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 /* Used only if there are !async pernet_operations registered */
33 DEFINE_MUTEX(net_mutex
);
35 LIST_HEAD(net_namespace_list
);
36 EXPORT_SYMBOL_GPL(net_namespace_list
);
38 struct net init_net
= {
39 .count
= REFCOUNT_INIT(1),
40 .dev_base_head
= LIST_HEAD_INIT(init_net
.dev_base_head
),
42 EXPORT_SYMBOL(init_net
);
44 static bool init_net_initialized
;
45 static unsigned nr_sync_pernet_ops
;
47 * net_sem: protects: pernet_list, net_generic_ids, nr_sync_pernet_ops,
48 * init_net_initialized and first_device pointer.
50 DECLARE_RWSEM(net_sem
);
52 #define MIN_PERNET_OPS_ID \
53 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
55 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
57 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
59 static struct net_generic
*net_alloc_generic(void)
61 struct net_generic
*ng
;
62 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
64 ng
= kzalloc(generic_size
, GFP_KERNEL
);
66 ng
->s
.len
= max_gen_ptrs
;
71 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
73 struct net_generic
*ng
, *old_ng
;
75 BUG_ON(id
< MIN_PERNET_OPS_ID
);
77 old_ng
= rcu_dereference_protected(net
->gen
,
78 lockdep_is_held(&net_sem
));
79 if (old_ng
->s
.len
> id
) {
80 old_ng
->ptr
[id
] = data
;
84 ng
= net_alloc_generic();
89 * Some synchronisation notes:
91 * The net_generic explores the net->gen array inside rcu
92 * read section. Besides once set the net->gen->ptr[x]
93 * pointer never changes (see rules in netns/generic.h).
95 * That said, we simply duplicate this array and schedule
96 * the old copy for kfree after a grace period.
99 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
100 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
103 rcu_assign_pointer(net
->gen
, ng
);
104 kfree_rcu(old_ng
, s
.rcu
);
108 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
113 if (ops
->id
&& ops
->size
) {
114 data
= kzalloc(ops
->size
, GFP_KERNEL
);
118 err
= net_assign_generic(net
, *ops
->id
, data
);
124 err
= ops
->init(net
);
135 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
137 if (ops
->id
&& ops
->size
) {
138 kfree(net_generic(net
, *ops
->id
));
142 static void ops_exit_list(const struct pernet_operations
*ops
,
143 struct list_head
*net_exit_list
)
147 list_for_each_entry(net
, net_exit_list
, exit_list
)
151 ops
->exit_batch(net_exit_list
);
154 static void ops_free_list(const struct pernet_operations
*ops
,
155 struct list_head
*net_exit_list
)
158 if (ops
->size
&& ops
->id
) {
159 list_for_each_entry(net
, net_exit_list
, exit_list
)
164 /* should be called with nsid_lock held */
165 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
167 int min
= 0, max
= 0;
174 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
177 /* This function is used by idr_for_each(). If net is equal to peer, the
178 * function returns the id so that idr_for_each() stops. Because we cannot
179 * returns the id 0 (idr_for_each() will not stop), we return the magic value
180 * NET_ID_ZERO (-1) for it.
182 #define NET_ID_ZERO -1
183 static int net_eq_idr(int id
, void *net
, void *peer
)
185 if (net_eq(net
, peer
))
186 return id
? : NET_ID_ZERO
;
190 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
191 * is set to true, thus the caller knows that the new id must be notified via
194 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
196 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
197 bool alloc_it
= *alloc
;
201 /* Magic value for id 0. */
202 if (id
== NET_ID_ZERO
)
208 id
= alloc_netid(net
, peer
, -1);
210 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
213 return NETNSA_NSID_NOT_ASSIGNED
;
216 /* should be called with nsid_lock held */
217 static int __peernet2id(struct net
*net
, struct net
*peer
)
221 return __peernet2id_alloc(net
, peer
, &no
);
224 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
225 /* This function returns the id of a peer netns. If no id is assigned, one will
226 * be allocated and returned.
