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
= REFCOUNT_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 * net_sem: protects: pernet_list, net_generic_ids,
46 * init_net_initialized and first_device pointer.
48 DECLARE_RWSEM(net_sem
);
50 #define MIN_PERNET_OPS_ID \
51 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
53 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
55 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
57 static struct net_generic
*net_alloc_generic(void)
59 struct net_generic
*ng
;
60 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
62 ng
= kzalloc(generic_size
, GFP_KERNEL
);
64 ng
->s
.len
= max_gen_ptrs
;
69 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
71 struct net_generic
*ng
, *old_ng
;
73 BUG_ON(!mutex_is_locked(&net_mutex
));
74 BUG_ON(id
< MIN_PERNET_OPS_ID
);
76 old_ng
= rcu_dereference_protected(net
->gen
,
77 lockdep_is_held(&net_mutex
));
78 if (old_ng
->s
.len
> id
) {
79 old_ng
->ptr
[id
] = data
;
83 ng
= net_alloc_generic();
88 * Some synchronisation notes:
90 * The net_generic explores the net->gen array inside rcu
91 * read section. Besides once set the net->gen->ptr[x]
92 * pointer never changes (see rules in netns/generic.h).
94 * That said, we simply duplicate this array and schedule
95 * the old copy for kfree after a grace period.
98 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
99 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
102 rcu_assign_pointer(net
->gen
, ng
);
103 kfree_rcu(old_ng
, s
.rcu
);
107 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
112 if (ops
->id
&& ops
->size
) {
113 data
= kzalloc(ops
->size
, GFP_KERNEL
);
117 err
= net_assign_generic(net
, *ops
->id
, data
);
123 err
= ops
->init(net
);
134 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
136 if (ops
->id
&& ops
->size
) {
137 kfree(net_generic(net
, *ops
->id
));
141 static void ops_exit_list(const struct pernet_operations
*ops
,
142 struct list_head
*net_exit_list
)
146 list_for_each_entry(net
, net_exit_list
, exit_list
)
150 ops
->exit_batch(net_exit_list
);
153 static void ops_free_list(const struct pernet_operations
*ops
,
154 struct list_head
*net_exit_list
)
157 if (ops
->size
&& ops
->id
) {
158 list_for_each_entry(net
, net_exit_list
, exit_list
)
163 /* should be called with nsid_lock held */
164 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
166 int min
= 0, max
= 0;
173 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
176 /* This function is used by idr_for_each(). If net is equal to peer, the
177 * function returns the id so that idr_for_each() stops. Because we cannot
178 * returns the id 0 (idr_for_each() will not stop), we return the magic value
179 * NET_ID_ZERO (-1) for it.
181 #define NET_ID_ZERO -1
182 static int net_eq_idr(int id
, void *net
, void *peer
)
184 if (net_eq(net
, peer
))
185 return id
? : NET_ID_ZERO
;
189 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
190 * is set to true, thus the caller knows that the new id must be notified via
193 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
195 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
196 bool alloc_it
= *alloc
;
200 /* Magic value for id 0. */
201 if (id
== NET_ID_ZERO
)
207 id
= alloc_netid(net
, peer
, -1);
209 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
212 return NETNSA_NSID_NOT_ASSIGNED
;
215 /* should be called with nsid_lock held */
216 static int __peernet2id(struct net
*net
, struct net
*peer
)
220 return __peernet2id_alloc(net
, peer
, &no
);
223 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
224 /* This function returns the id of a peer netns. If no id is assigned, one will
225 * be allocated and returned.
227 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
229 bool alloc
= false, alive
= false;
232 if (refcount_read(&net
->count
) == 0)
233 return NETNSA_NSID_NOT_ASSIGNED
;
234 spin_lock_bh(&net
->nsid_lock
);
236 * When peer is obtained from RCU lists, we may race with
237 * its cleanup. Check whether it's alive, and this guarantees
238 * we never hash a peer back to net->netns_ids, after it has
239 * just been idr_remove()'d from there in cleanup_net().
