1 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 #include <linux/workqueue.h>
4 #include <linux/rtnetlink.h>
5 #include <linux/cache.h>
6 #include <linux/slab.h>
7 #include <linux/list.h>
8 #include <linux/delay.h>
9 #include <linux/sched.h>
10 #include <linux/idr.h>
11 #include <linux/rculist.h>
12 #include <linux/nsproxy.h>
14 #include <linux/proc_ns.h>
15 #include <linux/file.h>
16 #include <linux/export.h>
17 #include <linux/user_namespace.h>
18 #include <linux/net_namespace.h>
19 #include <linux/sched/task.h>
20 #include <linux/uidgid.h>
23 #include <net/netlink.h>
24 #include <net/net_namespace.h>
25 #include <net/netns/generic.h>
28 * Our network namespace constructor/destructor lists
31 static LIST_HEAD(pernet_list
);
32 static struct list_head
*first_device
= &pernet_list
;
34 LIST_HEAD(net_namespace_list
);
35 EXPORT_SYMBOL_GPL(net_namespace_list
);
37 /* Protects net_namespace_list. Nests iside rtnl_lock() */
38 DECLARE_RWSEM(net_rwsem
);
39 EXPORT_SYMBOL_GPL(net_rwsem
);
41 struct net init_net
= {
42 .count
= REFCOUNT_INIT(1),
43 .dev_base_head
= LIST_HEAD_INIT(init_net
.dev_base_head
),
45 EXPORT_SYMBOL(init_net
);
47 static bool init_net_initialized
;
49 * pernet_ops_rwsem: protects: pernet_list, net_generic_ids,
50 * init_net_initialized and first_device pointer.
51 * This is internal net namespace object. Please, don't use it
54 DECLARE_RWSEM(pernet_ops_rwsem
);
55 EXPORT_SYMBOL_GPL(pernet_ops_rwsem
);
57 #define MIN_PERNET_OPS_ID \
58 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
60 #define INITIAL_NET_GEN_PTRS 13 /* +1 for len +2 for rcu_head */
62 static unsigned int max_gen_ptrs
= INITIAL_NET_GEN_PTRS
;
64 static struct net_generic
*net_alloc_generic(void)
66 struct net_generic
*ng
;
67 unsigned int generic_size
= offsetof(struct net_generic
, ptr
[max_gen_ptrs
]);
69 ng
= kzalloc(generic_size
, GFP_KERNEL
);
71 ng
->s
.len
= max_gen_ptrs
;
76 static int net_assign_generic(struct net
*net
, unsigned int id
, void *data
)
78 struct net_generic
*ng
, *old_ng
;
80 BUG_ON(id
< MIN_PERNET_OPS_ID
);
82 old_ng
= rcu_dereference_protected(net
->gen
,
83 lockdep_is_held(&pernet_ops_rwsem
));
84 if (old_ng
->s
.len
> id
) {
85 old_ng
->ptr
[id
] = data
;
89 ng
= net_alloc_generic();
94 * Some synchronisation notes:
96 * The net_generic explores the net->gen array inside rcu
97 * read section. Besides once set the net->gen->ptr[x]
98 * pointer never changes (see rules in netns/generic.h).
100 * That said, we simply duplicate this array and schedule
101 * the old copy for kfree after a grace period.