228 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
230 bool alloc
= false, alive
= false;
233 if (refcount_read(&net
->count
) == 0)
234 return NETNSA_NSID_NOT_ASSIGNED
;
235 spin_lock_bh(&net
->nsid_lock
);
237 * When peer is obtained from RCU lists, we may race with
238 * its cleanup. Check whether it's alive, and this guarantees
239 * we never hash a peer back to net->netns_ids, after it has
240 * just been idr_remove()'d from there in cleanup_net().
242 if (maybe_get_net(peer
))
243 alive
= alloc
= true;
244 id
= __peernet2id_alloc(net
, peer
, &alloc
);
245 spin_unlock_bh(&net
->nsid_lock
);
246 if (alloc
&& id
>= 0)
247 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
252 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
254 /* This function returns, if assigned, the id of a peer netns. */
255 int peernet2id(struct net
*net
, struct net
*peer
)
259 spin_lock_bh(&net
->nsid_lock
);
260 id
= __peernet2id(net
, peer
);
261 spin_unlock_bh(&net
->nsid_lock
);
264 EXPORT_SYMBOL(peernet2id
);
266 /* This function returns true is the peer netns has an id assigned into the
269 bool peernet_has_id(struct net
*net
, struct net
*peer
)
271 return peernet2id(net
, peer
) >= 0;
274 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
282 peer
= idr_find(&net
->netns_ids
, id
);
284 peer
= maybe_get_net(peer
);
291 * setup_net runs the initializers for the network namespace object.
293 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
295 /* Must be called with net_sem held */
296 const struct pernet_operations
*ops
, *saved_ops
;
298 LIST_HEAD(net_exit_list
);
300 refcount_set(&net
->count
, 1);
301 refcount_set(&net
->passive
, 1);
302 net
->dev_base_seq
= 1;
303 net
->user_ns
= user_ns
;
304 idr_init(&net
->netns_ids
);
305 spin_lock_init(&net
->nsid_lock
);
307 list_for_each_entry(ops
, &pernet_list
, list
) {
308 error
= ops_init(ops
, net
);
313 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
319 /* Walk through the list backwards calling the exit functions
320 * for the pernet modules whose init functions did not fail.
322 list_add(&net
->exit_list
, &net_exit_list
);
324 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
325 ops_exit_list(ops
, &net_exit_list
);
328 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
329 ops_free_list(ops
, &net_exit_list
);
335 static int __net_init
net_defaults_init_net(struct net
*net
)
337 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
341 static struct pernet_operations net_defaults_ops
= {
342 .init
= net_defaults_init_net
,
345 static __init
int net_defaults_init(void)
347 if (register_pernet_subsys(&net_defaults_ops
))
348 panic("Cannot initialize net default settings");
353 core_initcall(net_defaults_init
);
356 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
358 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
361 static void dec_net_namespaces(struct ucounts
*ucounts
)
363 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
366 static struct kmem_cache
*net_cachep
;
367 static struct workqueue_struct
*netns_wq
;
369 static struct net
*net_alloc(void)
371 struct net
*net
= NULL
;
372 struct net_generic
*ng
;
374 ng
= net_alloc_generic();
378 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
382 rcu_assign_pointer(net
->gen
, ng
);
391 static void net_free(struct net
*net
)
393 kfree(rcu_access_pointer(net
->gen
));
394 kmem_cache_free(net_cachep
, net
);
397 void net_drop_ns(void *p
)
400 if (ns
&& refcount_dec_and_test(&ns
->passive
))
404 struct net
*copy_net_ns(unsigned long flags
,
405 struct user_namespace
*user_ns
, struct net
*old_net
)
407 struct ucounts
*ucounts
;
411 if (!(flags
& CLONE_NEWNET
))
412 return get_net(old_net
);
414 ucounts
= inc_net_namespaces(user_ns
);
416 return ERR_PTR(-ENOSPC
);
423 refcount_set(&net
->passive
, 1);
424 net
->ucounts
= ucounts
;
425 get_user_ns(user_ns
);
427 rv
= down_read_killable(&net_sem
);
430 if (nr_sync_pernet_ops
) {
431 rv
= mutex_lock_killable(&net_mutex
);
435 rv
= setup_net(net
, user_ns
);
436 if (nr_sync_pernet_ops
)
437 mutex_unlock(&net_mutex
);
442 put_user_ns(user_ns
);
445 dec_net_namespaces(ucounts
);
451 static void unhash_nsid(struct net
*net
, struct net
*last
)
454 /* This function is only called from cleanup_net() work,
455 * and this work is the only process, that may delete
456 * a net from net_namespace_list. So, when the below
457 * is executing, the list may only grow. Thus, we do not
458 * use for_each_net_rcu() or rtnl_lock().