241 if (maybe_get_net(peer
))
242 alive
= alloc
= true;
243 id
= __peernet2id_alloc(net
, peer
, &alloc
);
244 spin_unlock_bh(&net
->nsid_lock
);
245 if (alloc
&& id
>= 0)
246 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
251 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
253 /* This function returns, if assigned, the id of a peer netns. */
254 int peernet2id(struct net
*net
, struct net
*peer
)
258 spin_lock_bh(&net
->nsid_lock
);
259 id
= __peernet2id(net
, peer
);
260 spin_unlock_bh(&net
->nsid_lock
);
263 EXPORT_SYMBOL(peernet2id
);
265 /* This function returns true is the peer netns has an id assigned into the
268 bool peernet_has_id(struct net
*net
, struct net
*peer
)
270 return peernet2id(net
, peer
) >= 0;
273 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
281 peer
= idr_find(&net
->netns_ids
, id
);
283 peer
= maybe_get_net(peer
);
290 * setup_net runs the initializers for the network namespace object.
292 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
294 /* Must be called with net_sem held */
295 const struct pernet_operations
*ops
, *saved_ops
;
297 LIST_HEAD(net_exit_list
);
299 refcount_set(&net
->count
, 1);
300 refcount_set(&net
->passive
, 1);
301 net
->dev_base_seq
= 1;
302 net
->user_ns
= user_ns
;
303 idr_init(&net
->netns_ids
);
304 spin_lock_init(&net
->nsid_lock
);
306 list_for_each_entry(ops
, &pernet_list
, list
) {
307 error
= ops_init(ops
, net
);
312 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
318 /* Walk through the list backwards calling the exit functions
319 * for the pernet modules whose init functions did not fail.
321 list_add(&net
->exit_list
, &net_exit_list
);
323 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
324 ops_exit_list(ops
, &net_exit_list
);
327 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
328 ops_free_list(ops
, &net_exit_list
);
334 static int __net_init
net_defaults_init_net(struct net
*net
)
336 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
340 static struct pernet_operations net_defaults_ops
= {
341 .init
= net_defaults_init_net
,
344 static __init
int net_defaults_init(void)
346 if (register_pernet_subsys(&net_defaults_ops
))
347 panic("Cannot initialize net default settings");
352 core_initcall(net_defaults_init
);
355 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
357 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
360 static void dec_net_namespaces(struct ucounts
*ucounts
)
362 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
365 static struct kmem_cache
*net_cachep
;
366 static struct workqueue_struct
*netns_wq
;
368 static struct net
*net_alloc(void)
370 struct net
*net
= NULL
;
371 struct net_generic
*ng
;
373 ng
= net_alloc_generic();
377 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
381 rcu_assign_pointer(net
->gen
, ng
);
390 static void net_free(struct net
*net
)
392 kfree(rcu_access_pointer(net
->gen
));
393 kmem_cache_free(net_cachep
, net
);
396 void net_drop_ns(void *p
)
399 if (ns
&& refcount_dec_and_test(&ns
->passive
))
403 struct net
*copy_net_ns(unsigned long flags
,
404 struct user_namespace
*user_ns
, struct net
*old_net
)
406 struct ucounts
*ucounts
;
410 if (!(flags
& CLONE_NEWNET
))
411 return get_net(old_net
);
413 ucounts
= inc_net_namespaces(user_ns
);
415 return ERR_PTR(-ENOSPC
);
422 refcount_set(&net
->passive
, 1);
423 net
->ucounts
= ucounts
;
424 get_user_ns(user_ns
);
426 rv
= down_read_killable(&net_sem
);
429 rv
= mutex_lock_killable(&net_mutex
);
432 rv
= setup_net(net
, user_ns
);
433 mutex_unlock(&net_mutex
);
438 put_user_ns(user_ns
);
441 dec_net_namespaces(ucounts
);
447 static void unhash_nsid(struct net
*net
, struct net
*last
)
450 /* This function is only called from cleanup_net() work,
451 * and this work is the only process, that may delete
452 * a net from net_namespace_list. So, when the below
453 * is executing, the list may only grow. Thus, we do not
454 * use for_each_net_rcu() or rtnl_lock().
