1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * NETLINK Kernel-user communication protocol.
4 *
5 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
6 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
7 * Patrick McHardy <kaber@trash.net>
8 *
9 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
10 * added netlink_proto_exit
11 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
12 * use nlk_sk, as sk->protinfo is on a diet 8)
13 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
14 * - inc module use count of module that owns
15 * the kernel socket in case userspace opens
16 * socket of same protocol
17 * - remove all module support, since netlink is
18 * mandatory if CONFIG_NET=y these days
19 */
20
21 #include <linux/module.h>
22
23 #include <linux/bpf.h>
24 #include <linux/capability.h>
25 #include <linux/kernel.h>
26 #include <linux/filter.h>
27 #include <linux/init.h>
28 #include <linux/signal.h>
29 #include <linux/sched.h>
30 #include <linux/errno.h>
31 #include <linux/string.h>
32 #include <linux/stat.h>
33 #include <linux/socket.h>
34 #include <linux/un.h>
35 #include <linux/fcntl.h>
36 #include <linux/termios.h>
37 #include <linux/sockios.h>
38 #include <linux/net.h>
39 #include <linux/fs.h>
40 #include <linux/slab.h>
41 #include <linux/uaccess.h>
42 #include <linux/skbuff.h>
43 #include <linux/netdevice.h>
44 #include <linux/rtnetlink.h>
45 #include <linux/proc_fs.h>
46 #include <linux/seq_file.h>
47 #include <linux/notifier.h>
48 #include <linux/security.h>
49 #include <linux/jhash.h>
50 #include <linux/jiffies.h>
51 #include <linux/random.h>
52 #include <linux/bitops.h>
53 #include <linux/mm.h>
54 #include <linux/types.h>
55 #include <linux/audit.h>
56 #include <linux/mutex.h>
57 #include <linux/vmalloc.h>
58 #include <linux/if_arp.h>
59 #include <linux/rhashtable.h>
60 #include <asm/cacheflush.h>
61 #include <linux/hash.h>
62 #include <linux/net_namespace.h>
63 #include <linux/nospec.h>
64 #include <linux/btf_ids.h>
65
66 #include <net/net_namespace.h>
67 #include <net/netns/generic.h>
68 #include <net/sock.h>
69 #include <net/scm.h>
70 #include <net/netlink.h>
71 #define CREATE_TRACE_POINTS
72 #include <trace/events/netlink.h>
73
74 #include "af_netlink.h"
75 #include "genetlink.h"
76
77 struct listeners {
78 struct rcu_head rcu;
79 unsigned long masks[];
80 };
81
82 /* state bits */
83 #define NETLINK_S_CONGESTED 0x0
84
85 static inline int netlink_is_kernel(struct sock *sk)
86 {
87 return nlk_test_bit(KERNEL_SOCKET, sk);
88 }
89
90 struct netlink_table *nl_table __read_mostly;
91 EXPORT_SYMBOL_GPL(nl_table);
92
93 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
94
95 static struct lock_class_key nlk_cb_mutex_keys[MAX_LINKS];
96
97 static const char *const nlk_cb_mutex_key_strings[MAX_LINKS + 1] = {
98 "nlk_cb_mutex-ROUTE",
99 "nlk_cb_mutex-1",
100 "nlk_cb_mutex-USERSOCK",
101 "nlk_cb_mutex-FIREWALL",
102 "nlk_cb_mutex-SOCK_DIAG",
103 "nlk_cb_mutex-NFLOG",
104 "nlk_cb_mutex-XFRM",
105 "nlk_cb_mutex-SELINUX",
106 "nlk_cb_mutex-ISCSI",
107 "nlk_cb_mutex-AUDIT",
108 "nlk_cb_mutex-FIB_LOOKUP",
109 "nlk_cb_mutex-CONNECTOR",
110 "nlk_cb_mutex-NETFILTER",
111 "nlk_cb_mutex-IP6_FW",
112 "nlk_cb_mutex-DNRTMSG",
113 "nlk_cb_mutex-KOBJECT_UEVENT",
114 "nlk_cb_mutex-GENERIC",
115 "nlk_cb_mutex-17",
116 "nlk_cb_mutex-SCSITRANSPORT",
117 "nlk_cb_mutex-ECRYPTFS",
118 "nlk_cb_mutex-RDMA",
119 "nlk_cb_mutex-CRYPTO",
120 "nlk_cb_mutex-SMC",
121 "nlk_cb_mutex-23",
122 "nlk_cb_mutex-24",
123 "nlk_cb_mutex-25",
124 "nlk_cb_mutex-26",
125 "nlk_cb_mutex-27",
126 "nlk_cb_mutex-28",
127 "nlk_cb_mutex-29",
128 "nlk_cb_mutex-30",
129 "nlk_cb_mutex-31",
130 "nlk_cb_mutex-MAX_LINKS"
131 };
132
133 static int netlink_dump(struct sock *sk, bool lock_taken);
134
135 /* nl_table locking explained:
136 * Lookup and traversal are protected with an RCU read-side lock. Insertion
137 * and removal are protected with per bucket lock while using RCU list
138 * modification primitives and may run in parallel to RCU protected lookups.
139 * Destruction of the Netlink socket may only occur *after* nl_table_lock has
140 * been acquired * either during or after the socket has been removed from
141 * the list and after an RCU grace period.
142 */
143 DEFINE_RWLOCK(nl_table_lock);
144 EXPORT_SYMBOL_GPL(nl_table_lock);
145 static atomic_t nl_table_users = ATOMIC_INIT(0);
146
147 #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
148
149 static BLOCKING_NOTIFIER_HEAD(netlink_chain);
150
151
152 static const struct rhashtable_params netlink_rhashtable_params;
153
154 void do_trace_netlink_extack(const char *msg)
155 {
156 trace_netlink_extack(msg);
157 }
158 EXPORT_SYMBOL(do_trace_netlink_extack);
159
160 static inline u32 netlink_group_mask(u32 group)
161 {
162 if (group > 32)
163 return 0;
164 return group ? 1 << (group - 1) : 0;
165 }
166
167 static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb,
168 gfp_t gfp_mask)
169 {
170 unsigned int len = skb->len;
171 struct sk_buff *new;
172
173 new = alloc_skb(len, gfp_mask);
174 if (new == NULL)
175 return NULL;
176
177 NETLINK_CB(new).portid = NETLINK_CB(skb).portid;
178 NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group;
179 NETLINK_CB(new).creds = NETLINK_CB(skb).creds;
180
181 skb_put_data(new, skb->data, len);
182 return new;
183 }
184
185 static unsigned int netlink_tap_net_id;
186
187 struct netlink_tap_net {
188 struct list_head netlink_tap_all;
189 struct mutex netlink_tap_lock;
190 };
191
192 int netlink_add_tap(struct netlink_tap *nt)
193 {
194 struct net *net = dev_net(nt->dev);
195 struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
196
197 if (unlikely(nt->dev->type != ARPHRD_NETLINK))
198 return -EINVAL;
199
200 mutex_lock(&nn->netlink_tap_lock);
201 list_add_rcu(&nt->list, &nn->netlink_tap_all);
202 mutex_unlock(&nn->netlink_tap_lock);
203
204 __module_get(nt->module);
205
206 return 0;
207 }
208 EXPORT_SYMBOL_GPL(netlink_add_tap);
209
210 static int __netlink_remove_tap(struct netlink_tap *nt)
211 {
212 struct net *net = dev_net(nt->dev);
213 struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
214 bool found = false;
215 struct netlink_tap *tmp;
216
217 mutex_lock(&nn->netlink_tap_lock);
218
219 list_for_each_entry(tmp, &nn->netlink_tap_all, list) {
220 if (nt == tmp) {
221 list_del_rcu(&nt->list);
222 found = true;
223 goto out;
224 }
225 }
226
227 pr_warn("__netlink_remove_tap: %p not found\n", nt);
228 out:
229 mutex_unlock(&nn->netlink_tap_lock);
230
231 if (found)
232 module_put(nt->module);
233
234 return found ? 0 : -ENODEV;
235 }
236
237 int netlink_remove_tap(struct netlink_tap *nt)
238 {
239 int ret;
240
241 ret = __netlink_remove_tap(nt);
242 synchronize_net();
243
244 return ret;
245 }
246 EXPORT_SYMBOL_GPL(netlink_remove_tap);
247
248 static __net_init int netlink_tap_init_net(struct net *net)
249 {
250 struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
251
252 INIT_LIST_HEAD(&nn->netlink_tap_all);
253 mutex_init(&nn->netlink_tap_lock);
254 return 0;
255 }
256
257 static struct pernet_operations netlink_tap_net_ops = {
258 .init = netlink_tap_init_net,
259 .id = &netlink_tap_net_id,
260 .size = sizeof(struct netlink_tap_net),
261 };
262
263 static bool netlink_filter_tap(const struct sk_buff *skb)
264 {
265 struct sock *sk = skb->sk;
266
267 /* We take the more conservative approach and
268 * whitelist socket protocols that may pass.
269 */
270 switch (sk->sk_protocol) {
271 case NETLINK_ROUTE:
272 case NETLINK_USERSOCK:
273 case NETLINK_SOCK_DIAG:
274 case NETLINK_NFLOG:
275 case NETLINK_XFRM:
276 case NETLINK_FIB_LOOKUP:
277 case NETLINK_NETFILTER:
278 case NETLINK_GENERIC:
279 return true;
280 }
281
282 return false;
283 }
284
285 static int __netlink_deliver_tap_skb(struct sk_buff *skb,
286 struct net_device *dev)
287 {
288 struct sk_buff *nskb;
289 struct sock *sk = skb->sk;
290 int ret = -ENOMEM;
291
292 if (!net_eq(dev_net(dev), sock_net(sk)))
293 return 0;
294
295 dev_hold(dev);
296
297 if (is_vmalloc_addr(skb->head))
298 nskb = netlink_to_full_skb(skb, GFP_ATOMIC);
299 else
300 nskb = skb_clone(skb, GFP_ATOMIC);
301 if (nskb) {
302 nskb->dev = dev;
303 nskb->protocol = htons((u16) sk->sk_protocol);
304 nskb->pkt_type = netlink_is_kernel(sk) ?