104 memcpy(&ng
->ptr
[MIN_PERNET_OPS_ID
], &old_ng
->ptr
[MIN_PERNET_OPS_ID
],
105 (old_ng
->s
.len
- MIN_PERNET_OPS_ID
) * sizeof(void *));
108 rcu_assign_pointer(net
->gen
, ng
);
109 kfree_rcu(old_ng
, s
.rcu
);
113 static int ops_init(const struct pernet_operations
*ops
, struct net
*net
)
118 if (ops
->id
&& ops
->size
) {
119 data
= kzalloc(ops
->size
, GFP_KERNEL
);
123 err
= net_assign_generic(net
, *ops
->id
, data
);
129 err
= ops
->init(net
);
140 static void ops_free(const struct pernet_operations
*ops
, struct net
*net
)
142 if (ops
->id
&& ops
->size
) {
143 kfree(net_generic(net
, *ops
->id
));
147 static void ops_exit_list(const struct pernet_operations
*ops
,
148 struct list_head
*net_exit_list
)
152 list_for_each_entry(net
, net_exit_list
, exit_list
)
156 ops
->exit_batch(net_exit_list
);
159 static void ops_free_list(const struct pernet_operations
*ops
,
160 struct list_head
*net_exit_list
)
163 if (ops
->size
&& ops
->id
) {
164 list_for_each_entry(net
, net_exit_list
, exit_list
)
169 /* should be called with nsid_lock held */
170 static int alloc_netid(struct net
*net
, struct net
*peer
, int reqid
)
172 int min
= 0, max
= 0;
179 return idr_alloc(&net
->netns_ids
, peer
, min
, max
, GFP_ATOMIC
);
182 /* This function is used by idr_for_each(). If net is equal to peer, the
183 * function returns the id so that idr_for_each() stops. Because we cannot
184 * returns the id 0 (idr_for_each() will not stop), we return the magic value
185 * NET_ID_ZERO (-1) for it.
187 #define NET_ID_ZERO -1
188 static int net_eq_idr(int id
, void *net
, void *peer
)
190 if (net_eq(net
, peer
))
191 return id
? : NET_ID_ZERO
;
195 /* Should be called with nsid_lock held. If a new id is assigned, the bool alloc
196 * is set to true, thus the caller knows that the new id must be notified via
199 static int __peernet2id_alloc(struct net
*net
, struct net
*peer
, bool *alloc
)
201 int id
= idr_for_each(&net
->netns_ids
, net_eq_idr
, peer
);
202 bool alloc_it
= *alloc
;
206 /* Magic value for id 0. */
207 if (id
== NET_ID_ZERO
)
213 id
= alloc_netid(net
, peer
, -1);
215 return id
>= 0 ? id
: NETNSA_NSID_NOT_ASSIGNED
;
218 return NETNSA_NSID_NOT_ASSIGNED
;
221 /* should be called with nsid_lock held */
222 static int __peernet2id(struct net
*net
, struct net
*peer
)
226 return __peernet2id_alloc(net
, peer
, &no
);
229 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
);
230 /* This function returns the id of a peer netns. If no id is assigned, one will
231 * be allocated and returned.
233 int peernet2id_alloc(struct net
*net
, struct net
*peer
)
235 bool alloc
= false, alive
= false;
238 if (refcount_read(&net
->count
) == 0)
239 return NETNSA_NSID_NOT_ASSIGNED
;
240 spin_lock_bh(&net
->nsid_lock
);
242 * When peer is obtained from RCU lists, we may race with
243 * its cleanup. Check whether it's alive, and this guarantees
244 * we never hash a peer back to net->netns_ids, after it has
245 * just been idr_remove()'d from there in cleanup_net().
247 if (maybe_get_net(peer
))
248 alive
= alloc
= true;
249 id
= __peernet2id_alloc(net
, peer
, &alloc
);
250 spin_unlock_bh(&net
->nsid_lock
);
251 if (alloc
&& id
>= 0)
252 rtnl_net_notifyid(net
, RTM_NEWNSID
, id
);
257 EXPORT_SYMBOL_GPL(peernet2id_alloc
);
259 /* This function returns, if assigned, the id of a peer netns. */
260 int peernet2id(struct net
*net
, struct net
*peer
)
264 spin_lock_bh(&net
->nsid_lock
);
265 id
= __peernet2id(net
, peer
);
266 spin_unlock_bh(&net
->nsid_lock
);
269 EXPORT_SYMBOL(peernet2id
);
271 /* This function returns true is the peer netns has an id assigned into the
274 bool peernet_has_id(struct net
*net
, struct net
*peer
)
276 return peernet2id(net
, peer
) >= 0;
279 struct net
*get_net_ns_by_id(struct net
*net
, int id
)
287 peer
= idr_find(&net
->netns_ids
, id
);
289 peer
= maybe_get_net(peer
);
296 * setup_net runs the initializers for the network namespace object.