459 spin_lock_bh(&tmp
->nsid_lock
);
460 id
= __peernet2id(tmp
, net
);
462 idr_remove(&tmp
->netns_ids
, id
);
463 spin_unlock_bh(&tmp
->nsid_lock
);
465 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
469 spin_lock_bh(&net
->nsid_lock
);
470 idr_destroy(&net
->netns_ids
);
471 spin_unlock_bh(&net
->nsid_lock
);
474 static DEFINE_SPINLOCK(cleanup_list_lock
);
475 static LIST_HEAD(cleanup_list
); /* Must hold cleanup_list_lock to touch */
477 static void cleanup_net(struct work_struct
*work
)
479 const struct pernet_operations
*ops
;
480 struct net
*net
, *tmp
, *last
;
481 struct list_head net_kill_list
;
482 LIST_HEAD(net_exit_list
);
484 /* Atomically snapshot the list of namespaces to cleanup */
485 spin_lock_irq(&cleanup_list_lock
);
486 list_replace_init(&cleanup_list
, &net_kill_list
);
487 spin_unlock_irq(&cleanup_list_lock
);
490 mutex_lock(&net_mutex
);
492 /* Don't let anyone else find us. */
494 list_for_each_entry(net
, &net_kill_list
, cleanup_list
)
495 list_del_rcu(&net
->list
);
496 /* Cache last net. After we unlock rtnl, no one new net
497 * added to net_namespace_list can assign nsid pointer
498 * to a net from net_kill_list (see peernet2id_alloc()).
499 * So, we skip them in unhash_nsid().
501 * Note, that unhash_nsid() does not delete nsid links
502 * between net_kill_list's nets, as they've already
503 * deleted from net_namespace_list. But, this would be
504 * useless anyway, as netns_ids are destroyed there.
506 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
509 list_for_each_entry(net
, &net_kill_list
, cleanup_list
) {
510 unhash_nsid(net
, last
);
511 list_add_tail(&net
->exit_list
, &net_exit_list
);
515 * Another CPU might be rcu-iterating the list, wait for it.
516 * This needs to be before calling the exit() notifiers, so
517 * the rcu_barrier() below isn't sufficient alone.
521 /* Run all of the network namespace exit methods */
522 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
523 ops_exit_list(ops
, &net_exit_list
);
525 mutex_unlock(&net_mutex
);
527 /* Free the net generic variables */
528 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
529 ops_free_list(ops
, &net_exit_list
);
533 /* Ensure there are no outstanding rcu callbacks using this
538 /* Finally it is safe to free my network namespace structure */
539 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
540 list_del_init(&net
->exit_list
);
541 dec_net_namespaces(net
->ucounts
);
542 put_user_ns(net
->user_ns
);
548 * net_ns_barrier - wait until concurrent net_cleanup_work is done
550 * cleanup_net runs from work queue and will first remove namespaces
551 * from the global list, then run net exit functions.
553 * Call this in module exit path to make sure that all netns
554 * ->exit ops have been invoked before the function is removed.