305 PACKET_KERNEL : PACKET_USER;
306 skb_reset_network_header(nskb);
307 ret = dev_queue_xmit(nskb);
308 if (unlikely(ret > 0))
309 ret = net_xmit_errno(ret);
310 }
311
312 dev_put(dev);
313 return ret;
314 }
315
316 static void __netlink_deliver_tap(struct sk_buff *skb, struct netlink_tap_net *nn)
317 {
318 int ret;
319 struct netlink_tap *tmp;
320
321 if (!netlink_filter_tap(skb))
322 return;
323
324 list_for_each_entry_rcu(tmp, &nn->netlink_tap_all, list) {
325 ret = __netlink_deliver_tap_skb(skb, tmp->dev);
326 if (unlikely(ret))
327 break;
328 }
329 }
330
331 static void netlink_deliver_tap(struct net *net, struct sk_buff *skb)
332 {
333 struct netlink_tap_net *nn = net_generic(net, netlink_tap_net_id);
334
335 rcu_read_lock();
336
337 if (unlikely(!list_empty(&nn->netlink_tap_all)))
338 __netlink_deliver_tap(skb, nn);
339
340 rcu_read_unlock();
341 }
342
343 static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src,
344 struct sk_buff *skb)
345 {
346 if (!(netlink_is_kernel(dst) && netlink_is_kernel(src)))
347 netlink_deliver_tap(sock_net(dst), skb);
348 }
349
350 static void netlink_overrun(struct sock *sk)
351 {
352 if (!nlk_test_bit(RECV_NO_ENOBUFS, sk)) {
353 if (!test_and_set_bit(NETLINK_S_CONGESTED,
354 &nlk_sk(sk)->state)) {
355 WRITE_ONCE(sk->sk_err, ENOBUFS);
356 sk_error_report(sk);
357 }
358 }
359 atomic_inc(&sk->sk_drops);
360 }
361
362 static void netlink_rcv_wake(struct sock *sk)
363 {
364 struct netlink_sock *nlk = nlk_sk(sk);
365
366 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
367 clear_bit(NETLINK_S_CONGESTED, &nlk->state);
368 if (!test_bit(NETLINK_S_CONGESTED, &nlk->state))
369 wake_up_interruptible(&nlk->wait);
370 }
371
372 static void netlink_skb_destructor(struct sk_buff *skb)
373 {
374 if (is_vmalloc_addr(skb->head)) {
375 if (!skb->cloned ||
376 !atomic_dec_return(&(skb_shinfo(skb)->dataref)))
377 vfree_atomic(skb->head);
378
379 skb->head = NULL;
380 }
381 if (skb->sk != NULL)
382 sock_rfree(skb);
383 }
384
385 static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
386 {
387 WARN_ON(skb->sk != NULL);
388 skb->sk = sk;
389 skb->destructor = netlink_skb_destructor;
390 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
391 sk_mem_charge(sk, skb->truesize);
392 }
393
394 static void netlink_sock_destruct(struct sock *sk)
395 {
396 struct netlink_sock *nlk = nlk_sk(sk);
397
398 if (nlk->cb_running) {
399 if (nlk->cb.done)
400 nlk->cb.done(&nlk->cb);
401 module_put(nlk->cb.module);
402 kfree_skb(nlk->cb.skb);
403 }
404
405 skb_queue_purge(&sk->sk_receive_queue);
406
407 if (!sock_flag(sk, SOCK_DEAD)) {
408 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
409 return;
410 }
411
412 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
413 WARN_ON(refcount_read(&sk->sk_wmem_alloc));
414 WARN_ON(nlk_sk(sk)->groups);
415 }
416
417 static void netlink_sock_destruct_work(struct work_struct *work)
418 {
419 struct netlink_sock *nlk = container_of(work, struct netlink_sock,
420 work);
421
422 sk_free(&nlk->sk);
423 }
424
425 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
426 * SMP. Look, when several writers sleep and reader wakes them up, all but one
427 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
428 * this, _but_ remember, it adds useless work on UP machines.
429 */
430
431 void netlink_table_grab(void)
432 __acquires(nl_table_lock)
433 {
434 might_sleep();
435
436 write_lock_irq(&nl_table_lock);
437
438 if (atomic_read(&nl_table_users)) {
439 DECLARE_WAITQUEUE(wait, current);
440
441 add_wait_queue_exclusive(&nl_table_wait, &wait);
442 for (;;) {
443 set_current_state(TASK_UNINTERRUPTIBLE);
444 if (atomic_read(&nl_table_users) == 0)
445 break;
446 write_unlock_irq(&nl_table_lock);
447 schedule();
448 write_lock_irq(&nl_table_lock);
449 }
450
451 __set_current_state(TASK_RUNNING);
452 remove_wait_queue(&nl_table_wait, &wait);
453 }
454 }
455
456 void netlink_table_ungrab(void)
457 __releases(nl_table_lock)
458 {
459 write_unlock_irq(&nl_table_lock);
460 wake_up(&nl_table_wait);
461 }
462
463 static inline void
464 netlink_lock_table(void)
465 {
466 unsigned long flags;
467
468 /* read_lock() synchronizes us to netlink_table_grab */
469
470 read_lock_irqsave(&nl_table_lock, flags);
471 atomic_inc(&nl_table_users);
472 read_unlock_irqrestore(&nl_table_lock, flags);
473 }
474
475 static inline void
476 netlink_unlock_table(void)
477 {
478 if (atomic_dec_and_test(&nl_table_users))
479 wake_up(&nl_table_wait);
480 }
481
482 struct netlink_compare_arg
483 {
484 possible_net_t pnet;
485 u32 portid;
486 };
487
488 /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
489 #define netlink_compare_arg_len \
490 (offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
491
492 static inline int netlink_compare(struct rhashtable_compare_arg *arg,
493 const void *ptr)
494 {
495 const struct netlink_compare_arg *x = arg->key;
496 const struct netlink_sock *nlk = ptr;
497
498 return nlk->portid != x->portid ||
499 !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet));
500 }
501
502 static void netlink_compare_arg_init(struct netlink_compare_arg *arg,
503 struct net *net, u32 portid)
504 {
505 memset(arg, 0, sizeof(*arg));
506 write_pnet(&arg->pnet, net);
507 arg->portid = portid;
508 }
509
510 static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
511 struct net *net)
512 {
513 struct netlink_compare_arg arg;
514
515 netlink_compare_arg_init(&arg, net, portid);
516 return rhashtable_lookup_fast(&table->hash, &arg,
517 netlink_rhashtable_params);
518 }
519
520 static int __netlink_insert(struct netlink_table *table, struct sock *sk)
521 {
522 struct netlink_compare_arg arg;
523
524 netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid);
525 return rhashtable_lookup_insert_key(&table->hash, &arg,
526 &nlk_sk(sk)->node,
527 netlink_rhashtable_params);
528 }
529
530 static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
531 {
532 struct netlink_table *table = &nl_table[protocol];
533 struct sock *sk;
534
535 rcu_read_lock();
536 sk = __netlink_lookup(table, portid, net);
537 if (sk)
538 sock_hold(sk);
539 rcu_read_unlock();
540
541 return sk;
542 }
543
544 static const struct proto_ops netlink_ops;
545
546 static void
547 netlink_update_listeners(struct sock *sk)
548 {
549 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
550 unsigned long mask;
551 unsigned int i;
552 struct listeners *listeners;
553
554 listeners = nl_deref_protected(tbl->listeners);
555 if (!listeners)
556 return;
557
558 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
559 mask = 0;
560 sk_for_each_bound(sk, &tbl->mc_list) {
561 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
562 mask |= nlk_sk(sk)->groups[i];
563 }
564 listeners->masks[i] = mask;
565 }
566 /* this function is only called with the netlink table "grabbed", which
567 * makes sure updates are visible before bind or setsockopt return. */
568 }
569
570 static int netlink_insert(struct sock *sk, u32 portid)
571 {
572 struct netlink_table *table = &nl_table[sk->sk_protocol];
573 int err;
574
575 lock_sock(sk);
576
577 err = nlk_sk(sk)->portid == portid ? 0 : -EBUSY;
578 if (nlk_sk(sk)->bound)
579 goto err;
580
581 /* portid can be read locklessly from netlink_getname(). */
582 WRITE_ONCE(nlk_sk(sk)->portid, portid);
583
584 sock_hold(sk);
585
586 err = __netlink_insert(table, sk);
587 if (err) {
588 /* In case the hashtable backend returns with -EBUSY
589 * from here, it must not escape to the caller.
590 */
591 if (unlikely(err == -EBUSY))
592 err = -EOVERFLOW;
593 if (err == -EEXIST)
594 err = -EADDRINUSE;
595 sock_put(sk);
596 goto err;
597 }
598
599 /* We need to ensure that the socket is hashed and visible. */
600 smp_wmb();
601 /* Paired with lockless reads from netlink_bind(),
602 * netlink_connect() and netlink_sendmsg().
603 */
604 WRITE_ONCE(nlk_sk(sk)->bound, portid);
605
606 err:
607 release_sock(sk);
608 return err;
609 }
610
611 static void netlink_remove(struct sock *sk)
612 {
613 struct netlink_table *table;
614
615 table = &nl_table[sk->sk_protocol];
616 if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node,
617 netlink_rhashtable_params)) {
618 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
619 __sock_put(sk);
620 }
621
622 netlink_table_grab();
623 if (nlk_sk(sk)->subscriptions) {
624 __sk_del_bind_node(sk);
625 netlink_update_listeners(sk);
626 }
627 if (sk->sk_protocol == NETLINK_GENERIC)
628 atomic_inc(&genl_sk_destructing_cnt);
629 netlink_table_ungrab();
630 }
631
632 static struct proto netlink_proto = {
633 .name = "NETLINK",
634 .owner = THIS_MODULE,
635 .obj_size = sizeof(struct netlink_sock),
636 };
637
638 static int __netlink_create(struct net *net, struct socket *sock,
639 int protocol, int kern)
640 {
641 struct sock *sk;
642 struct netlink_sock *nlk;
643
644 sock->ops = &netlink_ops;
645
646 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto, kern);
647 if (!sk)
648 return -ENOMEM;
649
650 sock_init_data(sock, sk);
651
652 nlk = nlk_sk(sk);
653 mutex_init(&nlk->nl_cb_mutex);
654 lockdep_set_class_and_name(&nlk->nl_cb_mutex,
655 nlk_cb_mutex_keys + protocol,
656 nlk_cb_mutex_key_strings[protocol]);
657 init_waitqueue_head(&nlk->wait);
658
659 sk->sk_destruct = netlink_sock_destruct;
660 sk->sk_protocol = protocol;
661 return 0;
662 }
663
664 static int netlink_create(struct net *net, struct socket *sock, int protocol,
665 int kern)
666 {
667 struct module *module = NULL;
668 struct netlink_sock *nlk;
669 int (*bind)(struct net *net, int group);
670 void (*unbind)(struct net *net, int group);
671 void (*release)(struct sock *sock, unsigned long *groups);
672 int err = 0;
673
674 sock->state = SS_UNCONNECTED;
675
676 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
677 return -ESOCKTNOSUPPORT;
678
679 if (protocol < 0 || protocol >= MAX_LINKS)
680 return -EPROTONOSUPPORT;
681 protocol = array_index_nospec(protocol, MAX_LINKS);
682
683 netlink_lock_table();
684 #ifdef CONFIG_MODULES
685 if (!nl_table[protocol].registered) {
686 netlink_unlock_table();
687 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
688 netlink_lock_table();
689 }
690 #endif
691 if (nl_table[protocol].registered &&
692 try_module_get(nl_table[protocol].module))
693 module = nl_table[protocol].module;
694 else
695 err = -EPROTONOSUPPORT;
696 bind = nl_table[protocol].bind;
697 unbind = nl_table[protocol].unbind;
698 release = nl_table[protocol].release;
699 netlink_unlock_table();
700
701 if (err < 0)
702 goto out;
703
704 err = __netlink_create(net, sock, protocol, kern);
705 if (err < 0)
706 goto out_module;
707
708 sock_prot_inuse_add(net, &netlink_proto, 1);
709
710 nlk = nlk_sk(sock->sk);
711 nlk->module = module;
712 nlk->netlink_bind = bind;
713 nlk->netlink_unbind = unbind;
714 nlk->netlink_release = release;
715 out:
716 return err;
717
718 out_module:
719 module_put(module);
720 goto out;
721 }
722
723 static void deferred_put_nlk_sk(struct rcu_head *head)
724 {
725 struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu);
726 struct sock *sk = &nlk->sk;
727
728 kfree(nlk->groups);
729 nlk->groups = NULL;
730
731 if (!refcount_dec_and_test(&sk->sk_refcnt))
732 return;
733
734 if (nlk->cb_running && nlk->cb.done) {
735 INIT_WORK(&nlk->work, netlink_sock_destruct_work);
736 schedule_work(&nlk->work);
737 return;
738 }
739
740 sk_free(sk);
741 }
742
743 static int netlink_release(struct socket *sock)
744 {
745 struct sock *sk = sock->sk;
746 struct netlink_sock *nlk;
747
748 if (!sk)
749 return 0;
750
751 netlink_remove(sk);
752 sock_orphan(sk);
753 nlk = nlk_sk(sk);
754
755 /*
756 * OK. Socket is unlinked, any packets that arrive now
757 * will be purged.