298 static __net_init
int setup_net(struct net
*net
, struct user_namespace
*user_ns
)
300 /* Must be called with pernet_ops_rwsem held */
301 const struct pernet_operations
*ops
, *saved_ops
;
303 LIST_HEAD(net_exit_list
);
305 refcount_set(&net
->count
, 1);
306 refcount_set(&net
->passive
, 1);
307 net
->dev_base_seq
= 1;
308 net
->user_ns
= user_ns
;
309 idr_init(&net
->netns_ids
);
310 spin_lock_init(&net
->nsid_lock
);
311 mutex_init(&net
->ipv4
.ra_mutex
);
313 list_for_each_entry(ops
, &pernet_list
, list
) {
314 error
= ops_init(ops
, net
);
318 down_write(&net_rwsem
);
319 list_add_tail_rcu(&net
->list
, &net_namespace_list
);
320 up_write(&net_rwsem
);
325 /* Walk through the list backwards calling the exit functions
326 * for the pernet modules whose init functions did not fail.
328 list_add(&net
->exit_list
, &net_exit_list
);
330 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
331 ops_exit_list(ops
, &net_exit_list
);
334 list_for_each_entry_continue_reverse(ops
, &pernet_list
, list
)
335 ops_free_list(ops
, &net_exit_list
);
341 static int __net_init
net_defaults_init_net(struct net
*net
)
343 net
->core
.sysctl_somaxconn
= SOMAXCONN
;
347 static struct pernet_operations net_defaults_ops
= {
348 .init
= net_defaults_init_net
,
351 static __init
int net_defaults_init(void)
353 if (register_pernet_subsys(&net_defaults_ops
))
354 panic("Cannot initialize net default settings");
359 core_initcall(net_defaults_init
);
362 static struct ucounts
*inc_net_namespaces(struct user_namespace
*ns
)
364 return inc_ucount(ns
, current_euid(), UCOUNT_NET_NAMESPACES
);
367 static void dec_net_namespaces(struct ucounts
*ucounts
)
369 dec_ucount(ucounts
, UCOUNT_NET_NAMESPACES
);
372 static struct kmem_cache
*net_cachep __ro_after_init
;
373 static struct workqueue_struct
*netns_wq
;
375 static struct net
*net_alloc(void)
377 struct net
*net
= NULL
;
378 struct net_generic
*ng
;
380 ng
= net_alloc_generic();
384 net
= kmem_cache_zalloc(net_cachep
, GFP_KERNEL
);
388 rcu_assign_pointer(net
->gen
, ng
);
397 static void net_free(struct net
*net
)
399 kfree(rcu_access_pointer(net
->gen
));
400 kmem_cache_free(net_cachep
, net
);
403 void net_drop_ns(void *p
)
406 if (ns
&& refcount_dec_and_test(&ns
->passive
))
410 struct net
*copy_net_ns(unsigned long flags
,
411 struct user_namespace
*user_ns
, struct net
*old_net
)
413 struct ucounts
*ucounts
;
417 if (!(flags
& CLONE_NEWNET
))
418 return get_net(old_net
);
420 ucounts
= inc_net_namespaces(user_ns
);
422 return ERR_PTR(-ENOSPC
);
429 refcount_set(&net
->passive
, 1);
430 net
->ucounts
= ucounts
;
431 get_user_ns(user_ns
);
433 rv
= down_read_killable(&pernet_ops_rwsem
);
437 rv
= setup_net(net
, user_ns
);
439 up_read(&pernet_ops_rwsem
);
443 put_user_ns(user_ns
);
446 dec_net_namespaces(ucounts
);
453 * net_ns_get_ownership - get sysfs ownership data for @net
454 * @net: network namespace in question (can be NULL)
455 * @uid: kernel user ID for sysfs objects
456 * @gid: kernel group ID for sysfs objects
458 * Returns the uid/gid pair of root in the user namespace associated with the
459 * given network namespace.