556 void net_ns_barrier(void)
558 down_write(&net_sem
);
559 mutex_lock(&net_mutex
);
560 mutex_unlock(&net_mutex
);
563 EXPORT_SYMBOL(net_ns_barrier
);
565 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
567 void __put_net(struct net
*net
)
569 /* Cleanup the network namespace in process context */
572 spin_lock_irqsave(&cleanup_list_lock
, flags
);
573 list_add(&net
->cleanup_list
, &cleanup_list
);
574 spin_unlock_irqrestore(&cleanup_list_lock
, flags
);
576 queue_work(netns_wq
, &net_cleanup_work
);
578 EXPORT_SYMBOL_GPL(__put_net
);
580 struct net
*get_net_ns_by_fd(int fd
)
583 struct ns_common
*ns
;
586 file
= proc_ns_fget(fd
);
588 return ERR_CAST(file
);
590 ns
= get_proc_ns(file_inode(file
));
591 if (ns
->ops
== &netns_operations
)
592 net
= get_net(container_of(ns
, struct net
, ns
));
594 net
= ERR_PTR(-EINVAL
);
601 struct net
*get_net_ns_by_fd(int fd
)
603 return ERR_PTR(-EINVAL
);
606 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
608 struct net
*get_net_ns_by_pid(pid_t pid
)
610 struct task_struct
*tsk
;
613 /* Lookup the network namespace */
614 net
= ERR_PTR(-ESRCH
);
616 tsk
= find_task_by_vpid(pid
);
618 struct nsproxy
*nsproxy
;
620 nsproxy
= tsk
->nsproxy
;
622 net
= get_net(nsproxy
->net_ns
);
628 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
630 static __net_init
int net_ns_net_init(struct net
*net
)
633 net
->ns
.ops
= &netns_operations
;
635 return ns_alloc_inum(&net
->ns
);
638 static __net_exit
void net_ns_net_exit(struct net
*net
)
640 ns_free_inum(&net
->ns
);
643 static struct pernet_operations __net_initdata net_ns_ops
= {
644 .init
= net_ns_net_init
,
645 .exit
= net_ns_net_exit
,
648 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
649 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
650 [NETNSA_NSID
] = { .type
= NLA_S32
},
651 [NETNSA_PID
] = { .type
= NLA_U32
},
652 [NETNSA_FD
] = { .type
= NLA_U32
},
655 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
656 struct netlink_ext_ack
*extack
)
658 struct net
*net
= sock_net(skb
->sk
);
659 struct nlattr
*tb
[NETNSA_MAX
+ 1];
664 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
665 rtnl_net_policy
, extack
);
668 if (!tb
[NETNSA_NSID
]) {
669 NL_SET_ERR_MSG(extack
, "nsid is missing");
672 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
674 if (tb
[NETNSA_PID
]) {
675 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
676 nla
= tb
[NETNSA_PID
];
677 } else if (tb
[NETNSA_FD
]) {
678 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
681 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
685 NL_SET_BAD_ATTR(extack
, nla
);
686 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
687 return PTR_ERR(peer
);
690 spin_lock_bh(&net
->nsid_lock
);
691 if (__peernet2id(net
, peer
) >= 0) {
692 spin_unlock_bh(&net
->nsid_lock
);
694 NL_SET_BAD_ATTR(extack
, nla
);
695 NL_SET_ERR_MSG(extack
,
696 "Peer netns already has a nsid assigned");
700 err
= alloc_netid(net
, peer
, nsid
);
701 spin_unlock_bh(&net
->nsid_lock
);
703 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
705 } else if (err
== -ENOSPC
&& nsid
>= 0) {
707 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
708 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
715 static int rtnl_net_get_size(void)
717 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
718 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
722 static int rtnl_net_fill(struct sk_buff
*skb
, u32 portid
, u32 seq
, int flags
,
723 int cmd
, struct net
*net
, int nsid
)
725 struct nlmsghdr
*nlh
;
726 struct rtgenmsg
*rth
;
728 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rth
), flags
);
732 rth
= nlmsg_data(nlh
);
733 rth
->rtgen_family
= AF_UNSPEC
;
735 if (nla_put_s32(skb
, NETNSA_NSID
, nsid
))
736 goto nla_put_failure
;
742 nlmsg_cancel(skb
, nlh
);
746 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
747 struct netlink_ext_ack
*extack
)
749 struct net
*net
= sock_net(skb
->sk
);
750 struct nlattr
*tb
[NETNSA_MAX
+ 1];
756 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
757 rtnl_net_policy
, extack
);
760 if (tb
[NETNSA_PID
]) {
761 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
762 nla
= tb
[NETNSA_PID
];
763 } else if (tb
[NETNSA_FD
]) {
764 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
767 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
772 NL_SET_BAD_ATTR(extack
, nla
);