758 */
759 if (nlk->netlink_release)
760 nlk->netlink_release(sk, nlk->groups);
761
762 /* must not acquire netlink_table_lock in any way again before unbind
763 * and notifying genetlink is done as otherwise it might deadlock
764 */
765 if (nlk->netlink_unbind) {
766 int i;
767
768 for (i = 0; i < nlk->ngroups; i++)
769 if (test_bit(i, nlk->groups))
770 nlk->netlink_unbind(sock_net(sk), i + 1);
771 }
772 if (sk->sk_protocol == NETLINK_GENERIC &&
773 atomic_dec_return(&genl_sk_destructing_cnt) == 0)
774 wake_up(&genl_sk_destructing_waitq);
775
776 sock->sk = NULL;
777 wake_up_interruptible_all(&nlk->wait);
778
779 skb_queue_purge(&sk->sk_write_queue);
780
781 if (nlk->portid && nlk->bound) {
782 struct netlink_notify n = {
783 .net = sock_net(sk),
784 .protocol = sk->sk_protocol,
785 .portid = nlk->portid,
786 };
787 blocking_notifier_call_chain(&netlink_chain,
788 NETLINK_URELEASE, &n);
789 }
790
791 module_put(nlk->module);
792
793 if (netlink_is_kernel(sk)) {
794 netlink_table_grab();
795 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
796 if (--nl_table[sk->sk_protocol].registered == 0) {
797 struct listeners *old;
798
799 old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
800 RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
801 kfree_rcu(old, rcu);
802 nl_table[sk->sk_protocol].module = NULL;
803 nl_table[sk->sk_protocol].bind = NULL;
804 nl_table[sk->sk_protocol].unbind = NULL;
805 nl_table[sk->sk_protocol].flags = 0;
806 nl_table[sk->sk_protocol].registered = 0;
807 }
808 netlink_table_ungrab();
809 }
810
811 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
812
813 /* Because struct net might disappear soon, do not keep a pointer. */
814 if (!sk->sk_net_refcnt && sock_net(sk) != &init_net) {
815 __netns_tracker_free(sock_net(sk), &sk->ns_tracker, false);
816 /* Because of deferred_put_nlk_sk and use of work queue,
817 * it is possible netns will be freed before this socket.
818 */
819 sock_net_set(sk, &init_net);
820 __netns_tracker_alloc(&init_net, &sk->ns_tracker,
821 false, GFP_KERNEL);
822 }
823 call_rcu(&nlk->rcu, deferred_put_nlk_sk);
824 return 0;
825 }
826
827 static int netlink_autobind(struct socket *sock)
828 {
829 struct sock *sk = sock->sk;
830 struct net *net = sock_net(sk);
831 struct netlink_table *table = &nl_table[sk->sk_protocol];
832 s32 portid = task_tgid_vnr(current);
833 int err;
834 s32 rover = -4096;
835 bool ok;
836
837 retry:
838 cond_resched();
839 rcu_read_lock();
840 ok = !__netlink_lookup(table, portid, net);
841 rcu_read_unlock();
842 if (!ok) {
843 /* Bind collision, search negative portid values. */
844 if (rover == -4096)
845 /* rover will be in range [S32_MIN, -4097] */
846 rover = S32_MIN + get_random_u32_below(-4096 - S32_MIN);
847 else if (rover >= -4096)
848 rover = -4097;
849 portid = rover--;
850 goto retry;
851 }
852
853 err = netlink_insert(sk, portid);
854 if (err == -EADDRINUSE)
855 goto retry;
856
857 /* If 2 threads race to autobind, that is fine. */
858 if (err == -EBUSY)
859 err = 0;
860
861 return err;
862 }
863
864 /**
865 * __netlink_ns_capable - General netlink message capability test
866 * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
867 * @user_ns: The user namespace of the capability to use
868 * @cap: The capability to use
869 *
870 * Test to see if the opener of the socket we received the message
871 * from had when the netlink socket was created and the sender of the
872 * message has the capability @cap in the user namespace @user_ns.
873 */
874 bool __netlink_ns_capable(const struct netlink_skb_parms *nsp,
875 struct user_namespace *user_ns, int cap)
876 {
877 return ((nsp->flags & NETLINK_SKB_DST) ||
878 file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) &&
879 ns_capable(user_ns, cap);
880 }
881 EXPORT_SYMBOL(__netlink_ns_capable);
882
883 /**
884 * netlink_ns_capable - General netlink message capability test
885 * @skb: socket buffer holding a netlink command from userspace
886 * @user_ns: The user namespace of the capability to use
887 * @cap: The capability to use
888 *
889 * Test to see if the opener of the socket we received the message
890 * from had when the netlink socket was created and the sender of the
891 * message has the capability @cap in the user namespace @user_ns.
892 */
893 bool netlink_ns_capable(const struct sk_buff *skb,
894 struct user_namespace *user_ns, int cap)
895 {
896 return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap);
897 }
898 EXPORT_SYMBOL(netlink_ns_capable);
899
900 /**
901 * netlink_capable - Netlink global message capability test
902 * @skb: socket buffer holding a netlink command from userspace
903 * @cap: The capability to use
904 *
905 * Test to see if the opener of the socket we received the message
906 * from had when the netlink socket was created and the sender of the
907 * message has the capability @cap in all user namespaces.
908 */
909 bool netlink_capable(const struct sk_buff *skb, int cap)
910 {
911 return netlink_ns_capable(skb, &init_user_ns, cap);
912 }
913 EXPORT_SYMBOL(netlink_capable);
914
915 /**
916 * netlink_net_capable - Netlink network namespace message capability test
917 * @skb: socket buffer holding a netlink command from userspace
918 * @cap: The capability to use
919 *
920 * Test to see if the opener of the socket we received the message
921 * from had when the netlink socket was created and the sender of the
922 * message has the capability @cap over the network namespace of
923 * the socket we received the message from.
924 */
925 bool netlink_net_capable(const struct sk_buff *skb, int cap)
926 {
927 return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap);
928 }
929 EXPORT_SYMBOL(netlink_net_capable);
930
931 static inline int netlink_allowed(const struct socket *sock, unsigned int flag)
932 {
933 return (nl_table[sock->sk->sk_protocol].flags & flag) ||
934 ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
935 }
936
937 static void
938 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
939 {
940 struct netlink_sock *nlk = nlk_sk(sk);
941
942 if (nlk->subscriptions && !subscriptions)
943 __sk_del_bind_node(sk);
944 else if (!nlk->subscriptions && subscriptions)
945 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
946 nlk->subscriptions = subscriptions;
947 }
948
949 static int netlink_realloc_groups(struct sock *sk)
950 {
951 struct netlink_sock *nlk = nlk_sk(sk);
952 unsigned int groups;
953 unsigned long *new_groups;
954 int err = 0;
955
956 netlink_table_grab();
957
958 groups = nl_table[sk->sk_protocol].groups;
959 if (!nl_table[sk->sk_protocol].registered) {
960 err = -ENOENT;
961 goto out_unlock;
962 }
963
964 if (nlk->ngroups >= groups)
965 goto out_unlock;
966
967 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
968 if (new_groups == NULL) {
969 err = -ENOMEM;
970 goto out_unlock;
971 }
972 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
973 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
974
975 nlk->groups = new_groups;
976 nlk->ngroups = groups;
977 out_unlock:
978 netlink_table_ungrab();
979 return err;
980 }
981
982 static void netlink_undo_bind(int group, long unsigned int groups,
983 struct sock *sk)
984 {
985 struct netlink_sock *nlk = nlk_sk(sk);
986 int undo;
987
988 if (!nlk->netlink_unbind)
989 return;
990
991 for (undo = 0; undo < group; undo++)
992 if (test_bit(undo, &groups))
993 nlk->netlink_unbind(sock_net(sk), undo + 1);
994 }
995
996 static int netlink_bind(struct socket *sock, struct sockaddr *addr,
997 int addr_len)
998 {
999 struct sock *sk = sock->sk;
1000 struct net *net = sock_net(sk);
1001 struct netlink_sock *nlk = nlk_sk(sk);
1002 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
1003 int err = 0;
1004 unsigned long groups;
1005 bool bound;
1006
1007 if (addr_len < sizeof(struct sockaddr_nl))
1008 return -EINVAL;
1009
1010 if (nladdr->nl_family != AF_NETLINK)
1011 return -EINVAL;
1012 groups = nladdr->nl_groups;
1013
1014 /* Only superuser is allowed to listen multicasts */
1015 if (groups) {
1016 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
1017 return -EPERM;
1018 err = netlink_realloc_groups(sk);
1019 if (err)
1020 return err;
1021 }
1022
1023 if (nlk->ngroups < BITS_PER_LONG)
1024 groups &= (1UL << nlk->ngroups) - 1;
1025
1026 /* Paired with WRITE_ONCE() in netlink_insert() */
1027 bound = READ_ONCE(nlk->bound);
1028 if (bound) {
1029 /* Ensure nlk->portid is up-to-date. */
1030 smp_rmb();
1031
1032 if (nladdr->nl_pid != nlk->portid)
1033 return -EINVAL;
1034 }
1035
1036 if (nlk->netlink_bind && groups) {
1037 int group;
1038
1039 /* nl_groups is a u32, so cap the maximum groups we can bind */
1040 for (group = 0; group < BITS_PER_TYPE(u32); group++) {
1041 if (!test_bit(group, &groups))
1042 continue;
1043 err = nlk->netlink_bind(net, group + 1);
1044 if (!err)
1045 continue;
1046 netlink_undo_bind(group, groups, sk);
1047 return err;
1048 }
1049 }
1050
1051 /* No need for barriers here as we return to user-space without
1052 * using any of the bound attributes.
1053 */
1054 netlink_lock_table();
1055 if (!bound) {
1056 err = nladdr->nl_pid ?
1057 netlink_insert(sk, nladdr->nl_pid) :
1058 netlink_autobind(sock);
1059 if (err) {
1060 netlink_undo_bind(BITS_PER_TYPE(u32), groups, sk);
1061 goto unlock;
1062 }
1063 }
1064
1065 if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
1066 goto unlock;
1067 netlink_unlock_table();
1068
1069 netlink_table_grab();
1070 netlink_update_subscriptions(sk, nlk->subscriptions +
1071 hweight32(groups) -
1072 hweight32(nlk->groups[0]));
1073 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups;
1074 netlink_update_listeners(sk);
1075 netlink_table_ungrab();
1076
1077 return 0;
1078
1079 unlock:
1080 netlink_unlock_table();
1081 return err;
1082 }
1083
1084 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
1085 int alen, int flags)
1086 {
1087 int err = 0;
1088 struct sock *sk = sock->sk;
1089 struct netlink_sock *nlk = nlk_sk(sk);
1090 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
1091
1092 if (alen < sizeof(addr->sa_family))
1093 return -EINVAL;
1094
1095 if (addr->sa_family == AF_UNSPEC) {
1096 /* paired with READ_ONCE() in netlink_getsockbyportid() */
1097 WRITE_ONCE(sk->sk_state, NETLINK_UNCONNECTED);
1098 /* dst_portid and dst_group can be read locklessly */
1099 WRITE_ONCE(nlk->dst_portid, 0);
1100 WRITE_ONCE(nlk->dst_group, 0);
1101 return 0;
1102 }
1103 if (addr->sa_family != AF_NETLINK)
1104 return -EINVAL;
1105
1106 if (alen < sizeof(struct sockaddr_nl))
1107 return -EINVAL;
1108
1109 if ((nladdr->nl_groups || nladdr->nl_pid) &&
1110 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
1111 return -EPERM;
1112
1113 /* No need for barriers here as we return to user-space without
1114 * using any of the bound attributes.
