This source file includes following definitions.
- net_alloc_generic
- net_assign_generic
- ops_init
- ops_free
- ops_pre_exit_list
- ops_exit_list
- ops_free_list
- alloc_netid
- net_eq_idr
- __peernet2id_alloc
- __peernet2id
- peernet2id_alloc
- peernet2id
- peernet_has_id
- get_net_ns_by_id
- setup_net
- net_defaults_init_net
- net_defaults_init
- inc_net_namespaces
- dec_net_namespaces
- net_alloc
- net_free
- net_drop_ns
- copy_net_ns
- net_ns_get_ownership
- unhash_nsid
- cleanup_net
- net_ns_barrier
- __put_net
- get_net_ns_by_fd
- get_net_ns_by_fd
- get_net_ns_by_pid
- net_ns_net_init
- net_ns_net_exit
- rtnl_net_newid
- rtnl_net_get_size
- rtnl_net_fill
- rtnl_net_valid_getid_req
- rtnl_net_getid
- rtnl_net_dumpid_one
- rtnl_valid_dump_net_req
- rtnl_net_dumpid
- rtnl_net_notifyid
- net_ns_init
- __register_pernet_operations
- __unregister_pernet_operations
- __register_pernet_operations
- __unregister_pernet_operations
- register_pernet_operations
- unregister_pernet_operations
- register_pernet_subsys
- unregister_pernet_subsys
- register_pernet_device
- unregister_pernet_device
- netns_get
- to_net_ns
- netns_put
- netns_install
- netns_owner
1
2 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3
4 #include <linux/workqueue.h>
5 #include <linux/rtnetlink.h>
6 #include <linux/cache.h>
7 #include <linux/slab.h>
8 #include <linux/list.h>
9 #include <linux/delay.h>
10 #include <linux/sched.h>
11 #include <linux/idr.h>
12 #include <linux/rculist.h>
13 #include <linux/nsproxy.h>
14 #include <linux/fs.h>
15 #include <linux/proc_ns.h>
16 #include <linux/file.h>
17 #include <linux/export.h>
18 #include <linux/user_namespace.h>
19 #include <linux/net_namespace.h>
20 #include <linux/sched/task.h>
21 #include <linux/uidgid.h>
22
23 #include <net/sock.h>
24 #include <net/netlink.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27
28
29
30
31
32 static LIST_HEAD(pernet_list);
33 static struct list_head *first_device = &pernet_list;
34
35 LIST_HEAD(net_namespace_list);
36 EXPORT_SYMBOL_GPL(net_namespace_list);
37
38
39 DECLARE_RWSEM(net_rwsem);
40 EXPORT_SYMBOL_GPL(net_rwsem);
41
42 #ifdef CONFIG_KEYS
43 static struct key_tag init_net_key_domain = { .usage = REFCOUNT_INIT(1) };
44 #endif
45
46 struct net init_net = {
47 .count = REFCOUNT_INIT(1),
48 .dev_base_head = LIST_HEAD_INIT(init_net.dev_base_head),
49 #ifdef CONFIG_KEYS
50 .key_domain = &init_net_key_domain,
51 #endif
52 };
53 EXPORT_SYMBOL(init_net);
54
55 static bool init_net_initialized;
56
57
58
59
60
61
62 DECLARE_RWSEM(pernet_ops_rwsem);
63 EXPORT_SYMBOL_GPL(pernet_ops_rwsem);
64
65 #define MIN_PERNET_OPS_ID \
66 ((sizeof(struct net_generic) + sizeof(void *) - 1) / sizeof(void *))
67
68 #define INITIAL_NET_GEN_PTRS 13
69
70 static unsigned int max_gen_ptrs = INITIAL_NET_GEN_PTRS;
71
72 static struct net_generic *net_alloc_generic(void)
73 {
74 struct net_generic *ng;
75 unsigned int generic_size = offsetof(struct net_generic, ptr[max_gen_ptrs]);
76
77 ng = kzalloc(generic_size, GFP_KERNEL);
78 if (ng)
79 ng->s.len = max_gen_ptrs;
80
81 return ng;
82 }
83
84 static int net_assign_generic(struct net *net, unsigned int id, void *data)
85 {
86 struct net_generic *ng, *old_ng;
87
88 BUG_ON(id < MIN_PERNET_OPS_ID);
89
90 old_ng = rcu_dereference_protected(net->gen,
91 lockdep_is_held(&pernet_ops_rwsem));
92 if (old_ng->s.len > id) {
93 old_ng->ptr[id] = data;
94 return 0;
95 }
96
97 ng = net_alloc_generic();
98 if (ng == NULL)
99 return -ENOMEM;
100
101
102
103
104
105
106
107
108
109
110
111
112 memcpy(&ng->ptr[MIN_PERNET_OPS_ID], &old_ng->ptr[MIN_PERNET_OPS_ID],
113 (old_ng->s.len - MIN_PERNET_OPS_ID) * sizeof(void *));
114 ng->ptr[id] = data;
115
116 rcu_assign_pointer(net->gen, ng);
117 kfree_rcu(old_ng, s.rcu);
118 return 0;
119 }
120
121 static int ops_init(const struct pernet_operations *ops, struct net *net)
122 {
123 int err = -ENOMEM;
124 void *data = NULL;
125
126 if (ops->id && ops->size) {
127 data = kzalloc(ops->size, GFP_KERNEL);
128 if (!data)
129 goto out;
130