773 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
774 return PTR_ERR(peer
);
777 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
783 id
= peernet2id(net
, peer
);
784 err
= rtnl_net_fill(msg
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
785 RTM_NEWNSID
, net
, id
);
789 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
799 struct rtnl_net_dump_cb
{
802 struct netlink_callback
*cb
;
807 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
809 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
812 if (net_cb
->idx
< net_cb
->s_idx
)
815 ret
= rtnl_net_fill(net_cb
->skb
, NETLINK_CB(net_cb
->cb
->skb
).portid
,
816 net_cb
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
817 RTM_NEWNSID
, net_cb
->net
, id
);
826 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
828 struct net
*net
= sock_net(skb
->sk
);
829 struct rtnl_net_dump_cb net_cb
= {
834 .s_idx
= cb
->args
[0],
837 spin_lock_bh(&net
->nsid_lock
);
838 idr_for_each(&net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
839 spin_unlock_bh(&net
->nsid_lock
);
841 cb
->args
[0] = net_cb
.idx
;
845 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
850 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
854 err
= rtnl_net_fill(msg
, 0, 0, 0, cmd
, net
, id
);
858 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
864 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
867 static int __init
net_ns_init(void)
869 struct net_generic
*ng
;
872 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
876 /* Create workqueue for cleanup */
877 netns_wq
= create_singlethread_workqueue("netns");
879 panic("Could not create netns workq");
882 ng
= net_alloc_generic();
884 panic("Could not allocate generic netns");
886 rcu_assign_pointer(init_net
.gen
, ng
);
888 down_write(&net_sem
);
889 if (setup_net(&init_net
, &init_user_ns
))
890 panic("Could not setup the initial network namespace");
892 init_net_initialized
= true;
895 register_pernet_subsys(&net_ns_ops
);
897 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
898 RTNL_FLAG_DOIT_UNLOCKED
);
899 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
900 RTNL_FLAG_DOIT_UNLOCKED
);
905 pure_initcall(net_ns_init
);
908 static int __register_pernet_operations(struct list_head
*list
,
909 struct pernet_operations
*ops
)
913 LIST_HEAD(net_exit_list
);
915 list_add_tail(&ops
->list
, list
);
916 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
918 error
= ops_init(ops
, net
);
921 list_add_tail(&net
->exit_list
, &net_exit_list
);
927 /* If I have an error cleanup all namespaces I initialized */
928 list_del(&ops
->list
);
929 ops_exit_list(ops
, &net_exit_list
);
930 ops_free_list(ops
, &net_exit_list
);
934 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
937 LIST_HEAD(net_exit_list
);
939 list_del(&ops
->list
);
941 list_add_tail(&net
->exit_list
, &net_exit_list
);
942 ops_exit_list(ops
, &net_exit_list
);
943 ops_free_list(ops
, &net_exit_list
);
948 static int __register_pernet_operations(struct list_head
*list
,
949 struct pernet_operations
*ops
)
951 if (!init_net_initialized
) {
952 list_add_tail(&ops
->list
, list
);
956 return ops_init(ops
, &init_net
);
959 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
961 if (!init_net_initialized
) {
962 list_del(&ops
->list
);
964 LIST_HEAD(net_exit_list
);
965 list_add(&init_net
.exit_list
, &net_exit_list
);
966 ops_exit_list(ops
, &net_exit_list
);
967 ops_free_list(ops
, &net_exit_list
);
971 #endif /* CONFIG_NET_NS */
973 static DEFINE_IDA(net_generic_ids
);
975 static int register_pernet_operations(struct list_head
*list
,
976 struct pernet_operations
*ops
)
982 error
= ida_get_new_above(&net_generic_ids
, MIN_PERNET_OPS_ID
, ops
->id
);
984 if (error
== -EAGAIN
) {
985 ida_pre_get(&net_generic_ids
, GFP_KERNEL
);
990 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
992 error
= __register_pernet_operations(list
, ops
);
996 ida_remove(&net_generic_ids
, *ops
->id
);
1002 static void unregister_pernet_operations(struct pernet_operations
*ops
)
1005 __unregister_pernet_operations(ops
);
1008 ida_remove(&net_generic_ids
, *ops
->id
);
1012 * register_pernet_subsys - register a network namespace subsystem
1013 * @ops: pernet operations structure for the subsystem
1015 * Register a subsystem which has init and exit functions
1016 * that are called when network namespaces are created and
1017 * destroyed respectively.