1115 * Paired with WRITE_ONCE() in netlink_insert().
1116 */
1117 if (!READ_ONCE(nlk->bound))
1118 err = netlink_autobind(sock);
1119
1120 if (err == 0) {
1121 /* paired with READ_ONCE() in netlink_getsockbyportid() */
1122 WRITE_ONCE(sk->sk_state, NETLINK_CONNECTED);
1123 /* dst_portid and dst_group can be read locklessly */
1124 WRITE_ONCE(nlk->dst_portid, nladdr->nl_pid);
1125 WRITE_ONCE(nlk->dst_group, ffs(nladdr->nl_groups));
1126 }
1127
1128 return err;
1129 }
1130
1131 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
1132 int peer)
1133 {
1134 struct sock *sk = sock->sk;
1135 struct netlink_sock *nlk = nlk_sk(sk);
1136 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
1137
1138 nladdr->nl_family = AF_NETLINK;
1139 nladdr->nl_pad = 0;
1140
1141 if (peer) {
1142 /* Paired with WRITE_ONCE() in netlink_connect() */
1143 nladdr->nl_pid = READ_ONCE(nlk->dst_portid);
1144 nladdr->nl_groups = netlink_group_mask(READ_ONCE(nlk->dst_group));
1145 } else {
1146 /* Paired with WRITE_ONCE() in netlink_insert() */
1147 nladdr->nl_pid = READ_ONCE(nlk->portid);
1148 netlink_lock_table();
1149 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
1150 netlink_unlock_table();
1151 }
1152 return sizeof(*nladdr);
1153 }
1154
1155 static int netlink_ioctl(struct socket *sock, unsigned int cmd,
1156 unsigned long arg)
1157 {
1158 /* try to hand this ioctl down to the NIC drivers.
1159 */
1160 return -ENOIOCTLCMD;
1161 }
1162
1163 static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
1164 {
1165 struct sock *sock;
1166 struct netlink_sock *nlk;
1167
1168 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
1169 if (!sock)
1170 return ERR_PTR(-ECONNREFUSED);
1171
1172 /* Don't bother queuing skb if kernel socket has no input function */
1173 nlk = nlk_sk(sock);
1174 /* dst_portid and sk_state can be changed in netlink_connect() */
1175 if (READ_ONCE(sock->sk_state) == NETLINK_CONNECTED &&
1176 READ_ONCE(nlk->dst_portid) != nlk_sk(ssk)->portid) {
1177 sock_put(sock);
1178 return ERR_PTR(-ECONNREFUSED);
1179 }
1180 return sock;
1181 }
1182
1183 struct sock *netlink_getsockbyfilp(struct file *filp)
1184 {
1185 struct inode *inode = file_inode(filp);
1186 struct sock *sock;
1187
1188 if (!S_ISSOCK(inode->i_mode))
1189 return ERR_PTR(-ENOTSOCK);
1190
1191 sock = SOCKET_I(inode)->sk;
1192 if (sock->sk_family != AF_NETLINK)
1193 return ERR_PTR(-EINVAL);
1194
1195 sock_hold(sock);
1196 return sock;
1197 }
1198
1199 struct sk_buff *netlink_alloc_large_skb(unsigned int size, int broadcast)
1200 {
1201 size_t head_size = SKB_HEAD_ALIGN(size);
1202 struct sk_buff *skb;
1203 void *data;
1204
1205 if (head_size <= PAGE_SIZE || broadcast)
1206 return alloc_skb(size, GFP_KERNEL);
1207
1208 data = kvmalloc(head_size, GFP_KERNEL);
1209 if (!data)
1210 return NULL;
1211
1212 skb = __build_skb(data, head_size);
1213 if (!skb)
1214 kvfree(data);
1215 else if (is_vmalloc_addr(data))
1216 skb->destructor = netlink_skb_destructor;
1217
1218 return skb;
1219 }
1220
1221 /*
1222 * Attach a skb to a netlink socket.
1223 * The caller must hold a reference to the destination socket. On error, the
1224 * reference is dropped. The skb is not send to the destination, just all
1225 * all error checks are performed and memory in the queue is reserved.
1226 * Return values:
1227 * < 0: error. skb freed, reference to sock dropped.
1228 * 0: continue
1229 * 1: repeat lookup - reference dropped while waiting for socket memory.
1230 */
1231 int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
1232 long *timeo, struct sock *ssk)
1233 {
1234 struct netlink_sock *nlk;
1235
1236 nlk = nlk_sk(sk);
1237
1238 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1239 test_bit(NETLINK_S_CONGESTED, &nlk->state))) {
1240 DECLARE_WAITQUEUE(wait, current);
1241 if (!*timeo) {
1242 if (!ssk || netlink_is_kernel(ssk))
1243 netlink_overrun(sk);
1244 sock_put(sk);
1245 kfree_skb(skb);
1246 return -EAGAIN;
1247 }
1248
1249 __set_current_state(TASK_INTERRUPTIBLE);
1250 add_wait_queue(&nlk->wait, &wait);
1251
1252 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1253 test_bit(NETLINK_S_CONGESTED, &nlk->state)) &&
1254 !sock_flag(sk, SOCK_DEAD))
1255 *timeo = schedule_timeout(*timeo);
1256
1257 __set_current_state(TASK_RUNNING);
1258 remove_wait_queue(&nlk->wait, &wait);
1259 sock_put(sk);
1260
1261 if (signal_pending(current)) {
1262 kfree_skb(skb);
1263 return sock_intr_errno(*timeo);
1264 }
1265 return 1;
1266 }
1267 netlink_skb_set_owner_r(skb, sk);
1268 return 0;
1269 }
1270
1271 static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1272 {
1273 int len = skb->len;
1274
1275 netlink_deliver_tap(sock_net(sk), skb);
1276
1277 skb_queue_tail(&sk->sk_receive_queue, skb);
1278 sk->sk_data_ready(sk);
1279 return len;
1280 }
1281
1282 int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1283 {
1284 int len = __netlink_sendskb(sk, skb);
1285
1286 sock_put(sk);
1287 return len;
1288 }
1289
1290 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
1291 {
1292 kfree_skb(skb);
1293 sock_put(sk);
1294 }
1295
1296 static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
1297 {
1298 int delta;
1299
1300 WARN_ON(skb->sk != NULL);
1301 delta = skb->end - skb->tail;
1302 if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize)
1303 return skb;
1304
1305 if (skb_shared(skb)) {
1306 struct sk_buff *nskb = skb_clone(skb, allocation);
1307 if (!nskb)
1308 return skb;
1309 consume_skb(skb);
1310 skb = nskb;
1311 }
1312
1313 pskb_expand_head(skb, 0, -delta,
1314 (allocation & ~__GFP_DIRECT_RECLAIM) |
1315 __GFP_NOWARN | __GFP_NORETRY);
1316 return skb;
1317 }
1318
1319 static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
1320 struct sock *ssk)
1321 {
1322 int ret;
1323 struct netlink_sock *nlk = nlk_sk(sk);
1324
1325 ret = -ECONNREFUSED;
1326 if (nlk->netlink_rcv != NULL) {
1327 ret = skb->len;
1328 netlink_skb_set_owner_r(skb, sk);
1329 NETLINK_CB(skb).sk = ssk;
1330 netlink_deliver_tap_kernel(sk, ssk, skb);
1331 nlk->netlink_rcv(skb);
1332 consume_skb(skb);
1333 } else {
1334 kfree_skb(skb);
1335 }
1336 sock_put(sk);
1337 return ret;
1338 }
1339
1340 int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
1341 u32 portid, int nonblock)
1342 {
1343 struct sock *sk;
1344 int err;
1345 long timeo;
1346
1347 skb = netlink_trim(skb, gfp_any());
1348
1349 timeo = sock_sndtimeo(ssk, nonblock);
1350 retry:
1351 sk = netlink_getsockbyportid(ssk, portid);
1352 if (IS_ERR(sk)) {
1353 kfree_skb(skb);
1354 return PTR_ERR(sk);
1355 }
1356 if (netlink_is_kernel(sk))
1357 return netlink_unicast_kernel(sk, skb, ssk);
1358
1359 if (sk_filter(sk, skb)) {
1360 err = skb->len;
1361 kfree_skb(skb);
1362 sock_put(sk);
1363 return err;
1364 }
1365
1366 err = netlink_attachskb(sk, skb, &timeo, ssk);
1367 if (err == 1)
1368 goto retry;
1369 if (err)
1370 return err;
1371
1372 return netlink_sendskb(sk, skb);
1373 }
1374 EXPORT_SYMBOL(netlink_unicast);
1375
1376 int netlink_has_listeners(struct sock *sk, unsigned int group)
1377 {
1378 int res = 0;
1379 struct listeners *listeners;
1380
1381 BUG_ON(!netlink_is_kernel(sk));
1382
1383 rcu_read_lock();
1384 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
1385
1386 if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
1387 res = test_bit(group - 1, listeners->masks);
1388
1389 rcu_read_unlock();
1390
1391 return res;
1392 }
1393 EXPORT_SYMBOL_GPL(netlink_has_listeners);
1394
1395 bool netlink_strict_get_check(struct sk_buff *skb)
1396 {
1397 return nlk_test_bit(STRICT_CHK, NETLINK_CB(skb).sk);
1398 }
1399 EXPORT_SYMBOL_GPL(netlink_strict_get_check);
1400
1401 static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
1402 {
1403 struct netlink_sock *nlk = nlk_sk(sk);
1404
1405 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
1406 !test_bit(NETLINK_S_CONGESTED, &nlk->state)) {
1407 netlink_skb_set_owner_r(skb, sk);
1408 __netlink_sendskb(sk, skb);
1409 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
1410 }
1411 return -1;
1412 }
1413
1414 struct netlink_broadcast_data {
1415 struct sock *exclude_sk;
1416 struct net *net;
1417 u32 portid;
1418 u32 group;
1419 int failure;
1420 int delivery_failure;
1421 int congested;
1422 int delivered;
1423 gfp_t allocation;
1424 struct sk_buff *skb, *skb2;
1425 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
1426 void *tx_data;
1427 };
1428
1429 static void do_one_broadcast(struct sock *sk,
1430 struct netlink_broadcast_data *p)
1431 {
1432 struct netlink_sock *nlk = nlk_sk(sk);
1433 int val;
1434
1435 if (p->exclude_sk == sk)
1436 return;
1437
1438 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
1439 !test_bit(p->group - 1, nlk->groups))
1440 return;
1441
1442 if (!net_eq(sock_net(sk), p->net)) {
1443 if (!nlk_test_bit(LISTEN_ALL_NSID, sk))
1444 return;
1445
1446 if (!peernet_has_id(sock_net(sk), p->net))
1447 return;
1448
1449 if (!file_ns_capable(sk->sk_socket->file, p->net->user_ns,
1450 CAP_NET_BROADCAST))
1451 return;
1452 }
1453
1454 if (p->failure) {
1455 netlink_overrun(sk);
1456 return;
1457 }
1458
1459 sock_hold(sk);
1460 if (p->skb2 == NULL) {
1461 if (skb_shared(p->skb)) {
1462 p->skb2 = skb_clone(p->skb, p->allocation);
1463 } else {
1464 p->skb2 = skb_get(p->skb);
1465 /*
1466 * skb ownership may have been set when
1467 * delivered to a previous socket.