131 err = net_assign_generic(net, *ops->id, data);
132 if (err)
133 goto cleanup;
134 }
135 err = 0;
136 if (ops->init)
137 err = ops->init(net);
138 if (!err)
139 return 0;
140
141 cleanup:
142 kfree(data);
143
144 out:
145 return err;
146 }
147
148 static void ops_free(const struct pernet_operations *ops, struct net *net)
149 {
150 if (ops->id && ops->size) {
151 kfree(net_generic(net, *ops->id));
152 }
153 }
154
155 static void ops_pre_exit_list(const struct pernet_operations *ops,
156 struct list_head *net_exit_list)
157 {
158 struct net *net;
159
160 if (ops->pre_exit) {
161 list_for_each_entry(net, net_exit_list, exit_list)
162 ops->pre_exit(net);
163 }
164 }
165
166 static void ops_exit_list(const struct pernet_operations *ops,
167 struct list_head *net_exit_list)
168 {
169 struct net *net;
170 if (ops->exit) {
171 list_for_each_entry(net, net_exit_list, exit_list)
172 ops->exit(net);
173 }
174 if (ops->exit_batch)
175 ops->exit_batch(net_exit_list);
176 }
177
178 static void ops_free_list(const struct pernet_operations *ops,
179 struct list_head *net_exit_list)
180 {
181 struct net *net;
182 if (ops->size && ops->id) {
183 list_for_each_entry(net, net_exit_list, exit_list)
184 ops_free(ops, net);
185 }
186 }
187
188
189 static int alloc_netid(struct net *net, struct net *peer, int reqid)
190 {
191 int min = 0, max = 0;
192
193 if (reqid >= 0) {
194 min = reqid;
195 max = reqid + 1;
196 }
197
198 return idr_alloc(&net->netns_ids, peer, min, max, GFP_ATOMIC);
199 }
200
201
202
203
204
205
206 #define NET_ID_ZERO -1
207 static int net_eq_idr(int id, void *net, void *peer)
208 {
209 if (net_eq(net, peer))
210 return id ? : NET_ID_ZERO;
211 return 0;
212 }
213
214
215
216
217
218 static int __peernet2id_alloc(struct net *net, struct net *peer, bool *alloc)
219 {
220 int id = idr_for_each(&net->netns_ids, net_eq_idr, peer);
221 bool alloc_it = *alloc;
222
223 *alloc = false;
224
225
226 if (id == NET_ID_ZERO)
227 return 0;
228 if (id > 0)
229 return id;
230
231 if (alloc_it) {
232 id = alloc_netid(net, peer, -1);
233 *alloc = true;
234 return id >= 0 ? id : NETNSA_NSID_NOT_ASSIGNED;
235 }
236
237 return NETNSA_NSID_NOT_ASSIGNED;
238 }
239
240
241 static int __peernet2id(struct net *net, struct net *peer)
242 {
243 bool no = false;
244
245 return __peernet2id_alloc(net, peer, &no);
246 }
247
248 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
249 struct nlmsghdr *nlh, gfp_t gfp);
250
251
252
253 int peernet2id_alloc(struct net *net, struct net *peer, gfp_t gfp)
254 {
255 bool alloc = false, alive = false;
256 int id;
257
258 if (refcount_read(&net->count) == 0)
259 return NETNSA_NSID_NOT_ASSIGNED;
260 spin_lock_bh(&net->nsid_lock);
261
262
263
264
265
266
267 if (maybe_get_net(peer))
268 alive = alloc = true;
269 id = __peernet2id_alloc(net, peer, &alloc);
270 spin_unlock_bh(&net->nsid_lock);
271 if (alloc && id >= 0)
272 rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL, gfp);
273 if (alive)
274 put_net(peer);
275 return id;
276 }
277 EXPORT_SYMBOL_GPL(peernet2id_alloc);
278
279
280 int peernet2id(struct net *net, struct net *peer)
281 {
282 int id;
283
284 spin_lock_bh(&net->nsid_lock);
285 id = __peernet2id(net, peer);
286 spin_unlock_bh(&net->nsid_lock);
287 return id;
288 }
289 EXPORT_SYMBOL(peernet2id);
290
291
292
293
294 bool peernet_has_id(struct net *net, struct net *peer)
295 {
296 return peernet2id(net, peer) >= 0;
297 }
298
299 struct net *get_net_ns_by_id(struct net *net, int id)
300 {
301 struct net *peer;
302
303 if (id < 0)
304 return NULL;
305
306 rcu_read_lock();
307 peer = idr_find(&net->netns_ids, id);
308 if (peer)
309 peer = maybe_get_net(peer);
310 rcu_read_unlock();
311
312 return peer;
313 }
314
315
316
317
318 static __net_init int setup_net(struct net *net, struct user_namespace *user_ns)
319 {
320
321 const struct pernet_operations *ops, *saved_ops;
322 int error = 0;
323 LIST_HEAD(net_exit_list);
324
325 refcount_set(&net->count, 1);
326 refcount_set(&net->passive, 1);
327 get_random_bytes(&net->hash_mix, sizeof(u32));