461 void net_ns_get_ownership(const struct net
*net
, kuid_t
*uid
, kgid_t
*gid
)
464 kuid_t ns_root_uid
= make_kuid(net
->user_ns
, 0);
465 kgid_t ns_root_gid
= make_kgid(net
->user_ns
, 0);
467 if (uid_valid(ns_root_uid
))
470 if (gid_valid(ns_root_gid
))
473 *uid
= GLOBAL_ROOT_UID
;
474 *gid
= GLOBAL_ROOT_GID
;
477 EXPORT_SYMBOL_GPL(net_ns_get_ownership
);
479 static void unhash_nsid(struct net
*net
, struct net
*last
)
482 /* This function is only called from cleanup_net() work,
483 * and this work is the only process, that may delete
484 * a net from net_namespace_list. So, when the below
485 * is executing, the list may only grow. Thus, we do not
486 * use for_each_net_rcu() or net_rwsem.
491 spin_lock_bh(&tmp
->nsid_lock
);
492 id
= __peernet2id(tmp
, net
);
494 idr_remove(&tmp
->netns_ids
, id
);
495 spin_unlock_bh(&tmp
->nsid_lock
);
497 rtnl_net_notifyid(tmp
, RTM_DELNSID
, id
);
501 spin_lock_bh(&net
->nsid_lock
);
502 idr_destroy(&net
->netns_ids
);
503 spin_unlock_bh(&net
->nsid_lock
);
506 static LLIST_HEAD(cleanup_list
);
508 static void cleanup_net(struct work_struct
*work
)
510 const struct pernet_operations
*ops
;
511 struct net
*net
, *tmp
, *last
;
512 struct llist_node
*net_kill_list
;
513 LIST_HEAD(net_exit_list
);
515 /* Atomically snapshot the list of namespaces to cleanup */
516 net_kill_list
= llist_del_all(&cleanup_list
);
518 down_read(&pernet_ops_rwsem
);
520 /* Don't let anyone else find us. */
521 down_write(&net_rwsem
);
522 llist_for_each_entry(net
, net_kill_list
, cleanup_list
)
523 list_del_rcu(&net
->list
);
524 /* Cache last net. After we unlock rtnl, no one new net
525 * added to net_namespace_list can assign nsid pointer
526 * to a net from net_kill_list (see peernet2id_alloc()).
527 * So, we skip them in unhash_nsid().
529 * Note, that unhash_nsid() does not delete nsid links
530 * between net_kill_list's nets, as they've already
531 * deleted from net_namespace_list. But, this would be
532 * useless anyway, as netns_ids are destroyed there.
534 last
= list_last_entry(&net_namespace_list
, struct net
, list
);
535 up_write(&net_rwsem
);
537 llist_for_each_entry(net
, net_kill_list
, cleanup_list
) {
538 unhash_nsid(net
, last
);
539 list_add_tail(&net
->exit_list
, &net_exit_list
);
543 * Another CPU might be rcu-iterating the list, wait for it.
544 * This needs to be before calling the exit() notifiers, so
545 * the rcu_barrier() below isn't sufficient alone.
549 /* Run all of the network namespace exit methods */
550 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
551 ops_exit_list(ops
, &net_exit_list
);
553 /* Free the net generic variables */
554 list_for_each_entry_reverse(ops
, &pernet_list
, list
)
555 ops_free_list(ops
, &net_exit_list
);
557 up_read(&pernet_ops_rwsem
);
559 /* Ensure there are no outstanding rcu callbacks using this
564 /* Finally it is safe to free my network namespace structure */
565 list_for_each_entry_safe(net
, tmp
, &net_exit_list
, exit_list
) {
566 list_del_init(&net
->exit_list
);
567 dec_net_namespaces(net
->ucounts
);
568 put_user_ns(net
->user_ns
);
574 * net_ns_barrier - wait until concurrent net_cleanup_work is done
576 * cleanup_net runs from work queue and will first remove namespaces
577 * from the global list, then run net exit functions.
579 * Call this in module exit path to make sure that all netns
580 * ->exit ops have been invoked before the function is removed.