1468 */
1469 skb_orphan(p->skb2);
1470 }
1471 }
1472 if (p->skb2 == NULL) {
1473 netlink_overrun(sk);
1474 /* Clone failed. Notify ALL listeners. */
1475 p->failure = 1;
1476 if (nlk_test_bit(BROADCAST_SEND_ERROR, sk))
1477 p->delivery_failure = 1;
1478 goto out;
1479 }
1480
1481 if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1482 kfree_skb(p->skb2);
1483 p->skb2 = NULL;
1484 goto out;
1485 }
1486
1487 if (sk_filter(sk, p->skb2)) {
1488 kfree_skb(p->skb2);
1489 p->skb2 = NULL;
1490 goto out;
1491 }
1492 NETLINK_CB(p->skb2).nsid = peernet2id(sock_net(sk), p->net);
1493 if (NETLINK_CB(p->skb2).nsid != NETNSA_NSID_NOT_ASSIGNED)
1494 NETLINK_CB(p->skb2).nsid_is_set = true;
1495 val = netlink_broadcast_deliver(sk, p->skb2);
1496 if (val < 0) {
1497 netlink_overrun(sk);
1498 if (nlk_test_bit(BROADCAST_SEND_ERROR, sk))
1499 p->delivery_failure = 1;
1500 } else {
1501 p->congested |= val;
1502 p->delivered = 1;
1503 p->skb2 = NULL;
1504 }
1505 out:
1506 sock_put(sk);
1507 }
1508
1509 int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb,
1510 u32 portid,
1511 u32 group, gfp_t allocation,
1512 netlink_filter_fn filter,
1513 void *filter_data)
1514 {
1515 struct net *net = sock_net(ssk);
1516 struct netlink_broadcast_data info;
1517 struct sock *sk;
1518
1519 skb = netlink_trim(skb, allocation);
1520
1521 info.exclude_sk = ssk;
1522 info.net = net;
1523 info.portid = portid;
1524 info.group = group;
1525 info.failure = 0;
1526 info.delivery_failure = 0;
1527 info.congested = 0;
1528 info.delivered = 0;
1529 info.allocation = allocation;
1530 info.skb = skb;
1531 info.skb2 = NULL;
1532 info.tx_filter = filter;
1533 info.tx_data = filter_data;
1534
1535 /* While we sleep in clone, do not allow to change socket list */
1536
1537 netlink_lock_table();
1538
1539 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1540 do_one_broadcast(sk, &info);
1541
1542 consume_skb(skb);
1543
1544 netlink_unlock_table();
1545
1546 if (info.delivery_failure) {
1547 kfree_skb(info.skb2);
1548 return -ENOBUFS;
1549 }
1550 consume_skb(info.skb2);
1551
1552 if (info.delivered) {
1553 if (info.congested && gfpflags_allow_blocking(allocation))
1554 yield();
1555 return 0;
1556 }
1557 return -ESRCH;
1558 }
1559 EXPORT_SYMBOL(netlink_broadcast_filtered);
1560
1561 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
1562 u32 group, gfp_t allocation)
1563 {
1564 return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
1565 NULL, NULL);
1566 }
1567 EXPORT_SYMBOL(netlink_broadcast);
1568
1569 struct netlink_set_err_data {
1570 struct sock *exclude_sk;
1571 u32 portid;
1572 u32 group;
1573 int code;
1574 };
1575
1576 static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
1577 {
1578 struct netlink_sock *nlk = nlk_sk(sk);
1579 int ret = 0;
1580
1581 if (sk == p->exclude_sk)
1582 goto out;
1583
1584 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
1585 goto out;
1586
1587 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
1588 !test_bit(p->group - 1, nlk->groups))
1589 goto out;
1590
1591 if (p->code == ENOBUFS && nlk_test_bit(RECV_NO_ENOBUFS, sk)) {
1592 ret = 1;
1593 goto out;
1594 }
1595
1596 WRITE_ONCE(sk->sk_err, p->code);
1597 sk_error_report(sk);
1598 out:
1599 return ret;
1600 }
1601
1602 /**
1603 * netlink_set_err - report error to broadcast listeners
1604 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1605 * @portid: the PORTID of a process that we want to skip (if any)
1606 * @group: the broadcast group that will notice the error
1607 * @code: error code, must be negative (as usual in kernelspace)
1608 *
1609 * This function returns the number of broadcast listeners that have set the
1610 * NETLINK_NO_ENOBUFS socket option.
1611 */
1612 int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
1613 {
1614 struct netlink_set_err_data info;
1615 unsigned long flags;
1616 struct sock *sk;
1617 int ret = 0;
1618
1619 info.exclude_sk = ssk;
1620 info.portid = portid;
1621 info.group = group;
1622 /* sk->sk_err wants a positive error value */
1623 info.code = -code;
1624
1625 read_lock_irqsave(&nl_table_lock, flags);
1626
1627 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1628 ret += do_one_set_err(sk, &info);
1629
1630 read_unlock_irqrestore(&nl_table_lock, flags);
1631 return ret;
1632 }
1633 EXPORT_SYMBOL(netlink_set_err);
1634
1635 /* must be called with netlink table grabbed */
1636 static void netlink_update_socket_mc(struct netlink_sock *nlk,
1637 unsigned int group,
1638 int is_new)
1639 {
1640 int old, new = !!is_new, subscriptions;
1641
1642 old = test_bit(group - 1, nlk->groups);
1643 subscriptions = nlk->subscriptions - old + new;
1644 __assign_bit(group - 1, nlk->groups, new);
1645 netlink_update_subscriptions(&nlk->sk, subscriptions);
1646 netlink_update_listeners(&nlk->sk);
1647 }
1648
1649 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1650 sockptr_t optval, unsigned int optlen)
1651 {
1652 struct sock *sk = sock->sk;
1653 struct netlink_sock *nlk = nlk_sk(sk);
1654 unsigned int val = 0;
1655 int nr = -1;
1656
1657 if (level != SOL_NETLINK)
1658 return -ENOPROTOOPT;
1659
1660 if (optlen >= sizeof(int) &&
1661 copy_from_sockptr(&val, optval, sizeof(val)))
1662 return -EFAULT;
1663
1664 switch (optname) {
1665 case NETLINK_PKTINFO:
1666 nr = NETLINK_F_RECV_PKTINFO;
1667 break;
1668 case NETLINK_ADD_MEMBERSHIP:
1669 case NETLINK_DROP_MEMBERSHIP: {
1670 int err;
1671
1672 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
1673 return -EPERM;
1674 err = netlink_realloc_groups(sk);
1675 if (err)
1676 return err;
1677 if (!val || val - 1 >= nlk->ngroups)
1678 return -EINVAL;
1679 if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) {
1680 err = nlk->netlink_bind(sock_net(sk), val);
1681 if (err)
1682 return err;
1683 }
1684 netlink_table_grab();
1685 netlink_update_socket_mc(nlk, val,
1686 optname == NETLINK_ADD_MEMBERSHIP);
1687 netlink_table_ungrab();
1688 if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind)
1689 nlk->netlink_unbind(sock_net(sk), val);
1690
1691 break;
1692 }
1693 case NETLINK_BROADCAST_ERROR:
1694 nr = NETLINK_F_BROADCAST_SEND_ERROR;
1695 break;
1696 case NETLINK_NO_ENOBUFS:
1697 assign_bit(NETLINK_F_RECV_NO_ENOBUFS, &nlk->flags, val);
1698 if (val) {
1699 clear_bit(NETLINK_S_CONGESTED, &nlk->state);
1700 wake_up_interruptible(&nlk->wait);
1701 }
1702 break;
1703 case NETLINK_LISTEN_ALL_NSID:
1704 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_BROADCAST))
1705 return -EPERM;
1706 nr = NETLINK_F_LISTEN_ALL_NSID;
1707 break;
1708 case NETLINK_CAP_ACK:
1709 nr = NETLINK_F_CAP_ACK;
1710 break;
1711 case NETLINK_EXT_ACK:
1712 nr = NETLINK_F_EXT_ACK;
1713 break;
1714 case NETLINK_GET_STRICT_CHK:
1715 nr = NETLINK_F_STRICT_CHK;
1716 break;
1717 default:
1718 return -ENOPROTOOPT;
1719 }
1720 if (nr >= 0)
1721 assign_bit(nr, &nlk->flags, val);
1722 return 0;
1723 }
1724
1725 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1726 char __user *optval, int __user *optlen)
1727 {
1728 struct sock *sk = sock->sk;
1729 struct netlink_sock *nlk = nlk_sk(sk);
1730 unsigned int flag;
1731 int len, val;
1732
1733 if (level != SOL_NETLINK)
1734 return -ENOPROTOOPT;
1735
1736 if (get_user(len, optlen))
1737 return -EFAULT;
1738 if (len < 0)
1739 return -EINVAL;
1740
1741 switch (optname) {
1742 case NETLINK_PKTINFO:
1743 flag = NETLINK_F_RECV_PKTINFO;
1744 break;
1745 case NETLINK_BROADCAST_ERROR:
1746 flag = NETLINK_F_BROADCAST_SEND_ERROR;
1747 break;
1748 case NETLINK_NO_ENOBUFS:
1749 flag = NETLINK_F_RECV_NO_ENOBUFS;
1750 break;
1751 case NETLINK_LIST_MEMBERSHIPS: {
1752 int pos, idx, shift, err = 0;
1753
1754 netlink_lock_table();
1755 for (pos = 0; pos * 8 < nlk->ngroups; pos += sizeof(u32)) {
1756 if (len - pos < sizeof(u32))
1757 break;
1758
1759 idx = pos / sizeof(unsigned long);
1760 shift = (pos % sizeof(unsigned long)) * 8;
1761 if (put_user((u32)(nlk->groups[idx] >> shift),
1762 (u32 __user *)(optval + pos))) {
1763 err = -EFAULT;
1764 break;
1765 }
1766 }
1767 if (put_user(ALIGN(BITS_TO_BYTES(nlk->ngroups), sizeof(u32)), optlen))
1768 err = -EFAULT;
1769 netlink_unlock_table();
1770 return err;
1771 }
1772 case NETLINK_LISTEN_ALL_NSID:
1773 flag = NETLINK_F_LISTEN_ALL_NSID;
1774 break;
1775 case NETLINK_CAP_ACK:
1776 flag = NETLINK_F_CAP_ACK;
1777 break;
1778 case NETLINK_EXT_ACK:
1779 flag = NETLINK_F_EXT_ACK;
1780 break;
1781 case NETLINK_GET_STRICT_CHK:
1782 flag = NETLINK_F_STRICT_CHK;
1783 break;
1784 default:
1785 return -ENOPROTOOPT;
1786 }
1787
1788 if (len < sizeof(int))
1789 return -EINVAL;
1790
1791 len = sizeof(int);
1792 val = test_bit(flag, &nlk->flags);
1793
1794 if (put_user(len, optlen) ||
1795 copy_to_user(optval, &val, len))
1796 return -EFAULT;
1797
1798 return 0;
1799 }
1800
1801 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1802 {