328 net->dev_base_seq = 1;
329 net->user_ns = user_ns;
330 idr_init(&net->netns_ids);
331 spin_lock_init(&net->nsid_lock);
332 mutex_init(&net->ipv4.ra_mutex);
333
334 list_for_each_entry(ops, &pernet_list, list) {
335 error = ops_init(ops, net);
336 if (error < 0)
337 goto out_undo;
338 }
339 down_write(&net_rwsem);
340 list_add_tail_rcu(&net->list, &net_namespace_list);
341 up_write(&net_rwsem);
342 out:
343 return error;
344
345 out_undo:
346
347
348
349 list_add(&net->exit_list, &net_exit_list);
350 saved_ops = ops;
351 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
352 ops_pre_exit_list(ops, &net_exit_list);
353
354 synchronize_rcu();
355
356 ops = saved_ops;
357 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
358 ops_exit_list(ops, &net_exit_list);
359
360 ops = saved_ops;
361 list_for_each_entry_continue_reverse(ops, &pernet_list, list)
362 ops_free_list(ops, &net_exit_list);
363
364 rcu_barrier();
365 goto out;
366 }
367
368 static int __net_init net_defaults_init_net(struct net *net)
369 {
370 net->core.sysctl_somaxconn = SOMAXCONN;
371 return 0;
372 }
373
374 static struct pernet_operations net_defaults_ops = {
375 .init = net_defaults_init_net,
376 };
377
378 static __init int net_defaults_init(void)
379 {
380 if (register_pernet_subsys(&net_defaults_ops))
381 panic("Cannot initialize net default settings");
382
383 return 0;
384 }
385
386 core_initcall(net_defaults_init);
387
388 #ifdef CONFIG_NET_NS
389 static struct ucounts *inc_net_namespaces(struct user_namespace *ns)
390 {
391 return inc_ucount(ns, current_euid(), UCOUNT_NET_NAMESPACES);
392 }
393
394 static void dec_net_namespaces(struct ucounts *ucounts)
395 {
396 dec_ucount(ucounts, UCOUNT_NET_NAMESPACES);
397 }
398
399 static struct kmem_cache *net_cachep __ro_after_init;
400 static struct workqueue_struct *netns_wq;
401
402 static struct net *net_alloc(void)
403 {
404 struct net *net = NULL;
405 struct net_generic *ng;
406
407 ng = net_alloc_generic();
408 if (!ng)
409 goto out;
410
411 net = kmem_cache_zalloc(net_cachep, GFP_KERNEL);
412 if (!net)
413 goto out_free;
414
415 #ifdef CONFIG_KEYS
416 net->key_domain = kzalloc(sizeof(struct key_tag), GFP_KERNEL);
417 if (!net->key_domain)
418 goto out_free_2;
419 refcount_set(&net->key_domain->usage, 1);
420 #endif
421
422 rcu_assign_pointer(net->gen, ng);
423 out:
424 return net;
425
426 #ifdef CONFIG_KEYS
427 out_free_2:
428 kmem_cache_free(net_cachep, net);
429 net = NULL;
430 #endif
431 out_free:
432 kfree(ng);
433 goto out;
434 }
435
436 static void net_free(struct net *net)
437 {
438 kfree(rcu_access_pointer(net->gen));
439 kmem_cache_free(net_cachep, net);
440 }
441
442 void net_drop_ns(void *p)
443 {
444 struct net *ns = p;
445 if (ns && refcount_dec_and_test(&ns->passive))
446 net_free(ns);
447 }
448
449 struct net *copy_net_ns(unsigned long flags,
450 struct user_namespace *user_ns, struct net *old_net)
451 {
452 struct ucounts *ucounts;
453 struct net *net;
454 int rv;
455
456 if (!(flags & CLONE_NEWNET))
457 return get_net(old_net);
458
459 ucounts = inc_net_namespaces(user_ns);
460 if (!ucounts)
461 return ERR_PTR(-ENOSPC);
462
463 net = net_alloc();
464 if (!net) {
465 rv = -ENOMEM;
466 goto dec_ucounts;
467 }
468 refcount_set(&net->passive, 1);
469 net->ucounts = ucounts;
470 get_user_ns(user_ns);
471
472 rv = down_read_killable(&pernet_ops_rwsem);
473 if (rv < 0)
474 goto put_userns;
475
476 rv = setup_net(net, user_ns);
477
478 up_read(&pernet_ops_rwsem);
479
480 if (rv < 0) {
481 put_userns:
482 key_remove_domain(net->key_domain);
483 put_user_ns(user_ns);
484 net_drop_ns(net);
485 dec_ucounts:
486 dec_net_namespaces(ucounts);
487 return ERR_PTR(rv);
488 }
489 return net;
490 }
491
492
493
494
495
496
497
498
499
500
501 void net_ns_get_ownership(const struct net *net, kuid_t *uid, kgid_t *gid)
502 {
503 if (net) {
504 kuid_t ns_root_uid = make_kuid(net->user_ns, 0);
505 kgid_t ns_root_gid = make_kgid(net->user_ns, 0);
506
507 if (uid_valid(ns_root_uid))
508 *uid = ns_root_uid;
509