582 void net_ns_barrier(void)
584 down_write(&pernet_ops_rwsem
);
585 up_write(&pernet_ops_rwsem
);
587 EXPORT_SYMBOL(net_ns_barrier
);
589 static DECLARE_WORK(net_cleanup_work
, cleanup_net
);
591 void __put_net(struct net
*net
)
593 /* Cleanup the network namespace in process context */
594 if (llist_add(&net
->cleanup_list
, &cleanup_list
))
595 queue_work(netns_wq
, &net_cleanup_work
);
597 EXPORT_SYMBOL_GPL(__put_net
);
599 struct net
*get_net_ns_by_fd(int fd
)
602 struct ns_common
*ns
;
605 file
= proc_ns_fget(fd
);
607 return ERR_CAST(file
);
609 ns
= get_proc_ns(file_inode(file
));
610 if (ns
->ops
== &netns_operations
)
611 net
= get_net(container_of(ns
, struct net
, ns
));
613 net
= ERR_PTR(-EINVAL
);
620 struct net
*get_net_ns_by_fd(int fd
)
622 return ERR_PTR(-EINVAL
);
625 EXPORT_SYMBOL_GPL(get_net_ns_by_fd
);
627 struct net
*get_net_ns_by_pid(pid_t pid
)
629 struct task_struct
*tsk
;
632 /* Lookup the network namespace */
633 net
= ERR_PTR(-ESRCH
);
635 tsk
= find_task_by_vpid(pid
);
637 struct nsproxy
*nsproxy
;
639 nsproxy
= tsk
->nsproxy
;
641 net
= get_net(nsproxy
->net_ns
);
647 EXPORT_SYMBOL_GPL(get_net_ns_by_pid
);
649 static __net_init
int net_ns_net_init(struct net
*net
)
652 net
->ns
.ops
= &netns_operations
;
654 return ns_alloc_inum(&net
->ns
);
657 static __net_exit
void net_ns_net_exit(struct net
*net
)
659 ns_free_inum(&net
->ns
);
662 static struct pernet_operations __net_initdata net_ns_ops
= {
663 .init
= net_ns_net_init
,
664 .exit
= net_ns_net_exit
,
667 static const struct nla_policy rtnl_net_policy
[NETNSA_MAX
+ 1] = {
668 [NETNSA_NONE
] = { .type
= NLA_UNSPEC
},
669 [NETNSA_NSID
] = { .type
= NLA_S32
},
670 [NETNSA_PID
] = { .type
= NLA_U32
},
671 [NETNSA_FD
] = { .type
= NLA_U32
},
674 static int rtnl_net_newid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
675 struct netlink_ext_ack
*extack
)
677 struct net
*net
= sock_net(skb
->sk
);
678 struct nlattr
*tb
[NETNSA_MAX
+ 1];
683 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
684 rtnl_net_policy
, extack
);
687 if (!tb
[NETNSA_NSID
]) {
688 NL_SET_ERR_MSG(extack
, "nsid is missing");
691 nsid
= nla_get_s32(tb
[NETNSA_NSID
]);
693 if (tb
[NETNSA_PID
]) {
694 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
695 nla
= tb
[NETNSA_PID
];
696 } else if (tb
[NETNSA_FD
]) {
697 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
700 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
704 NL_SET_BAD_ATTR(extack
, nla
);
705 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
706 return PTR_ERR(peer
);
709 spin_lock_bh(&net
->nsid_lock
);
710 if (__peernet2id(net
, peer
) >= 0) {
711 spin_unlock_bh(&net
->nsid_lock
);
713 NL_SET_BAD_ATTR(extack
, nla
);
714 NL_SET_ERR_MSG(extack
,
715 "Peer netns already has a nsid assigned");
719 err
= alloc_netid(net
, peer
, nsid
);
720 spin_unlock_bh(&net
->nsid_lock
);
722 rtnl_net_notifyid(net
, RTM_NEWNSID
, err
);
724 } else if (err
== -ENOSPC
&& nsid
>= 0) {
726 NL_SET_BAD_ATTR(extack
, tb
[NETNSA_NSID
]);
727 NL_SET_ERR_MSG(extack
, "The specified nsid is already used");
