904 ops_exit_list(ops
, &net_exit_list
);
905 ops_free_list(ops
, &net_exit_list
);
910 static int __register_pernet_operations(struct list_head
*list
,
911 struct pernet_operations
*ops
)
913 if (!init_net_initialized
) {
914 list_add_tail(&ops
->list
, list
);
918 return ops_init(ops
, &init_net
);
921 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
923 if (!init_net_initialized
) {
924 list_del(&ops
->list
);
926 LIST_HEAD(net_exit_list
);
927 list_add(&init_net
.exit_list
, &net_exit_list
);
928 ops_exit_list(ops
, &net_exit_list
);
929 ops_free_list(ops
, &net_exit_list
);
933 #endif /* CONFIG_NET_NS */
935 static DEFINE_IDA(net_generic_ids
);
937 static int register_pernet_operations(struct list_head
*list
,
938 struct pernet_operations
*ops
)
944 error
= ida_get_new_above(&net_generic_ids
, MIN_PERNET_OPS_ID
, ops
->id
);
946 if (error
== -EAGAIN
) {
947 ida_pre_get(&net_generic_ids
, GFP_KERNEL
);
952 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
954 error
= __register_pernet_operations(list
, ops
);
958 ida_remove(&net_generic_ids
, *ops
->id
);
964 static void unregister_pernet_operations(struct pernet_operations
*ops
)
967 __unregister_pernet_operations(ops
);
970 ida_remove(&net_generic_ids
, *ops
->id
);
974 * register_pernet_subsys - register a network namespace subsystem
975 * @ops: pernet operations structure for the subsystem
977 * Register a subsystem which has init and exit functions
978 * that are called when network namespaces are created and
979 * destroyed respectively.
981 * When registered all network namespace init functions are
982 * called for every existing network namespace. Allowing kernel
983 * modules to have a race free view of the set of network namespaces.
985 * When a new network namespace is created all of the init
986 * methods are called in the order in which they were registered.
988 * When a network namespace is destroyed all of the exit methods
989 * are called in the reverse of the order with which they were
992 int register_pernet_subsys(struct pernet_operations
*ops
)
995 mutex_lock(&net_mutex
);
996 error
= register_pernet_operations(first_device
, ops
);
997 mutex_unlock(&net_mutex
);
1000 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1003 * unregister_pernet_subsys - unregister a network namespace subsystem
1004 * @ops: pernet operations structure to manipulate
1006 * Remove the pernet operations structure from the list to be
1007 * used when network namespaces are created or destroyed. In
1008 * addition run the exit method for all existing network
1011 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1013 mutex_lock(&net_mutex
);
1014 unregister_pernet_operations(ops
);
1015 mutex_unlock(&net_mutex
);
1017 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1020 * register_pernet_device - register a network namespace device
1021 * @ops: pernet operations structure for the subsystem
1023 * Register a device which has init and exit functions
1024 * that are called when network namespaces are created and
1025 * destroyed respectively.
1027 * When registered all network namespace init functions are
1028 * called for every existing network namespace. Allowing kernel
1029 * modules to have a race free view of the set of network namespaces.
1031 * When a new network namespace is created all of the init
1032 * methods are called in the order in which they were registered.
1034 * When a network namespace is destroyed all of the exit methods
1035 * are called in the reverse of the order with which they were
1038 int register_pernet_device(struct pernet_operations
*ops
)
1041 mutex_lock(&net_mutex
);
1042 error
= register_pernet_operations(&pernet_list
, ops
);
1043 if (!error
&& (first_device
== &pernet_list
))
1044 first_device
= &ops
->list
;
1045 mutex_unlock(&net_mutex
);
1048 EXPORT_SYMBOL_GPL(register_pernet_device
);
1051 * unregister_pernet_device - unregister a network namespace netdevice
1052 * @ops: pernet operations structure to manipulate
1054 * Remove the pernet operations structure from the list to be
1055 * used when network namespaces are created or destroyed. In
1056 * addition run the exit method for all existing network
1059 void unregister_pernet_device(struct pernet_operations
*ops
)
1061 mutex_lock(&net_mutex
);
1062 if (&ops
->list
== first_device
)
1063 first_device
= first_device
->next
;
1064 unregister_pernet_operations(ops
);
1065 mutex_unlock(&net_mutex
);
1067 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1069 #ifdef CONFIG_NET_NS
1070 static struct ns_common
*netns_get(struct task_struct
*task
)
1072 struct net
*net
= NULL
;
1073 struct nsproxy
*nsproxy
;
1076 nsproxy
= task
->nsproxy
;
1078 net
= get_net(nsproxy
->net_ns
);
1081 return net
? &net
->ns
: NULL
;
1084 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1086 return container_of(ns
, struct net
, ns
);
1089 static void netns_put(struct ns_common
*ns
)
1091 put_net(to_net_ns(ns
));
1094 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1096 struct net
*net
= to_net_ns(ns
);
1098 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1099 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1102 put_net(nsproxy
->net_ns
);
1103 nsproxy
->net_ns
= get_net(net
);
1107 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1109 return to_net_ns(ns
)->user_ns
;
1112 const struct proc_ns_operations netns_operations
= {
1114 .type
= CLONE_NEWNET
,
1117 .install
= netns_install
,
1118 .owner
= netns_owner
,