1019 * When registered all network namespace init functions are
1020 * called for every existing network namespace. Allowing kernel
1021 * modules to have a race free view of the set of network namespaces.
1023 * When a new network namespace is created all of the init
1024 * methods are called in the order in which they were registered.
1026 * When a network namespace is destroyed all of the exit methods
1027 * are called in the reverse of the order with which they were
1030 int register_pernet_subsys(struct pernet_operations
*ops
)
1033 down_write(&net_sem
);
1034 error
= register_pernet_operations(first_device
, ops
);
1038 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1041 * unregister_pernet_subsys - unregister a network namespace subsystem
1042 * @ops: pernet operations structure to manipulate
1044 * Remove the pernet operations structure from the list to be
1045 * used when network namespaces are created or destroyed. In
1046 * addition run the exit method for all existing network
1049 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1051 down_write(&net_sem
);
1052 unregister_pernet_operations(ops
);
1055 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1058 * register_pernet_device - register a network namespace device
1059 * @ops: pernet operations structure for the subsystem
1061 * Register a device which has init and exit functions
1062 * that are called when network namespaces are created and
1063 * destroyed respectively.
1065 * When registered all network namespace init functions are
1066 * called for every existing network namespace. Allowing kernel
1067 * modules to have a race free view of the set of network namespaces.
1069 * When a new network namespace is created all of the init
1070 * methods are called in the order in which they were registered.
1072 * When a network namespace is destroyed all of the exit methods
1073 * are called in the reverse of the order with which they were
1076 int register_pernet_device(struct pernet_operations
*ops
)
1079 down_write(&net_sem
);
1080 error
= register_pernet_operations(&pernet_list
, ops
);
1081 if (!error
&& (first_device
== &pernet_list
))
1082 first_device
= &ops
->list
;
1086 EXPORT_SYMBOL_GPL(register_pernet_device
);
1089 * unregister_pernet_device - unregister a network namespace netdevice
1090 * @ops: pernet operations structure to manipulate
1092 * Remove the pernet operations structure from the list to be
1093 * used when network namespaces are created or destroyed. In
1094 * addition run the exit method for all existing network
1097 void unregister_pernet_device(struct pernet_operations
*ops
)
1099 down_write(&net_sem
);
1100 if (&ops
->list
== first_device
)
1101 first_device
= first_device
->next
;
1102 unregister_pernet_operations(ops
);
1105 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1107 #ifdef CONFIG_NET_NS
1108 static struct ns_common
*netns_get(struct task_struct
*task
)
1110 struct net
*net
= NULL
;
1111 struct nsproxy
*nsproxy
;
1114 nsproxy
= task
->nsproxy
;
1116 net
= get_net(nsproxy
->net_ns
);
1119 return net
? &net
->ns
: NULL
;
1122 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1124 return container_of(ns
, struct net
, ns
);
1127 static void netns_put(struct ns_common
*ns
)
1129 put_net(to_net_ns(ns
));
1132 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1134 struct net
*net
= to_net_ns(ns
);
1136 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1137 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1140 put_net(nsproxy
->net_ns
);
1141 nsproxy
->net_ns
= get_net(net
);
1145 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1147 return to_net_ns(ns
)->user_ns
;
1150 const struct proc_ns_operations netns_operations
= {
1152 .type
= CLONE_NEWNET
,
1155 .install
= netns_install
,
1156 .owner
= netns_owner
,