1803 struct nl_pktinfo info;
1804
1805 info.group = NETLINK_CB(skb).dst_group;
1806 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1807 }
1808
1809 static void netlink_cmsg_listen_all_nsid(struct sock *sk, struct msghdr *msg,
1810 struct sk_buff *skb)
1811 {
1812 if (!NETLINK_CB(skb).nsid_is_set)
1813 return;
1814
1815 put_cmsg(msg, SOL_NETLINK, NETLINK_LISTEN_ALL_NSID, sizeof(int),
1816 &NETLINK_CB(skb).nsid);
1817 }
1818
1819 static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
1820 {
1821 struct sock *sk = sock->sk;
1822 struct netlink_sock *nlk = nlk_sk(sk);
1823 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
1824 u32 dst_portid;
1825 u32 dst_group;
1826 struct sk_buff *skb;
1827 int err;
1828 struct scm_cookie scm;
1829 u32 netlink_skb_flags = 0;
1830
1831 if (msg->msg_flags & MSG_OOB)
1832 return -EOPNOTSUPP;
1833
1834 if (len == 0) {
1835 pr_warn_once("Zero length message leads to an empty skb\n");
1836 return -ENODATA;
1837 }
1838
1839 err = scm_send(sock, msg, &scm, true);
1840 if (err < 0)
1841 return err;
1842
1843 if (msg->msg_namelen) {
1844 err = -EINVAL;
1845 if (msg->msg_namelen < sizeof(struct sockaddr_nl))
1846 goto out;
1847 if (addr->nl_family != AF_NETLINK)
1848 goto out;
1849 dst_portid = addr->nl_pid;
1850 dst_group = ffs(addr->nl_groups);
1851 err = -EPERM;
1852 if ((dst_group || dst_portid) &&
1853 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
1854 goto out;
1855 netlink_skb_flags |= NETLINK_SKB_DST;
1856 } else {
1857 /* Paired with WRITE_ONCE() in netlink_connect() */
1858 dst_portid = READ_ONCE(nlk->dst_portid);
1859 dst_group = READ_ONCE(nlk->dst_group);
1860 }
1861
1862 /* Paired with WRITE_ONCE() in netlink_insert() */
1863 if (!READ_ONCE(nlk->bound)) {
1864 err = netlink_autobind(sock);
1865 if (err)
1866 goto out;
1867 } else {
1868 /* Ensure nlk is hashed and visible. */
1869 smp_rmb();
1870 }
1871
1872 err = -EMSGSIZE;
1873 if (len > sk->sk_sndbuf - 32)
1874 goto out;
1875 err = -ENOBUFS;
1876 skb = netlink_alloc_large_skb(len, dst_group);
1877 if (skb == NULL)
1878 goto out;
1879
1880 NETLINK_CB(skb).portid = nlk->portid;
1881 NETLINK_CB(skb).dst_group = dst_group;
1882 NETLINK_CB(skb).creds = scm.creds;
1883 NETLINK_CB(skb).flags = netlink_skb_flags;
1884
1885 err = -EFAULT;
1886 if (memcpy_from_msg(skb_put(skb, len), msg, len)) {
1887 kfree_skb(skb);
1888 goto out;
1889 }
1890
1891 err = security_netlink_send(sk, skb);
1892 if (err) {
1893 kfree_skb(skb);
1894 goto out;
1895 }
1896
1897 if (dst_group) {
1898 refcount_inc(&skb->users);
1899 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
1900 }
1901 err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags & MSG_DONTWAIT);
1902
1903 out:
1904 scm_destroy(&scm);
1905 return err;
1906 }
1907
1908 static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
1909 int flags)
1910 {
1911 struct scm_cookie scm;
1912 struct sock *sk = sock->sk;
1913 struct netlink_sock *nlk = nlk_sk(sk);
1914 size_t copied, max_recvmsg_len;
1915 struct sk_buff *skb, *data_skb;
1916 int err, ret;
1917
1918 if (flags & MSG_OOB)
1919 return -EOPNOTSUPP;
1920
1921 copied = 0;
1922
1923 skb = skb_recv_datagram(sk, flags, &err);
1924 if (skb == NULL)
1925 goto out;
1926
1927 data_skb = skb;
1928
1929 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1930 if (unlikely(skb_shinfo(skb)->frag_list)) {
1931 /*
1932 * If this skb has a frag_list, then here that means that we
1933 * will have to use the frag_list skb's data for compat tasks
1934 * and the regular skb's data for normal (non-compat) tasks.
1935 *
1936 * If we need to send the compat skb, assign it to the
1937 * 'data_skb' variable so that it will be used below for data
1938 * copying. We keep 'skb' for everything else, including
1939 * freeing both later.
1940 */
1941 if (flags & MSG_CMSG_COMPAT)
1942 data_skb = skb_shinfo(skb)->frag_list;
1943 }
1944 #endif
1945
1946 /* Record the max length of recvmsg() calls for future allocations */
1947 max_recvmsg_len = max(READ_ONCE(nlk->max_recvmsg_len), len);
1948 max_recvmsg_len = min_t(size_t, max_recvmsg_len,
1949 SKB_WITH_OVERHEAD(32768));
1950 WRITE_ONCE(nlk->max_recvmsg_len, max_recvmsg_len);
1951
1952 copied = data_skb->len;
1953 if (len < copied) {
1954 msg->msg_flags |= MSG_TRUNC;
1955 copied = len;
1956 }
1957
1958 err = skb_copy_datagram_msg(data_skb, 0, msg, copied);
1959
1960 if (msg->msg_name) {
1961 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
1962 addr->nl_family = AF_NETLINK;
1963 addr->nl_pad = 0;
1964 addr->nl_pid = NETLINK_CB(skb).portid;
1965 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1966 msg->msg_namelen = sizeof(*addr);
1967 }
1968
1969 if (nlk_test_bit(RECV_PKTINFO, sk))
1970 netlink_cmsg_recv_pktinfo(msg, skb);
1971 if (nlk_test_bit(LISTEN_ALL_NSID, sk))
1972 netlink_cmsg_listen_all_nsid(sk, msg, skb);
1973
1974 memset(&scm, 0, sizeof(scm));
1975 scm.creds = *NETLINK_CREDS(skb);
1976 if (flags & MSG_TRUNC)
1977 copied = data_skb->len;
1978
1979 skb_free_datagram(sk, skb);
1980
1981 if (READ_ONCE(nlk->cb_running) &&
1982 atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
1983 ret = netlink_dump(sk, false);
1984 if (ret) {
1985 WRITE_ONCE(sk->sk_err, -ret);
1986 sk_error_report(sk);
1987 }
1988 }
1989
1990 scm_recv(sock, msg, &scm, flags);
1991 out:
1992 netlink_rcv_wake(sk);
1993 return err ? : copied;
1994 }
1995
1996 static void netlink_data_ready(struct sock *sk)
1997 {
1998 BUG();
1999 }
2000
2001 /*
2002 * We export these functions to other modules. They provide a
2003 * complete set of kernel non-blocking support for message
2004 * queueing.
2005 */
2006
2007 struct sock *
2008 __netlink_kernel_create(struct net *net, int unit, struct module *module,
2009 struct netlink_kernel_cfg *cfg)
2010 {
2011 struct socket *sock;
2012 struct sock *sk;
2013 struct netlink_sock *nlk;
2014 struct listeners *listeners = NULL;
2015 unsigned int groups;
2016
2017 BUG_ON(!nl_table);
2018
2019 if (unit < 0 || unit >= MAX_LINKS)
2020 return NULL;
2021
2022 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
2023 return NULL;
2024
2025 if (__netlink_create(net, sock, unit, 1) < 0)
2026 goto out_sock_release_nosk;
2027
2028 sk = sock->sk;
2029
2030 if (!cfg || cfg->groups < 32)
2031 groups = 32;
2032 else
2033 groups = cfg->groups;
2034
2035 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
2036 if (!listeners)
2037 goto out_sock_release;
2038
2039 sk->sk_data_ready = netlink_data_ready;
2040 if (cfg && cfg->input)
2041 nlk_sk(sk)->netlink_rcv = cfg->input;
2042
2043 if (netlink_insert(sk, 0))
2044 goto out_sock_release;
2045
2046 nlk = nlk_sk(sk);
2047 set_bit(NETLINK_F_KERNEL_SOCKET, &nlk->flags);
2048
2049 netlink_table_grab();
2050 if (!nl_table[unit].registered) {
2051 nl_table[unit].groups = groups;
2052 rcu_assign_pointer(nl_table[unit].listeners, listeners);
2053 nl_table[unit].module = module;
2054 if (cfg) {
2055 nl_table[unit].bind = cfg->bind;
2056 nl_table[unit].unbind = cfg->unbind;
2057 nl_table[unit].release = cfg->release;
2058 nl_table[unit].flags = cfg->flags;
2059 }
2060 nl_table[unit].registered = 1;
2061 } else {
2062 kfree(listeners);
2063 nl_table[unit].registered++;
2064 }
2065 netlink_table_ungrab();
2066 return sk;
2067
2068 out_sock_release:
2069 kfree(listeners);
2070 netlink_kernel_release(sk);
2071 return NULL;
2072
2073 out_sock_release_nosk:
2074 sock_release(sock);
2075 return NULL;
2076 }
2077 EXPORT_SYMBOL(__netlink_kernel_create);
2078
2079 void
2080 netlink_kernel_release(struct sock *sk)
2081 {
2082 if (sk == NULL || sk->sk_socket == NULL)
2083 return;
2084
2085 sock_release(sk->sk_socket);
2086 }
2087 EXPORT_SYMBOL(netlink_kernel_release);
2088
2089 int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
2090 {
2091 struct listeners *new, *old;
2092 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
2093
2094 if (groups < 32)
2095 groups = 32;
2096
2097 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
2098 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
2099 if (!new)
2100 return -ENOMEM;
2101 old = nl_deref_protected(tbl->listeners);
2102 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
2103 rcu_assign_pointer(tbl->listeners, new);
2104
2105 kfree_rcu(old, rcu);
2106 }
2107 tbl->groups = groups;
2108
2109 return 0;
2110 }
2111
2112 /**
2113 * netlink_change_ngroups - change number of multicast groups
2114 *
2115 * This changes the number of multicast groups that are available
2116 * on a certain netlink family. Note that it is not possible to
2117 * change the number of groups to below 32. Also note that it does
2118 * not implicitly call netlink_clear_multicast_users() when the
2119 * number of groups is reduced.
2120 *
2121 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
2122 * @groups: The new number of groups.