463 spin_lock_bh(&tmp
->nsid_lock
);
464 id
= __peernet2id(tmp
, net
);
466 idr_remove(&tmp
->netns_ids
, id
);
467 spin_unlock_bh(&tmp
->nsid_lock
);
469 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
473 spin_lock_bh(&net
->nsid_lock
);
474 idr_destroy(&net
->netns_ids
);
475 spin_unlock_bh(&net
->nsid_lock
);
478 static DEFINE_SPINLOCK(cleanup_list_lock
);
479 static LIST_HEAD(cleanup_list
); /* Must hold cleanup_list_lock to touch */
481 static void cleanup_net(struct work_struct
*work
)
483 const struct pernet_operations
*ops
;
484 struct net
*net
, *tmp
, *last
;
485 struct list_head net_kill_list
;
486 LIST_HEAD(net_exit_list
);
488 /* Atomically snapshot the list of namespaces to cleanup */
489 spin_lock_irq(&cleanup_list_lock
);
490 list_replace_init(&cleanup_list
, &net_kill_list
);
491 spin_unlock_irq(&cleanup_list_lock
);
494 if (nr_sync_pernet_ops
)
495 mutex_lock(&net_mutex
);
497 /* Don't let anyone else find us. */
499 list_for_each_entry(net
, &net_kill_list
, cleanup_list
)
500 list_del_rcu(&net
->list
);
501 /* Cache last net. After we unlock rtnl, no one new net
502 * added to net_namespace_list can assign nsid pointer
503 * to a net from net_kill_list (see peernet2id_alloc()).
504 * So, we skip them in unhash_nsid().
506 * Note, that unhash_nsid() does not delete nsid links
507 * between net_kill_list's nets, as they've already
508 * deleted from net_namespace_list. But, this would be
509 * useless anyway, as netns_ids are destroyed there.
511 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
514 list_for_each_entry(net
, &net_kill_list
, cleanup_list
) {
515 unhash_nsid(net
, last
);
516 list_add_tail(&net
->exit_list
, &net_exit_list
);
520 * Another CPU might be rcu-iterating the list, wait for it.
521 * This needs to be before calling the exit() notifiers, so
522 * the rcu_barrier() below isn't sufficient alone.
526 /* Run all of the network namespace exit methods */
527 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
528 ops_exit_list(ops
, &net_exit_list
);
530 if (nr_sync_pernet_ops
)
531 mutex_unlock(&net_mutex
);
533 /* Free the net generic variables */
534 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
535 ops_free_list(ops
, &net_exit_list
);
539 /* Ensure there are no outstanding rcu callbacks using this
544 /* Finally it is safe to free my network namespace structure */
545 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
546 list_del_init(&net
->exit_list
);
547 dec_net_namespaces(net
->ucounts
);
548 put_user_ns(net
->user_ns
);
554 * net_ns_barrier - wait until concurrent net_cleanup_work is done
556 * cleanup_net runs from work queue and will first remove namespaces
557 * from the global list, then run net exit functions.
559 * Call this in module exit path to make sure that all netns
560 * ->exit ops have been invoked before the function is removed.