510 if (gid_valid(ns_root_gid))
511 *gid = ns_root_gid;
512 } else {
513 *uid = GLOBAL_ROOT_UID;
514 *gid = GLOBAL_ROOT_GID;
515 }
516 }
517 EXPORT_SYMBOL_GPL(net_ns_get_ownership);
518
519 static void unhash_nsid(struct net *net, struct net *last)
520 {
521 struct net *tmp;
522
523
524
525
526
527
528 for_each_net(tmp) {
529 int id;
530
531 spin_lock_bh(&tmp->nsid_lock);
532 id = __peernet2id(tmp, net);
533 if (id >= 0)
534 idr_remove(&tmp->netns_ids, id);
535 spin_unlock_bh(&tmp->nsid_lock);
536 if (id >= 0)
537 rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL,
538 GFP_KERNEL);
539 if (tmp == last)
540 break;
541 }
542 spin_lock_bh(&net->nsid_lock);
543 idr_destroy(&net->netns_ids);
544 spin_unlock_bh(&net->nsid_lock);
545 }
546
547 static LLIST_HEAD(cleanup_list);
548
549 static void cleanup_net(struct work_struct *work)
550 {
551 const struct pernet_operations *ops;
552 struct net *net, *tmp, *last;
553 struct llist_node *net_kill_list;
554 LIST_HEAD(net_exit_list);
555
556
557 net_kill_list = llist_del_all(&cleanup_list);
558
559 down_read(&pernet_ops_rwsem);
560
561
562 down_write(&net_rwsem);
563 llist_for_each_entry(net, net_kill_list, cleanup_list)
564 list_del_rcu(&net->list);
565
566
567
568
569
570
571
572
573
574
575 last = list_last_entry(&net_namespace_list, struct net, list);
576 up_write(&net_rwsem);
577
578 llist_for_each_entry(net, net_kill_list, cleanup_list) {
579 unhash_nsid(net, last);
580 list_add_tail(&net->exit_list, &net_exit_list);
581 }
582
583
584 list_for_each_entry_reverse(ops, &pernet_list, list)
585 ops_pre_exit_list(ops, &net_exit_list);
586
587
588
589
590
591
592
593 synchronize_rcu();
594
595
596 list_for_each_entry_reverse(ops, &pernet_list, list)
597 ops_exit_list(ops, &net_exit_list);
598
599
600 list_for_each_entry_reverse(ops, &pernet_list, list)
601 ops_free_list(ops, &net_exit_list);
602
603 up_read(&pernet_ops_rwsem);
604
605
606
607
608 rcu_barrier();
609
610
611 list_for_each_entry_safe(net, tmp, &net_exit_list, exit_list) {
612 list_del_init(&net->exit_list);
613 dec_net_namespaces(net->ucounts);
614 key_remove_domain(net->key_domain);
615 put_user_ns(net->user_ns);
616 net_drop_ns(net);
617 }
618 }
619
620
621
622
623
624
625
626
627
628
629 void net_ns_barrier(void)
630 {
631 down_write(&pernet_ops_rwsem);
632 up_write(&pernet_ops_rwsem);
633 }
634 EXPORT_SYMBOL(net_ns_barrier);
635
636 static DECLARE_WORK(net_cleanup_work, cleanup_net);
637
638 void __put_net(struct net *net)
639 {
640
641 if (llist_add(&net->cleanup_list, &cleanup_list))
642 queue_work(netns_wq, &net_cleanup_work);
643 }
644 EXPORT_SYMBOL_GPL(__put_net);
645
646 struct net *get_net_ns_by_fd(int fd)
647 {
648 struct file *file;
649 struct ns_common *ns;
650 struct net *net;
651
652 file = proc_ns_fget(fd);
653 if (IS_ERR(file))
654 return ERR_CAST(file);
655
656 ns = get_proc_ns(file_inode(file));
657 if (ns->ops == &netns_operations)
658 net = get_net(container_of(ns, struct net, ns));
659 else
660 net = ERR_PTR(-EINVAL);
661
662 fput(file);
663 return net;
664 }
665
666 #else
667 struct net *get_net_ns_by_fd(int fd)
668 {
669 return ERR_PTR(-EINVAL);
670 }
671 #endif
672 EXPORT_SYMBOL_GPL(get_net_ns_by_fd);
673
674 struct net *get_net_ns_by_pid(pid_t pid)
675 {
676 struct task_struct *tsk;
677 struct net *net;
678
679
680 net = ERR_PTR(-ESRCH);
681 rcu_read_lock();
682 tsk = find_task_by_vpid(pid);
683 if (tsk) {
684 struct nsproxy *nsproxy;
685 task_lock(tsk);
686 nsproxy = tsk->nsproxy;
687 if (nsproxy)
688 net = get_net(nsproxy->net_ns);
689 task_unlock(tsk);
690 }
691 rcu_read_unlock();
692 return net;
693 }
694 EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
695
696 static __net_init int net_ns_net_init(struct net *net)
697 {
698 #ifdef CONFIG_NET_NS
699 net->ns.ops = &netns_operations;
700 #endif
701 return ns_alloc_inum(&net->ns);
702 }
703
704 static __net_exit void net_ns_net_exit(struct net *net)
705 {
706 ns_free_inum(&net->ns);
707 }
708
709 static struct pernet_operations __net_initdata net_ns_ops = {