734 static int rtnl_net_get_size(void)
736 return NLMSG_ALIGN(sizeof(struct rtgenmsg
))
737 + nla_total_size(sizeof(s32
)) /* NETNSA_NSID */
741 static int rtnl_net_fill(struct sk_buff
*skb
, u32 portid
, u32 seq
, int flags
,
742 int cmd
, struct net
*net
, int nsid
)
744 struct nlmsghdr
*nlh
;
745 struct rtgenmsg
*rth
;
747 nlh
= nlmsg_put(skb
, portid
, seq
, cmd
, sizeof(*rth
), flags
);
751 rth
= nlmsg_data(nlh
);
752 rth
->rtgen_family
= AF_UNSPEC
;
754 if (nla_put_s32(skb
, NETNSA_NSID
, nsid
))
755 goto nla_put_failure
;
761 nlmsg_cancel(skb
, nlh
);
765 static int rtnl_net_getid(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
766 struct netlink_ext_ack
*extack
)
768 struct net
*net
= sock_net(skb
->sk
);
769 struct nlattr
*tb
[NETNSA_MAX
+ 1];
775 err
= nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, NETNSA_MAX
,
776 rtnl_net_policy
, extack
);
779 if (tb
[NETNSA_PID
]) {
780 peer
= get_net_ns_by_pid(nla_get_u32(tb
[NETNSA_PID
]));
781 nla
= tb
[NETNSA_PID
];
782 } else if (tb
[NETNSA_FD
]) {
783 peer
= get_net_ns_by_fd(nla_get_u32(tb
[NETNSA_FD
]));
786 NL_SET_ERR_MSG(extack
, "Peer netns reference is missing");
791 NL_SET_BAD_ATTR(extack
, nla
);
792 NL_SET_ERR_MSG(extack
, "Peer netns reference is invalid");
793 return PTR_ERR(peer
);
796 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
802 id
= peernet2id(net
, peer
);
803 err
= rtnl_net_fill(msg
, NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
804 RTM_NEWNSID
, net
, id
);
808 err
= rtnl_unicast(msg
, net
, NETLINK_CB(skb
).portid
);
818 struct rtnl_net_dump_cb
{
821 struct netlink_callback
*cb
;
826 static int rtnl_net_dumpid_one(int id
, void *peer
, void *data
)
828 struct rtnl_net_dump_cb
*net_cb
= (struct rtnl_net_dump_cb
*)data
;
831 if (net_cb
->idx
< net_cb
->s_idx
)
834 ret
= rtnl_net_fill(net_cb
->skb
, NETLINK_CB(net_cb
->cb
->skb
).portid
,
835 net_cb
->cb
->nlh
->nlmsg_seq
, NLM_F_MULTI
,
836 RTM_NEWNSID
, net_cb
->net
, id
);
845 static int rtnl_net_dumpid(struct sk_buff
*skb
, struct netlink_callback
*cb
)
847 struct net
*net
= sock_net(skb
->sk
);
848 struct rtnl_net_dump_cb net_cb
= {
853 .s_idx
= cb
->args
[0],
856 spin_lock_bh(&net
->nsid_lock
);
857 idr_for_each(&net
->netns_ids
, rtnl_net_dumpid_one
, &net_cb
);
858 spin_unlock_bh(&net
->nsid_lock
);
860 cb
->args
[0] = net_cb
.idx
;
864 static void rtnl_net_notifyid(struct net
*net
, int cmd
, int id
)
869 msg
= nlmsg_new(rtnl_net_get_size(), GFP_KERNEL
);
873 err
= rtnl_net_fill(msg
, 0, 0, 0, cmd
, net
, id
);
877 rtnl_notify(msg
, net
, 0, RTNLGRP_NSID
, NULL
, 0);
883 rtnl_set_sk_err(net
, RTNLGRP_NSID
, err
);
886 static int __init
net_ns_init(void)
888 struct net_generic
*ng
;
891 net_cachep
= kmem_cache_create("net_namespace", sizeof(struct net
),
893 SLAB_PANIC
|SLAB_ACCOUNT
, NULL
);
895 /* Create workqueue for cleanup */
896 netns_wq
= create_singlethread_workqueue("netns");
898 panic("Could not create netns workq");
901 ng
= net_alloc_generic();
903 panic("Could not allocate generic netns");
905 rcu_assign_pointer(init_net
.gen
, ng
);
907 down_write(&pernet_ops_rwsem
);