2123 */
2124 int netlink_change_ngroups(struct sock *sk, unsigned int groups)
2125 {
2126 int err;
2127
2128 netlink_table_grab();
2129 err = __netlink_change_ngroups(sk, groups);
2130 netlink_table_ungrab();
2131
2132 return err;
2133 }
2134
2135 void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
2136 {
2137 struct sock *sk;
2138 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
2139
2140 sk_for_each_bound(sk, &tbl->mc_list)
2141 netlink_update_socket_mc(nlk_sk(sk), group, 0);
2142 }
2143
2144 struct nlmsghdr *
2145 __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
2146 {
2147 struct nlmsghdr *nlh;
2148 int size = nlmsg_msg_size(len);
2149
2150 nlh = skb_put(skb, NLMSG_ALIGN(size));
2151 nlh->nlmsg_type = type;
2152 nlh->nlmsg_len = size;
2153 nlh->nlmsg_flags = flags;
2154 nlh->nlmsg_pid = portid;
2155 nlh->nlmsg_seq = seq;
2156 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
2157 memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
2158 return nlh;
2159 }
2160 EXPORT_SYMBOL(__nlmsg_put);
2161
2162 static size_t
2163 netlink_ack_tlv_len(struct netlink_sock *nlk, int err,
2164 const struct netlink_ext_ack *extack)
2165 {
2166 size_t tlvlen;
2167
2168 if (!extack || !test_bit(NETLINK_F_EXT_ACK, &nlk->flags))
2169 return 0;
2170
2171 tlvlen = 0;
2172 if (extack->_msg)
2173 tlvlen += nla_total_size(strlen(extack->_msg) + 1);
2174 if (extack->cookie_len)
2175 tlvlen += nla_total_size(extack->cookie_len);
2176
2177 /* Following attributes are only reported as error (not warning) */
2178 if (!err)
2179 return tlvlen;
2180
2181 if (extack->bad_attr)
2182 tlvlen += nla_total_size(sizeof(u32));
2183 if (extack->policy)
2184 tlvlen += netlink_policy_dump_attr_size_estimate(extack->policy);
2185 if (extack->miss_type)
2186 tlvlen += nla_total_size(sizeof(u32));
2187 if (extack->miss_nest)
2188 tlvlen += nla_total_size(sizeof(u32));
2189
2190 return tlvlen;
2191 }
2192
2193 static void
2194 netlink_ack_tlv_fill(struct sk_buff *in_skb, struct sk_buff *skb,
2195 const struct nlmsghdr *nlh, int err,
2196 const struct netlink_ext_ack *extack)
2197 {
2198 if (extack->_msg)
2199 WARN_ON(nla_put_string(skb, NLMSGERR_ATTR_MSG, extack->_msg));
2200 if (extack->cookie_len)
2201 WARN_ON(nla_put(skb, NLMSGERR_ATTR_COOKIE,
2202 extack->cookie_len, extack->cookie));
2203
2204 if (!err)
2205 return;
2206
2207 if (extack->bad_attr &&
2208 !WARN_ON((u8 *)extack->bad_attr < in_skb->data ||
2209 (u8 *)extack->bad_attr >= in_skb->data + in_skb->len))
2210 WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_OFFS,
2211 (u8 *)extack->bad_attr - (const u8 *)nlh));
2212 if (extack->policy)
2213 netlink_policy_dump_write_attr(skb, extack->policy,
2214 NLMSGERR_ATTR_POLICY);
2215 if (extack->miss_type)
2216 WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_MISS_TYPE,
2217 extack->miss_type));
2218 if (extack->miss_nest &&
2219 !WARN_ON((u8 *)extack->miss_nest < in_skb->data ||
2220 (u8 *)extack->miss_nest > in_skb->data + in_skb->len))
2221 WARN_ON(nla_put_u32(skb, NLMSGERR_ATTR_MISS_NEST,
2222 (u8 *)extack->miss_nest - (const u8 *)nlh));
2223 }
2224
2225 /*
2226 * It looks a bit ugly.
2227 * It would be better to create kernel thread.
2228 */
2229
2230 static int netlink_dump_done(struct netlink_sock *nlk, struct sk_buff *skb,
2231 struct netlink_callback *cb,
2232 struct netlink_ext_ack *extack)
2233 {
2234 struct nlmsghdr *nlh;
2235 size_t extack_len;
2236
2237 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(nlk->dump_done_errno),
2238 NLM_F_MULTI | cb->answer_flags);
2239 if (WARN_ON(!nlh))
2240 return -ENOBUFS;
2241
2242 nl_dump_check_consistent(cb, nlh);
2243 memcpy(nlmsg_data(nlh), &nlk->dump_done_errno, sizeof(nlk->dump_done_errno));
2244
2245 extack_len = netlink_ack_tlv_len(nlk, nlk->dump_done_errno, extack);
2246 if (extack_len) {
2247 nlh->nlmsg_flags |= NLM_F_ACK_TLVS;
2248 if (skb_tailroom(skb) >= extack_len) {
2249 netlink_ack_tlv_fill(cb->skb, skb, cb->nlh,
2250 nlk->dump_done_errno, extack);
2251 nlmsg_end(skb, nlh);
2252 }
2253 }
2254
2255 return 0;
2256 }
2257
2258 static int netlink_dump(struct sock *sk, bool lock_taken)
2259 {
2260 struct netlink_sock *nlk = nlk_sk(sk);
2261 struct netlink_ext_ack extack = {};
2262 struct netlink_callback *cb;
2263 struct sk_buff *skb = NULL;
2264 size_t max_recvmsg_len;
2265 struct module *module;
2266 int err = -ENOBUFS;
2267 int alloc_min_size;
2268 int alloc_size;
2269
2270 if (!lock_taken)
2271 mutex_lock(&nlk->nl_cb_mutex);
2272 if (!nlk->cb_running) {
2273 err = -EINVAL;
2274 goto errout_skb;
2275 }
2276
2277 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
2278 goto errout_skb;
2279
2280 /* NLMSG_GOODSIZE is small to avoid high order allocations being
2281 * required, but it makes sense to _attempt_ a 16K bytes allocation
2282 * to reduce number of system calls on dump operations, if user
2283 * ever provided a big enough buffer.
2284 */
2285 cb = &nlk->cb;
2286 alloc_min_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
2287
2288 max_recvmsg_len = READ_ONCE(nlk->max_recvmsg_len);
2289 if (alloc_min_size < max_recvmsg_len) {
2290 alloc_size = max_recvmsg_len;
2291 skb = alloc_skb(alloc_size,
2292 (GFP_KERNEL & ~__GFP_DIRECT_RECLAIM) |
2293 __GFP_NOWARN | __GFP_NORETRY);
2294 }
2295 if (!skb) {
2296 alloc_size = alloc_min_size;
2297 skb = alloc_skb(alloc_size, GFP_KERNEL);
2298 }
2299 if (!skb)
2300 goto errout_skb;
2301
2302 /* Trim skb to allocated size. User is expected to provide buffer as
2303 * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at
2304 * netlink_recvmsg())). dump will pack as many smaller messages as
2305 * could fit within the allocated skb. skb is typically allocated
2306 * with larger space than required (could be as much as near 2x the
2307 * requested size with align to next power of 2 approach). Allowing
2308 * dump to use the excess space makes it difficult for a user to have a
2309 * reasonable static buffer based on the expected largest dump of a
2310 * single netdev. The outcome is MSG_TRUNC error.
2311 */
2312 skb_reserve(skb, skb_tailroom(skb) - alloc_size);
2313
2314 /* Make sure malicious BPF programs can not read unitialized memory
2315 * from skb->head -> skb->data
2316 */
2317 skb_reset_network_header(skb);
2318 skb_reset_mac_header(skb);
2319
2320 netlink_skb_set_owner_r(skb, sk);
2321
2322 if (nlk->dump_done_errno > 0) {
2323 cb->extack = &extack;
2324
2325 nlk->dump_done_errno = cb->dump(skb, cb);
2326
2327 /* EMSGSIZE plus something already in the skb means
2328 * that there's more to dump but current skb has filled up.
2329 * If the callback really wants to return EMSGSIZE to user space
2330 * it needs to do so again, on the next cb->dump() call,
2331 * without putting data in the skb.
2332 */
2333 if (nlk->dump_done_errno == -EMSGSIZE && skb->len)
2334 nlk->dump_done_errno = skb->len;
2335
2336 cb->extack = NULL;
2337 }
2338
2339 if (nlk->dump_done_errno > 0 ||
2340 skb_tailroom(skb) < nlmsg_total_size(sizeof(nlk->dump_done_errno))) {
2341 mutex_unlock(&nlk->nl_cb_mutex);
2342
2343 if (sk_filter(sk, skb))
2344 kfree_skb(skb);
2345 else
2346 __netlink_sendskb(sk, skb);
2347 return 0;
2348 }
2349
2350 if (netlink_dump_done(nlk, skb, cb, &extack))
2351 goto errout_skb;
2352
2353 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
2354 /* frag_list skb's data is used for compat tasks
2355 * and the regular skb's data for normal (non-compat) tasks.
2356 * See netlink_recvmsg().
2357 */
2358 if (unlikely(skb_shinfo(skb)->frag_list)) {
2359 if (netlink_dump_done(nlk, skb_shinfo(skb)->frag_list, cb, &extack))
2360 goto errout_skb;
2361 }
2362 #endif
2363
2364 if (sk_filter(sk, skb))
2365 kfree_skb(skb);
2366 else
2367 __netlink_sendskb(sk, skb);
2368
2369 if (cb->done)
2370 cb->done(cb);
2371
2372 WRITE_ONCE(nlk->cb_running, false);
2373 module = cb->module;
2374 skb = cb->skb;
2375 mutex_unlock(&nlk->nl_cb_mutex);
2376 module_put(module);
2377 consume_skb(skb);
2378 return 0;
2379
2380 errout_skb:
2381 mutex_unlock(&nlk->nl_cb_mutex);
2382 kfree_skb(skb);
2383 return err;
2384 }
2385
2386 int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
2387 const struct nlmsghdr *nlh,
2388 struct netlink_dump_control *control)
2389 {
2390 struct netlink_callback *cb;
2391 struct netlink_sock *nlk;
2392 struct sock *sk;
2393 int ret;
2394
2395 refcount_inc(&skb->users);
2396
2397 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
2398 if (sk == NULL) {
2399 ret = -ECONNREFUSED;
2400 goto error_free;
2401 }
2402
2403 nlk = nlk_sk(sk);
2404 mutex_lock(&nlk->nl_cb_mutex);
2405 /* A dump is in progress... */
2406 if (nlk->cb_running) {
2407 ret = -EBUSY;
2408 goto error_unlock;
2409 }
2410 /* add reference of module which cb->dump belongs to */
2411 if (!try_module_get(control->module)) {
2412 ret = -EPROTONOSUPPORT;
2413 goto error_unlock;
2414 }
2415
2416 cb = &nlk->cb;
2417 memset(cb, 0, sizeof(*cb));
2418 cb->dump = control->dump;
2419 cb->done = control->done;
2420 cb->nlh = nlh;
2421 cb->data = control->data;
2422 cb->module = control->module;
2423 cb->min_dump_alloc = control->min_dump_alloc;
2424 cb->flags = control->flags;
2425 cb->skb = skb;
2426
2427 cb->strict_check = nlk_test_bit(STRICT_CHK, NETLINK_CB(skb).sk);
2428
2429 if (control->start) {
2430 cb->extack = control->extack;
2431 ret = control->start(cb);
2432 cb->extack = NULL;
2433 if (ret)
2434 goto error_put;
2435 }
2436
2437 WRITE_ONCE(nlk->cb_running, true);
2438 nlk->dump_done_errno = INT_MAX;
2439
2440 ret = netlink_dump(sk, true);
2441
2442 sock_put(sk);
2443
2444 if (ret)
2445 return ret;
2446
2447 /* We successfully started a dump, by returning -EINTR we
2448 * signal not to send ACK even if it was requested.