710 .init = net_ns_net_init,
711 .exit = net_ns_net_exit,
712 };
713
714 static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
715 [NETNSA_NONE] = { .type = NLA_UNSPEC },
716 [NETNSA_NSID] = { .type = NLA_S32 },
717 [NETNSA_PID] = { .type = NLA_U32 },
718 [NETNSA_FD] = { .type = NLA_U32 },
719 [NETNSA_TARGET_NSID] = { .type = NLA_S32 },
720 };
721
722 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
723 struct netlink_ext_ack *extack)
724 {
725 struct net *net = sock_net(skb->sk);
726 struct nlattr *tb[NETNSA_MAX + 1];
727 struct nlattr *nla;
728 struct net *peer;
729 int nsid, err;
730
731 err = nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg), tb,
732 NETNSA_MAX, rtnl_net_policy, extack);
733 if (err < 0)
734 return err;
735 if (!tb[NETNSA_NSID]) {
736 NL_SET_ERR_MSG(extack, "nsid is missing");
737 return -EINVAL;
738 }
739 nsid = nla_get_s32(tb[NETNSA_NSID]);
740
741 if (tb[NETNSA_PID]) {
742 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
743 nla = tb[NETNSA_PID];
744 } else if (tb[NETNSA_FD]) {
745 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
746 nla = tb[NETNSA_FD];
747 } else {
748 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
749 return -EINVAL;
750 }
751 if (IS_ERR(peer)) {
752 NL_SET_BAD_ATTR(extack, nla);
753 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
754 return PTR_ERR(peer);
755 }
756
757 spin_lock_bh(&net->nsid_lock);
758 if (__peernet2id(net, peer) >= 0) {
759 spin_unlock_bh(&net->nsid_lock);
760 err = -EEXIST;
761 NL_SET_BAD_ATTR(extack, nla);
762 NL_SET_ERR_MSG(extack,
763 "Peer netns already has a nsid assigned");
764 goto out;
765 }
766
767 err = alloc_netid(net, peer, nsid);
768 spin_unlock_bh(&net->nsid_lock);
769 if (err >= 0) {
770 rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
771 nlh, GFP_KERNEL);
772 err = 0;
773 } else if (err == -ENOSPC && nsid >= 0) {
774 err = -EEXIST;
775 NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
776 NL_SET_ERR_MSG(extack, "The specified nsid is already used");
777 }
778 out:
779 put_net(peer);
780 return err;
781 }
782
783 static int rtnl_net_get_size(void)
784 {
785 return NLMSG_ALIGN(sizeof(struct rtgenmsg))
786 + nla_total_size(sizeof(s32))
787 + nla_total_size(sizeof(s32))
788 ;
789 }
790
791 struct net_fill_args {
792 u32 portid;
793 u32 seq;
794 int flags;
795 int cmd;
796 int nsid;
797 bool add_ref;
798 int ref_nsid;
799 };
800
801 static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
802 {
803 struct nlmsghdr *nlh;
804 struct rtgenmsg *rth;
805
806 nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
807 args->flags);
808 if (!nlh)
809 return -EMSGSIZE;
810
811 rth = nlmsg_data(nlh);
812 rth->rtgen_family = AF_UNSPEC;
813
814 if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
815 goto nla_put_failure;
816
817 if (args->add_ref &&
818 nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
819 goto nla_put_failure;
820
821 nlmsg_end(skb, nlh);
822 return 0;
823
824 nla_put_failure:
825 nlmsg_cancel(skb, nlh);
826 return -EMSGSIZE;
827 }
828
829 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
830 const struct nlmsghdr *nlh,
831 struct nlattr **tb,
832 struct netlink_ext_ack *extack)
833 {
834 int i, err;
835
836 if (!netlink_strict_get_check(skb))
837 return nlmsg_parse_deprecated(nlh, sizeof(struct rtgenmsg),
838 tb, NETNSA_MAX, rtnl_net_policy,
839 extack);
840
841 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
842 NETNSA_MAX, rtnl_net_policy,
843 extack);
844 if (err)
845 return err;
846
847 for (i = 0; i <= NETNSA_MAX; i++) {
848 if (!tb[i])
849 continue;
850
851 switch (i) {
852 case NETNSA_PID:
853 case NETNSA_FD:
854 case NETNSA_NSID:
855 case NETNSA_TARGET_NSID:
856 break;
857 default:
858 NL_SET_ERR_MSG(extack, "Unsupported attribute in peer netns getid request");
859 return -EINVAL;
860 }
861 }
862
863 return 0;
864 }
865
866 static int rtnl_net_getid(struct sk_buff *skb, struct nlmsghdr *nlh,
867 struct netlink_ext_ack *extack)
868 {
869 struct net *net = sock_net(skb->sk);
870 struct nlattr *tb[NETNSA_MAX + 1];