908 if (setup_net(&init_net
, &init_user_ns
))
909 panic("Could not setup the initial network namespace");
911 init_net_initialized
= true;
912 up_write(&pernet_ops_rwsem
);
914 register_pernet_subsys(&net_ns_ops
);
916 rtnl_register(PF_UNSPEC
, RTM_NEWNSID
, rtnl_net_newid
, NULL
,
917 RTNL_FLAG_DOIT_UNLOCKED
);
918 rtnl_register(PF_UNSPEC
, RTM_GETNSID
, rtnl_net_getid
, rtnl_net_dumpid
,
919 RTNL_FLAG_DOIT_UNLOCKED
);
924 pure_initcall(net_ns_init
);
927 static int __register_pernet_operations(struct list_head
*list
,
928 struct pernet_operations
*ops
)
932 LIST_HEAD(net_exit_list
);
934 list_add_tail(&ops
->list
, list
);
935 if (ops
->init
|| (ops
->id
&& ops
->size
)) {
936 /* We held write locked pernet_ops_rwsem, and parallel
937 * setup_net() and cleanup_net() are not possible.
940 error
= ops_init(ops
, net
);
943 list_add_tail(&net
->exit_list
, &net_exit_list
);
949 /* If I have an error cleanup all namespaces I initialized */
950 list_del(&ops
->list
);
951 ops_exit_list(ops
, &net_exit_list
);
952 ops_free_list(ops
, &net_exit_list
);
956 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
959 LIST_HEAD(net_exit_list
);
961 list_del(&ops
->list
);
962 /* See comment in __register_pernet_operations() */
964 list_add_tail(&net
->exit_list
, &net_exit_list
);
965 ops_exit_list(ops
, &net_exit_list
);
966 ops_free_list(ops
, &net_exit_list
);
971 static int __register_pernet_operations(struct list_head
*list
,
972 struct pernet_operations
*ops
)
974 if (!init_net_initialized
) {
975 list_add_tail(&ops
->list
, list
);
979 return ops_init(ops
, &init_net
);
982 static void __unregister_pernet_operations(struct pernet_operations
*ops
)
984 if (!init_net_initialized
) {
985 list_del(&ops
->list
);
987 LIST_HEAD(net_exit_list
);
988 list_add(&init_net
.exit_list
, &net_exit_list
);
989 ops_exit_list(ops
, &net_exit_list
);
990 ops_free_list(ops
, &net_exit_list
);
994 #endif /* CONFIG_NET_NS */
996 static DEFINE_IDA(net_generic_ids
);
998 static int register_pernet_operations(struct list_head
*list
,
999 struct pernet_operations
*ops
)
1004 error
= ida_alloc_min(&net_generic_ids
, MIN_PERNET_OPS_ID
,
1009 max_gen_ptrs
= max(max_gen_ptrs
, *ops
->id
+ 1);
1011 error
= __register_pernet_operations(list
, ops
);
1015 ida_free(&net_generic_ids
, *ops
->id
);
1021 static void unregister_pernet_operations(struct pernet_operations
*ops
)
1023 __unregister_pernet_operations(ops
);
1026 ida_free(&net_generic_ids
, *ops
->id
);
1030 * register_pernet_subsys - register a network namespace subsystem
1031 * @ops: pernet operations structure for the subsystem
1033 * Register a subsystem which has init and exit functions
1034 * that are called when network namespaces are created and
1035 * destroyed respectively.
1037 * When registered all network namespace init functions are
1038 * called for every existing network namespace. Allowing kernel
1039 * modules to have a race free view of the set of network namespaces.
1041 * When a new network namespace is created all of the init
1042 * methods are called in the order in which they were registered.