562 void net_ns_barrier(void)
564 down_write(&net_sem
);
565 mutex_lock(&net_mutex
);
566 mutex_unlock(&net_mutex
);
569 EXPORT_SYMBOL(net_ns_barrier
);
571 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
573 void __put_net(struct net
*net
)
575 /* Cleanup the network namespace in process context */
578 spin_lock_irqsave(&cleanup_list_lock
, flags
);
579 list_add(&net
->cleanup_list
, &cleanup_list
);
580 spin_unlock_irqrestore(&cleanup_list_lock
, flags
);
582 queue_work(netns_wq
, &net_cleanup_work
);
584 EXPORT_SYMBOL_GPL(__put_net
);
586 struct net
*get_net_ns_by_fd(int fd
)
589 struct ns_common
*ns
;
592 file
= proc_ns_fget(fd
);
594 return ERR_CAST(file
);
596 ns
= get_proc_ns(file_inode(file
));
597 if (ns
->ops
== &netns_operations
)
598 net
= get_net(container_of(ns
, struct net
, ns
));
600 net
= ERR_PTR(-EINVAL
);
607 struct net
*get_net_ns_by_fd(int fd
)
609 return ERR_PTR(-EINVAL
);
612 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
614 struct net
*get_net_ns_by_pid(pid_t pid
)
616 struct task_struct
*tsk
;
619 /* Lookup the network namespace */
620 net
= ERR_PTR(-ESRCH
);
622 tsk
= find_task_by_vpid(pid
);
624 struct nsproxy
*nsproxy
;
626 nsproxy
= tsk
->nsproxy
;
628 net
= get_net(nsproxy
->net_ns
);
634 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
636 static __net_init
int net_ns_net_init(struct net
*net
)
639 net
->ns
.ops
= &netns_operations
;
641 return ns_alloc_inum(&net
->ns
);
644 static __net_exit
void net_ns_net_exit(struct net
*net
)
646 ns_free_inum(&net
->ns
);
649 static struct pernet_operations __net_initdata net_ns_ops
= {
650 .init
= net_ns_net_init
,
651 .exit
= net_ns_net_exit
,
654 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
655 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
656 [NETNSA_NSID
] = { .type
= NLA_S32
},
657 [NETNSA_PID
] = { .type
= NLA_U32
},
658 [NETNSA_FD
] = { .type
= NLA_U32
},
661 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
662 struct netlink_ext_ack
*extack
)
664 struct net
*net
= sock_net(skb
->sk
);
665 struct nlattr
*tb
[NETNSA_MAX
+ 1];
670 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
671 rtnl_net_policy
, extack
);
674 if (!tb
[NETNSA_NSID
]) {
675 NL_SET_ERR_MSG(extack
, "nsid is missing");
678 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
680 if (tb
[NETNSA_PID
]) {
681 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
682 nla
= tb
[NETNSA_PID
];
683 } else if (tb
[NETNSA_FD
]) {
684 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
687 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
691 NL_SET_BAD_ATTR(extack
, nla
);
692 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
693 return PTR_ERR(peer
);
696 spin_lock_bh(&net
->nsid_lock
);
697 if (__peernet2id(net
, peer
) >= 0) {
698 spin_unlock_bh(&net
->nsid_lock
);
700 NL_SET_BAD_ATTR(extack
, nla
);
701 NL_SET_ERR_MSG(extack
,
702 "Peer netns already has a nsid assigned");
706 err
= alloc_netid(net
, peer
, nsid
);
707 spin_unlock_bh(&net
->nsid_lock
);
709 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
711 } else if (err
== -ENOSPC
&& nsid
>= 0) {
713 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
714 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
721 static int rtnl_net_get_size(void)
723 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
724 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
728 static int rtnl_net_fill(struct sk_buff
*skb
, u32 portid
, u32 seq
, int flags
,
729 int cmd
, struct net
*net
, int nsid
)
731 struct nlmsghdr
*nlh
;
732 struct rtgenmsg
*rth
;
734 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rth
), flags
);
738 rth
= nlmsg_data(nlh
);
739 rth
->rtgen_family
= AF_UNSPEC
;
741 if (nla_put_s32(skb
, NETNSA_NSID
, nsid
))
742 goto nla_put_failure
;
748 nlmsg_cancel(skb
, nlh
);
752 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
753 struct netlink_ext_ack
*extack
)
755 struct net
*net
= sock_net(skb
->sk
);
756 struct nlattr
*tb
[NETNSA_MAX
+ 1];
762 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
763 rtnl_net_policy
, extack
);
766 if (tb
[NETNSA_PID
]) {
767 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
768 nla
= tb
[NETNSA_PID
];
769 } else if (tb
[NETNSA_FD
]) {
770 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
773 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
778 NL_SET_BAD_ATTR(extack
, nla
);
779 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
780 return PTR_ERR(peer
);