871 struct net_fill_args fillargs = {
872 .portid = NETLINK_CB(skb).portid,
873 .seq = nlh->nlmsg_seq,
874 .cmd = RTM_NEWNSID,
875 };
876 struct net *peer, *target = net;
877 struct nlattr *nla;
878 struct sk_buff *msg;
879 int err;
880
881 err = rtnl_net_valid_getid_req(skb, nlh, tb, extack);
882 if (err < 0)
883 return err;
884 if (tb[NETNSA_PID]) {
885 peer = get_net_ns_by_pid(nla_get_u32(tb[NETNSA_PID]));
886 nla = tb[NETNSA_PID];
887 } else if (tb[NETNSA_FD]) {
888 peer = get_net_ns_by_fd(nla_get_u32(tb[NETNSA_FD]));
889 nla = tb[NETNSA_FD];
890 } else if (tb[NETNSA_NSID]) {
891 peer = get_net_ns_by_id(net, nla_get_s32(tb[NETNSA_NSID]));
892 if (!peer)
893 peer = ERR_PTR(-ENOENT);
894 nla = tb[NETNSA_NSID];
895 } else {
896 NL_SET_ERR_MSG(extack, "Peer netns reference is missing");
897 return -EINVAL;
898 }
899
900 if (IS_ERR(peer)) {
901 NL_SET_BAD_ATTR(extack, nla);
902 NL_SET_ERR_MSG(extack, "Peer netns reference is invalid");
903 return PTR_ERR(peer);
904 }
905
906 if (tb[NETNSA_TARGET_NSID]) {
907 int id = nla_get_s32(tb[NETNSA_TARGET_NSID]);
908
909 target = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, id);
910 if (IS_ERR(target)) {
911 NL_SET_BAD_ATTR(extack, tb[NETNSA_TARGET_NSID]);
912 NL_SET_ERR_MSG(extack,
913 "Target netns reference is invalid");
914 err = PTR_ERR(target);
915 goto out;
916 }
917 fillargs.add_ref = true;
918 fillargs.ref_nsid = peernet2id(net, peer);
919 }
920
921 msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
922 if (!msg) {
923 err = -ENOMEM;
924 goto out;
925 }
926
927 fillargs.nsid = peernet2id(target, peer);
928 err = rtnl_net_fill(msg, &fillargs);
929 if (err < 0)
930 goto err_out;
931
932 err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
933 goto out;
934
935 err_out:
936 nlmsg_free(msg);
937 out:
938 if (fillargs.add_ref)
939 put_net(target);
940 put_net(peer);
941 return err;
942 }
943
944 struct rtnl_net_dump_cb {
945 struct net *tgt_net;
946 struct net *ref_net;
947 struct sk_buff *skb;
948 struct net_fill_args fillargs;
949 int idx;
950 int s_idx;
951 };
952
953 static int rtnl_net_dumpid_one(int id, void *peer, void *data)
954 {
955 struct rtnl_net_dump_cb *net_cb = (struct rtnl_net_dump_cb *)data;
956 int ret;
957
958 if (net_cb->idx < net_cb->s_idx)
959 goto cont;
960
961 net_cb->fillargs.nsid = id;
962 if (net_cb->fillargs.add_ref)
963 net_cb->fillargs.ref_nsid = __peernet2id(net_cb->ref_net, peer);
964 ret = rtnl_net_fill(net_cb->skb, &net_cb->fillargs);
965 if (ret < 0)
966 return ret;
967
968 cont:
969 net_cb->idx++;
970 return 0;
971 }
972
973 static int rtnl_valid_dump_net_req(const struct nlmsghdr *nlh, struct sock *sk,
974 struct rtnl_net_dump_cb *net_cb,
975 struct netlink_callback *cb)
976 {
977 struct netlink_ext_ack *extack = cb->extack;
978 struct nlattr *tb[NETNSA_MAX + 1];
979 int err, i;
980
981 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct rtgenmsg), tb,
982 NETNSA_MAX, rtnl_net_policy,
983 extack);
984 if (err < 0)
985 return err;
986
987 for (i = 0; i <= NETNSA_MAX; i++) {
988 if (!tb[i])
989 continue;
990
991 if (i == NETNSA_TARGET_NSID) {
992 struct net *net;
993
994 net = rtnl_get_net_ns_capable(sk, nla_get_s32(tb[i]));
995 if (IS_ERR(net)) {
996 NL_SET_BAD_ATTR(extack, tb[i]);
997 NL_SET_ERR_MSG(extack,
998 "Invalid target network namespace id");
999 return PTR_ERR(net);
1000 }
1001 net_cb->fillargs.add_ref = true;
1002 net_cb->ref_net = net_cb->tgt_net;
1003 net_cb->tgt_net = net;
1004 } else {
1005 NL_SET_BAD_ATTR(extack, tb[i]);
1006 NL_SET_ERR_MSG(extack,
1007 "Unsupported attribute in dump request");
1008 return -EINVAL;
1009 }
1010 }
1011
1012 return 0;
1013 }
1014
1015 static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
1016 {
1017 struct rtnl_net_dump_cb net_cb = {
1018 .tgt_net = sock_net(skb->sk),
1019 .skb = skb,
1020 .fillargs = {
1021 .portid = NETLINK_CB(cb->skb).portid,
1022 .seq = cb->nlh->nlmsg_seq,
1023 .flags = NLM_F_MULTI,
1024 .cmd = RTM_NEWNSID,
1025 },
1026 .idx = 0,
1027 .s_idx = cb->args[0],