1044 * When a network namespace is destroyed all of the exit methods
1045 * are called in the reverse of the order with which they were
1048 int register_pernet_subsys(struct pernet_operations
*ops
)
1051 down_write(&pernet_ops_rwsem
);
1052 error
= register_pernet_operations(first_device
, ops
);
1053 up_write(&pernet_ops_rwsem
);
1056 EXPORT_SYMBOL_GPL(register_pernet_subsys
);
1059 * unregister_pernet_subsys - unregister a network namespace subsystem
1060 * @ops: pernet operations structure to manipulate
1062 * Remove the pernet operations structure from the list to be
1063 * used when network namespaces are created or destroyed. In
1064 * addition run the exit method for all existing network
1067 void unregister_pernet_subsys(struct pernet_operations
*ops
)
1069 down_write(&pernet_ops_rwsem
);
1070 unregister_pernet_operations(ops
);
1071 up_write(&pernet_ops_rwsem
);
1073 EXPORT_SYMBOL_GPL(unregister_pernet_subsys
);
1076 * register_pernet_device - register a network namespace device
1077 * @ops: pernet operations structure for the subsystem
1079 * Register a device which has init and exit functions
1080 * that are called when network namespaces are created and
1081 * destroyed respectively.
1083 * When registered all network namespace init functions are
1084 * called for every existing network namespace. Allowing kernel
1085 * modules to have a race free view of the set of network namespaces.
1087 * When a new network namespace is created all of the init
1088 * methods are called in the order in which they were registered.
1090 * When a network namespace is destroyed all of the exit methods
1091 * are called in the reverse of the order with which they were
1094 int register_pernet_device(struct pernet_operations
*ops
)
1097 down_write(&pernet_ops_rwsem
);
1098 error
= register_pernet_operations(&pernet_list
, ops
);
1099 if (!error
&& (first_device
== &pernet_list
))
1100 first_device
= &ops
->list
;
1101 up_write(&pernet_ops_rwsem
);
1104 EXPORT_SYMBOL_GPL(register_pernet_device
);
1107 * unregister_pernet_device - unregister a network namespace netdevice
1108 * @ops: pernet operations structure to manipulate
1110 * Remove the pernet operations structure from the list to be
1111 * used when network namespaces are created or destroyed. In
1112 * addition run the exit method for all existing network
1115 void unregister_pernet_device(struct pernet_operations
*ops
)
1117 down_write(&pernet_ops_rwsem
);
1118 if (&ops
->list
== first_device
)
1119 first_device
= first_device
->next
;
1120 unregister_pernet_operations(ops
);
1121 up_write(&pernet_ops_rwsem
);
1123 EXPORT_SYMBOL_GPL(unregister_pernet_device
);
1125 #ifdef CONFIG_NET_NS
1126 static struct ns_common
*netns_get(struct task_struct
*task
)
1128 struct net
*net
= NULL
;
1129 struct nsproxy
*nsproxy
;
1132 nsproxy
= task
->nsproxy
;
1134 net
= get_net(nsproxy
->net_ns
);
1137 return net
? &net
->ns
: NULL
;
1140 static inline struct net
*to_net_ns(struct ns_common
*ns
)
1142 return container_of(ns
, struct net
, ns
);
1145 static void netns_put(struct ns_common
*ns
)
1147 put_net(to_net_ns(ns
));
1150 static int netns_install(struct nsproxy
*nsproxy
, struct ns_common
*ns
)
1152 struct net
*net
= to_net_ns(ns
);
1154 if (!ns_capable(net
->user_ns
, CAP_SYS_ADMIN
) ||
1155 !ns_capable(current_user_ns(), CAP_SYS_ADMIN
))
1158 put_net(nsproxy
->net_ns
);
1159 nsproxy
->net_ns
= get_net(net
);
1163 static struct user_namespace
*netns_owner(struct ns_common
*ns
)
1165 return to_net_ns(ns
)->user_ns
;
1168 const struct proc_ns_operations netns_operations
= {
1170 .type
= CLONE_NEWNET
,
1173 .install
= netns_install
,
1174 .owner
= netns_owner
,