783 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
789 id
= peernet2id(net
, peer
);
790 err
= rtnl_net_fill(msg
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
791 RTM_NEWNSID
, net
, id
);
795 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
805 struct rtnl_net_dump_cb
{
808 struct netlink_callback
*cb
;
813 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
815 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
818 if (net_cb
->idx
< net_cb
->s_idx
)
821 ret
= rtnl_net_fill(net_cb
->skb
, NETLINK_CB(net_cb
->cb
->skb
).portid
,
822 net_cb
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
823 RTM_NEWNSID
, net_cb
->net
, id
);
832 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
834 struct net
*net
= sock_net(skb
->sk
);
835 struct rtnl_net_dump_cb net_cb
= {
840 .s_idx
= cb
->args
[0],
843 spin_lock_bh(&net
->nsid_lock
);
844 idr_for_each(&net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
845 spin_unlock_bh(&net
->nsid_lock
);
847 cb
->args
[0] = net_cb
.idx
;
851 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
856 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
860 err
= rtnl_net_fill(msg
, 0, 0, 0, cmd
, net
, id
);
864 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
870 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
873 static int __init
net_ns_init(void)
875 struct net_generic
*ng
;
878 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
882 /* Create workqueue for cleanup */
883 netns_wq
= create_singlethread_workqueue("netns");
885 panic("Could not create netns workq");
888 ng
= net_alloc_generic();
890 panic("Could not allocate generic netns");
892 rcu_assign_pointer(init_net
.gen
, ng
);
894 down_write(&net_sem
);
895 if (setup_net(&init_net
, &init_user_ns
))
896 panic("Could not setup the initial network namespace");
898 init_net_initialized
= true;
901 register_pernet_subsys(&net_ns_ops
);
903 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
904 RTNL_FLAG_DOIT_UNLOCKED
);
905 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
906 RTNL_FLAG_DOIT_UNLOCKED
);
911 pure_initcall(net_ns_init
);
914 static int __register_pernet_operations(struct list_head
*list
,
915 struct pernet_operations
*ops
)
919 LIST_HEAD(net_exit_list
);
921 list_add_tail(&ops
->list
, list
);
922 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
924 error
= ops_init(ops
, net
);
927 list_add_tail(&net
->exit_list
, &net_exit_list
);
933 /* If I have an error cleanup all namespaces I initialized */
934 list_del(&ops
->list
);
935 ops_exit_list(ops
, &net_exit_list
);
936 ops_free_list(ops
, &net_exit_list
);
940 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
943 LIST_HEAD(net_exit_list
);
945 list_del(&ops
->list
);
947 list_add_tail(&net
->exit_list
, &net_exit_list
);
948 ops_exit_list(ops
, &net_exit_list
);
949 ops_free_list(ops
, &net_exit_list
);
954 static int __register_pernet_operations(struct list_head
*list
,
955 struct pernet_operations
*ops
)
957 if (!init_net_initialized
) {
958 list_add_tail(&ops
->list
, list
);
962 return ops_init(ops
, &init_net
);
965 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
967 if (!init_net_initialized
) {
968 list_del(&ops
->list
);
970 LIST_HEAD(net_exit_list
);
971 list_add(&init_net
.exit_list
, &net_exit_list
);
972 ops_exit_list(ops
, &net_exit_list
);
973 ops_free_list(ops
, &net_exit_list
);
977 #endif /* CONFIG_NET_NS */
979 static DEFINE_IDA(net_generic_ids
);
981 static int register_pernet_operations(struct list_head
*list
,
982 struct pernet_operations
*ops
)
988 error
= ida_get_new_above(&net_generic_ids
, MIN_PERNET_OPS_ID
, ops
->id
);
990 if (error
== -EAGAIN
) {
991 ida_pre_get(&net_generic_ids
, GFP_KERNEL
);
996 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
998 error
= __register_pernet_operations(list
, ops
);
1002 ida_remove(&net_generic_ids
, *ops
->id
);
1003 } else if (!ops
->async
) {
1004 pr_info_once("Pernet operations %ps are sync.\n", ops
);
1005 nr_sync_pernet_ops
++;
1011 static void unregister_pernet_operations(struct pernet_operations
*ops
)
1014 BUG_ON(nr_sync_pernet_ops
-- == 0);
1015 __unregister_pernet_operations(ops
);
1018 ida_remove(&net_generic_ids
, *ops
->id
);
1022 * register_pernet_subsys - register a network namespace subsystem
1023 * @ops: pernet operations structure for the subsystem
1025 * Register a subsystem which has init and exit functions
1026 * that are called when network namespaces are created and
1027 * destroyed respectively.