1028 };
1029 int err = 0;
1030
1031 if (cb->strict_check) {
1032 err = rtnl_valid_dump_net_req(cb->nlh, skb->sk, &net_cb, cb);
1033 if (err < 0)
1034 goto end;
1035 }
1036
1037 spin_lock_bh(&net_cb.tgt_net->nsid_lock);
1038 if (net_cb.fillargs.add_ref &&
1039 !net_eq(net_cb.ref_net, net_cb.tgt_net) &&
1040 !spin_trylock_bh(&net_cb.ref_net->nsid_lock)) {
1041 spin_unlock_bh(&net_cb.tgt_net->nsid_lock);
1042 err = -EAGAIN;
1043 goto end;
1044 }
1045 idr_for_each(&net_cb.tgt_net->netns_ids, rtnl_net_dumpid_one, &net_cb);
1046 if (net_cb.fillargs.add_ref &&
1047 !net_eq(net_cb.ref_net, net_cb.tgt_net))
1048 spin_unlock_bh(&net_cb.ref_net->nsid_lock);
1049 spin_unlock_bh(&net_cb.tgt_net->nsid_lock);
1050
1051 cb->args[0] = net_cb.idx;
1052 end:
1053 if (net_cb.fillargs.add_ref)
1054 put_net(net_cb.tgt_net);
1055 return err < 0 ? err : skb->len;
1056 }
1057
1058 static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
1059 struct nlmsghdr *nlh, gfp_t gfp)
1060 {
1061 struct net_fill_args fillargs = {
1062 .portid = portid,
1063 .seq = nlh ? nlh->nlmsg_seq : 0,
1064 .cmd = cmd,
1065 .nsid = id,
1066 };
1067 struct sk_buff *msg;
1068 int err = -ENOMEM;
1069
1070 msg = nlmsg_new(rtnl_net_get_size(), gfp);
1071 if (!msg)
1072 goto out;
1073
1074 err = rtnl_net_fill(msg, &fillargs);
1075 if (err < 0)
1076 goto err_out;
1077
1078 rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, gfp);
1079 return;
1080
1081 err_out:
1082 nlmsg_free(msg);
1083 out:
1084 rtnl_set_sk_err(net, RTNLGRP_NSID, err);
1085 }
1086
1087 static int __init net_ns_init(void)
1088 {
1089 struct net_generic *ng;
1090
1091 #ifdef CONFIG_NET_NS
1092 net_cachep = kmem_cache_create("net_namespace", sizeof(struct net),
1093 SMP_CACHE_BYTES,
1094 SLAB_PANIC|SLAB_ACCOUNT, NULL);
1095
1096
1097 netns_wq = create_singlethread_workqueue("netns");
1098 if (!netns_wq)
1099 panic("Could not create netns workq");
1100 #endif
1101
1102 ng = net_alloc_generic();
1103 if (!ng)
1104 panic("Could not allocate generic netns");
1105
1106 rcu_assign_pointer(init_net.gen, ng);
1107
1108 down_write(&pernet_ops_rwsem);
1109 if (setup_net(&init_net, &init_user_ns))
1110 panic("Could not setup the initial network namespace");
1111
1112 init_net_initialized = true;
1113 up_write(&pernet_ops_rwsem);
1114
1115 if (register_pernet_subsys(&net_ns_ops))
1116 panic("Could not register network namespace subsystems");
1117
1118 rtnl_register(PF_UNSPEC, RTM_NEWNSID, rtnl_net_newid, NULL,
1119 RTNL_FLAG_DOIT_UNLOCKED);
1120 rtnl_register(PF_UNSPEC, RTM_GETNSID, rtnl_net_getid, rtnl_net_dumpid,
1121 RTNL_FLAG_DOIT_UNLOCKED);
1122
1123 return 0;
1124 }
1125
1126 pure_initcall(net_ns_init);
1127
1128 #ifdef CONFIG_NET_NS
1129 static int __register_pernet_operations(struct list_head *list,
1130 struct pernet_operations *ops)
1131 {
1132 struct net *net;
1133 int error;
1134 LIST_HEAD(net_exit_list);
1135
1136 list_add_tail(&ops->list, list);
1137 if (ops->init || (ops->id && ops->size)) {
1138
1139
1140
1141 for_each_net(net) {
1142 error = ops_init(ops, net);
1143 if (error)
1144 goto out_undo;
1145 list_add_tail(&net->exit_list, &net_exit_list);
1146 }
1147 }
1148 return 0;
1149
1150 out_undo:
1151
1152 list_del(&ops->list);
1153 ops_pre_exit_list(ops, &net_exit_list);
1154 synchronize_rcu();
1155 ops_exit_list(ops, &net_exit_list);
1156 ops_free_list(ops, &net_exit_list);
1157 return error;
1158 }
1159
1160 static void __unregister_pernet_operations(struct pernet_operations *ops)
1161 {
1162 struct net *net;
1163 LIST_HEAD(net_exit_list);
1164
1165 list_del(&ops->list);
1166
1167 for_each_net(net)
1168 list_add_tail(&net->exit_list, &net_exit_list);
1169 ops_pre_exit_list(ops, &net_exit_list);
1170 synchronize_rcu();
1171 ops_exit_list(ops, &net_exit_list);
1172 ops_free_list(ops, &net_exit_list);
1173 }
1174
1175 #else
1176
1177 static int __register_pernet_operations(struct list_head *list,
1178 struct pernet_operations *ops)
1179 {
1180 if (!init_net_initialized) {
1181 list_add_tail(&ops->list, list);
1182 return 0;
1183 }
1184
1185 return ops_init(ops, &init_net);
1186 }
1187
1188 static void __unregister_pernet_operations(struct pernet_operations *ops)
1189 {
1190 if (!init_net_initialized) {
1191 list_del(&ops->list);
1192 } else {
1193 LIST_HEAD(net_exit_list);
1194 list_add(&init_net.exit_list, &net_exit_list);
1195 ops_pre_exit_list(ops, &net_exit_list);
1196 synchronize_rcu();
1197 ops_exit_list(ops, &net_exit_list);
1198 ops_free_list(ops, &net_exit_list);
1199 }
1200 }
1201
1202 #endif
1203
1204 static DEFINE_IDA(net_generic_ids);
1205
1206 static int register_pernet_operations(struct list_head *list,
1207 struct pernet_operations *ops)
1208 {
1209 int error;
1210
1211 if (ops->id) {
1212 error = ida_alloc_min(&net_generic_ids, MIN_PERNET_OPS_ID,
1213 GFP_KERNEL);
1214 if (error < 0)
1215 return error;
1216 *ops->id = error;
1217 max_gen_ptrs = max(max_gen_ptrs, *ops->id + 1);
1218 }
1219 error = __register_pernet_operations(list, ops);
1220 if (error) {
1221 rcu_barrier();
1222 if (ops->id)
1223 ida_free(&net_generic_ids, *ops->id);
1224 }
1225
1226 return error;
1227 }
1228
1229 static void unregister_pernet_operations(struct pernet_operations *ops)
1230 {
1231 __unregister_pernet_operations(ops);
1232 rcu_barrier();
1233 if (ops->id)
1234 ida_free(&net_generic_ids, *ops->id);
1235 }
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256 int register_pernet_subsys(struct pernet_operations *ops)
1257 {
1258 int error;
1259 down_write(&pernet_ops_rwsem);
1260 error = register_pernet_operations(first_device, ops);
1261 up_write(&pernet_ops_rwsem);
1262 return error;
1263 }
1264 EXPORT_SYMBOL_GPL(register_pernet_subsys);
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275 void unregister_pernet_subsys(struct pernet_operations *ops)
1276 {
1277 down_write(&pernet_ops_rwsem);
1278 unregister_pernet_operations(ops);
1279 up_write(&pernet_ops_rwsem);
1280 }
1281 EXPORT_SYMBOL_GPL(unregister_pernet_subsys);
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302 int register_pernet_device(struct pernet_operations *ops)
1303 {
1304 int error;
1305 down_write(&pernet_ops_rwsem);
1306 error = register_pernet_operations(&pernet_list, ops);
1307 if (!error && (first_device == &pernet_list))
1308 first_device = &ops->list;
1309 up_write(&pernet_ops_rwsem);
1310 return error;
1311 }
1312 EXPORT_SYMBOL_GPL(register_pernet_device);
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323 void unregister_pernet_device(struct pernet_operations *ops)
1324 {
1325 down_write(&pernet_ops_rwsem);
1326 if (&ops->list == first_device)
1327 first_device = first_device->next;
1328 unregister_pernet_operations(ops);
1329 up_write(&pernet_ops_rwsem);
1330 }
1331 EXPORT_SYMBOL_GPL(unregister_pernet_device);
1332
1333 #ifdef CONFIG_NET_NS
1334 static struct ns_common *netns_get(struct task_struct *task)
1335 {
1336 struct net *net = NULL;
1337 struct nsproxy *nsproxy;
1338
1339 task_lock(task);
1340 nsproxy = task->nsproxy;
1341 if (nsproxy)
1342 net = get_net(nsproxy->net_ns);
1343 task_unlock(task);
1344
1345 return net ? &net->ns : NULL;
1346 }
1347
1348 static inline struct net *to_net_ns(struct ns_common *ns)
1349 {
1350 return container_of(ns, struct net, ns);
1351 }
1352
1353 static void netns_put(struct ns_common *ns)
1354 {
1355 put_net(to_net_ns(ns));
1356 }
1357
1358 static int netns_install(struct nsproxy *nsproxy, struct ns_common *ns)
1359 {
1360 struct net *net = to_net_ns(ns);
1361
1362 if (!ns_capable(net->user_ns, CAP_SYS_ADMIN) ||
1363 !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
1364 return -EPERM;
1365
1366 put_net(nsproxy->net_ns);
1367 nsproxy->net_ns = get_net(net);
1368 return 0;
1369 }
1370
1371 static struct user_namespace *netns_owner(struct ns_common *ns)
1372 {
1373 return to_net_ns(ns)->user_ns;
1374 }
1375
1376 const struct proc_ns_operations netns_operations = {
1377 .name = "net",
1378 .type = CLONE_NEWNET,
1379 .get = netns_get,
1380 .put = netns_put,
1381 .install = netns_install,
1382 .owner = netns_owner,
1383 };
1384 #endif