1029 * When registered all network namespace init functions are
1030 * called for every existing network namespace. Allowing kernel
1031 * modules to have a race free view of the set of network namespaces.
1033 * When a new network namespace is created all of the init
1034 * methods are called in the order in which they were registered.
1036 * When a network namespace is destroyed all of the exit methods
1037 * are called in the reverse of the order with which they were
1040 int register_pernet_subsys(struct pernet_operations
*ops
)
1043 down_write(&net_sem
);
1044 error
= register_pernet_operations(first_device
, ops
);
1048 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1051 * unregister_pernet_subsys - unregister a network namespace subsystem
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_subsys(struct pernet_operations
*ops
)
1061 down_write(&net_sem
);
1062 unregister_pernet_operations(ops
);
1065 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1068 * register_pernet_device - register a network namespace device
1069 * @ops: pernet operations structure for the subsystem
1071 * Register a device which has init and exit functions
1072 * that are called when network namespaces are created and
1073 * destroyed respectively.
1075 * When registered all network namespace init functions are
1076 * called for every existing network namespace. Allowing kernel
1077 * modules to have a race free view of the set of network namespaces.
1079 * When a new network namespace is created all of the init
1080 * methods are called in the order in which they were registered.
1082 * When a network namespace is destroyed all of the exit methods
1083 * are called in the reverse of the order with which they were
1086 int register_pernet_device(struct pernet_operations
*ops
)
1089 down_write(&net_sem
);
1090 error
= register_pernet_operations(&pernet_list
, ops
);
1091 if (!error
&& (first_device
== &pernet_list
))
1092 first_device
= &ops
->list
;
1096 EXPORT_SYMBOL_GPL(register_pernet_device
);
1099 * unregister_pernet_device - unregister a network namespace netdevice
1100 * @ops: pernet operations structure to manipulate
1102 * Remove the pernet operations structure from the list to be
1103 * used when network namespaces are created or destroyed. In
1104 * addition run the exit method for all existing network
1107 void unregister_pernet_device(struct pernet_operations
*ops
)
1109 down_write(&net_sem
);
1110 if (&ops
->list
== first_device
)
1111 first_device
= first_device
->next
;
1112 unregister_pernet_operations(ops
);
1115 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1117 #ifdef CONFIG_NET_NS
1118 static struct ns_common
*netns_get(struct task_struct
*task
)
1120 struct net
*net
= NULL
;
1121 struct nsproxy
*nsproxy
;
1124 nsproxy
= task
->nsproxy
;
1126 net
= get_net(nsproxy
->net_ns
);
1129 return net
? &net
->ns
: NULL
;
1132 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1134 return container_of(ns
, struct net
, ns
);
1137 static void netns_put(struct ns_common
*ns
)
1139 put_net(to_net_ns(ns
));
1142 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1144 struct net
*net
= to_net_ns(ns
);
1146 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1147 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1150 put_net(nsproxy
->net_ns
);
1151 nsproxy
->net_ns
= get_net(net
);
1155 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1157 return to_net_ns(ns
)->user_ns
;
1160 const struct proc_ns_operations netns_operations
= {
1162 .type
= CLONE_NEWNET
,
1165 .install
= netns_install
,
1166 .owner
= netns_owner
,