2449 */
2450 return -EINTR;
2451
2452 error_put:
2453 module_put(control->module);
2454 error_unlock:
2455 sock_put(sk);
2456 mutex_unlock(&nlk->nl_cb_mutex);
2457 error_free:
2458 kfree_skb(skb);
2459 return ret;
2460 }
2461 EXPORT_SYMBOL(__netlink_dump_start);
2462
2463 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err,
2464 const struct netlink_ext_ack *extack)
2465 {
2466 struct sk_buff *skb;
2467 struct nlmsghdr *rep;
2468 struct nlmsgerr *errmsg;
2469 size_t payload = sizeof(*errmsg);
2470 struct netlink_sock *nlk = nlk_sk(NETLINK_CB(in_skb).sk);
2471 unsigned int flags = 0;
2472 size_t tlvlen;
2473
2474 /* Error messages get the original request appened, unless the user
2475 * requests to cap the error message, and get extra error data if
2476 * requested.
2477 */
2478 if (err && !test_bit(NETLINK_F_CAP_ACK, &nlk->flags))
2479 payload += nlmsg_len(nlh);
2480 else
2481 flags |= NLM_F_CAPPED;
2482
2483 tlvlen = netlink_ack_tlv_len(nlk, err, extack);
2484 if (tlvlen)
2485 flags |= NLM_F_ACK_TLVS;
2486
2487 skb = nlmsg_new(payload + tlvlen, GFP_KERNEL);
2488 if (!skb)
2489 goto err_skb;
2490
2491 rep = nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
2492 NLMSG_ERROR, sizeof(*errmsg), flags);
2493 if (!rep)
2494 goto err_bad_put;
2495 errmsg = nlmsg_data(rep);
2496 errmsg->error = err;
2497 errmsg->msg = *nlh;
2498
2499 if (!(flags & NLM_F_CAPPED)) {
2500 if (!nlmsg_append(skb, nlmsg_len(nlh)))
2501 goto err_bad_put;
2502
2503 memcpy(nlmsg_data(&errmsg->msg), nlmsg_data(nlh),
2504 nlmsg_len(nlh));
2505 }
2506
2507 if (tlvlen)
2508 netlink_ack_tlv_fill(in_skb, skb, nlh, err, extack);
2509
2510 nlmsg_end(skb, rep);
2511
2512 nlmsg_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid);
2513
2514 return;
2515
2516 err_bad_put:
2517 nlmsg_free(skb);
2518 err_skb:
2519 WRITE_ONCE(NETLINK_CB(in_skb).sk->sk_err, ENOBUFS);
2520 sk_error_report(NETLINK_CB(in_skb).sk);
2521 }
2522 EXPORT_SYMBOL(netlink_ack);
2523
2524 int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
2525 struct nlmsghdr *,
2526 struct netlink_ext_ack *))
2527 {
2528 struct netlink_ext_ack extack;
2529 struct nlmsghdr *nlh;
2530 int err;
2531
2532 while (skb->len >= nlmsg_total_size(0)) {
2533 int msglen;
2534
2535 memset(&extack, 0, sizeof(extack));
2536 nlh = nlmsg_hdr(skb);
2537 err = 0;
2538
2539 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
2540 return 0;
2541
2542 /* Only requests are handled by the kernel */
2543 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
2544 goto ack;
2545
2546 /* Skip control messages */
2547 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
2548 goto ack;
2549
2550 err = cb(skb, nlh, &extack);
2551 if (err == -EINTR)
2552 goto skip;
2553
2554 ack:
2555 if (nlh->nlmsg_flags & NLM_F_ACK || err)
2556 netlink_ack(skb, nlh, err, &extack);
2557
2558 skip:
2559 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
2560 if (msglen > skb->len)
2561 msglen = skb->len;
2562 skb_pull(skb, msglen);
2563 }
2564
2565 return 0;
2566 }
2567 EXPORT_SYMBOL(netlink_rcv_skb);
2568
2569 /**
2570 * nlmsg_notify - send a notification netlink message
2571 * @sk: netlink socket to use
2572 * @skb: notification message
2573 * @portid: destination netlink portid for reports or 0
2574 * @group: destination multicast group or 0
2575 * @report: 1 to report back, 0 to disable
2576 * @flags: allocation flags
2577 */
2578 int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
2579 unsigned int group, int report, gfp_t flags)
2580 {
2581 int err = 0;
2582
2583 if (group) {
2584 int exclude_portid = 0;
2585
2586 if (report) {
2587 refcount_inc(&skb->users);
2588 exclude_portid = portid;
2589 }
2590
2591 /* errors reported via destination sk->sk_err, but propagate
2592 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
2593 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
2594 if (err == -ESRCH)
2595 err = 0;
2596 }
2597
2598 if (report) {
2599 int err2;
2600
2601 err2 = nlmsg_unicast(sk, skb, portid);
2602 if (!err)
2603 err = err2;
2604 }
2605
2606 return err;
2607 }
2608 EXPORT_SYMBOL(nlmsg_notify);
2609
2610 #ifdef CONFIG_PROC_FS
2611 struct nl_seq_iter {
2612 struct seq_net_private p;
2613 struct rhashtable_iter hti;
2614 int link;
2615 };
2616
2617 static void netlink_walk_start(struct nl_seq_iter *iter)
2618 {
2619 rhashtable_walk_enter(&nl_table[iter->link].hash, &iter->hti);
2620 rhashtable_walk_start(&iter->hti);
2621 }
2622
2623 static void netlink_walk_stop(struct nl_seq_iter *iter)
2624 {
2625 rhashtable_walk_stop(&iter->hti);
2626 rhashtable_walk_exit(&iter->hti);
2627 }
2628
2629 static void *__netlink_seq_next(struct seq_file *seq)
2630 {
2631 struct nl_seq_iter *iter = seq->private;
2632 struct netlink_sock *nlk;
2633
2634 do {
2635 for (;;) {
2636 nlk = rhashtable_walk_next(&iter->hti);
2637
2638 if (IS_ERR(nlk)) {
2639 if (PTR_ERR(nlk) == -EAGAIN)
2640 continue;
2641
2642 return nlk;
2643 }
2644
2645 if (nlk)
2646 break;
2647
2648 netlink_walk_stop(iter);
2649 if (++iter->link >= MAX_LINKS)
2650 return NULL;
2651
2652 netlink_walk_start(iter);
2653 }
2654 } while (sock_net(&nlk->sk) != seq_file_net(seq));
2655
2656 return nlk;
2657 }
2658
2659 static void *netlink_seq_start(struct seq_file *seq, loff_t *posp)
2660 __acquires(RCU)
2661 {
2662 struct nl_seq_iter *iter = seq->private;
2663 void *obj = SEQ_START_TOKEN;
2664 loff_t pos;
2665
2666 iter->link = 0;
2667
2668 netlink_walk_start(iter);
2669
2670 for (pos = *posp; pos && obj && !IS_ERR(obj); pos--)
2671 obj = __netlink_seq_next(seq);
2672
2673 return obj;
2674 }
2675
2676 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2677 {
2678 ++*pos;
2679 return __netlink_seq_next(seq);
2680 }
2681
2682 static void netlink_native_seq_stop(struct seq_file *seq, void *v)
2683 {
2684 struct nl_seq_iter *iter = seq->private;
2685
2686 if (iter->link >= MAX_LINKS)
2687 return;
2688
2689 netlink_walk_stop(iter);
2690 }
2691
2692
2693 static int netlink_native_seq_show(struct seq_file *seq, void *v)
2694 {
2695 if (v == SEQ_START_TOKEN) {
2696 seq_puts(seq,
2697 "sk Eth Pid Groups "
2698 "Rmem Wmem Dump Locks Drops Inode\n");
2699 } else {
2700 struct sock *s = v;
2701 struct netlink_sock *nlk = nlk_sk(s);
2702
2703 seq_printf(seq, "%pK %-3d %-10u %08x %-8d %-8d %-5d %-8d %-8u %-8lu\n",
2704 s,
2705 s->sk_protocol,
2706 nlk->portid,
2707 nlk->groups ? (u32)nlk->groups[0] : 0,
2708 sk_rmem_alloc_get(s),
2709 sk_wmem_alloc_get(s),
2710 READ_ONCE(nlk->cb_running),
2711 refcount_read(&s->sk_refcnt),
2712 atomic_read(&s->sk_drops),
2713 sock_i_ino(s)
2714 );
2715
2716 }
2717 return 0;
2718 }
2719
2720 #ifdef CONFIG_BPF_SYSCALL
2721 struct bpf_iter__netlink {
2722 __bpf_md_ptr(struct bpf_iter_meta *, meta);
2723 __bpf_md_ptr(struct netlink_sock *, sk);
2724 };
2725
2726 DEFINE_BPF_ITER_FUNC(netlink, struct bpf_iter_meta *meta, struct netlink_sock *sk)
2727
2728 static int netlink_prog_seq_show(struct bpf_prog *prog,
2729 struct bpf_iter_meta *meta,
2730 void *v)
2731 {
2732 struct bpf_iter__netlink ctx;
2733
2734 meta->seq_num--; /* skip SEQ_START_TOKEN */
2735 ctx.meta = meta;
2736 ctx.sk = nlk_sk((struct sock *)v);
2737 return bpf_iter_run_prog(prog, &ctx);
2738 }
2739
2740 static int netlink_seq_show(struct seq_file *seq, void *v)
2741 {
2742 struct bpf_iter_meta meta;
2743 struct bpf_prog *prog;
2744
2745 meta.seq = seq;
2746 prog = bpf_iter_get_info(&meta, false);
2747 if (!prog)
2748 return netlink_native_seq_show(seq, v);
2749
2750 if (v != SEQ_START_TOKEN)
2751 return netlink_prog_seq_show(prog, &meta, v);
2752
2753 return 0;
2754 }
2755
2756 static void netlink_seq_stop(struct seq_file *seq, void *v)
2757 {
2758 struct bpf_iter_meta meta;
2759 struct bpf_prog *prog;
2760
2761 if (!v) {
2762 meta.seq = seq;
2763 prog = bpf_iter_get_info(&meta, true);
2764 if (prog)
2765 (void)netlink_prog_seq_show(prog, &meta, v);
2766 }
2767
2768 netlink_native_seq_stop(seq, v);
2769 }
2770 #else
2771 static int netlink_seq_show(struct seq_file *seq, void *v)
2772 {
2773 return netlink_native_seq_show(seq, v);
2774 }
2775
2776 static void netlink_seq_stop(struct seq_file *seq, void *v)
2777 {
2778 netlink_native_seq_stop(seq, v);
2779 }
2780 #endif
2781
2782 static const struct seq_operations netlink_seq_ops = {
2783 .start = netlink_seq_start,
2784 .next = netlink_seq_next,
2785 .stop = netlink_seq_stop,
2786 .show = netlink_seq_show,
2787 };
2788 #endif
2789
2790 int netlink_register_notifier(struct notifier_block *nb)
2791 {
2792 return blocking_notifier_chain_register(&netlink_chain, nb);
2793 }
2794 EXPORT_SYMBOL(netlink_register_notifier);
2795
2796 int netlink_unregister_notifier(struct notifier_block *nb)
2797 {
2798 return blocking_notifier_chain_unregister(&netlink_chain, nb);
2799 }
2800 EXPORT_SYMBOL(netlink_unregister_notifier);
2801
2802 static const struct proto_ops netlink_ops = {
2803 .family = PF_NETLINK,
2804 .owner = THIS_MODULE,
2805 .release = netlink_release,
2806 .bind = netlink_bind,
2807 .connect = netlink_connect,
2808 .socketpair = sock_no_socketpair,
2809 .accept = sock_no_accept,
2810 .getname = netlink_getname,
2811 .poll = datagram_poll,
2812 .ioctl = netlink_ioctl,
2813 .listen = sock_no_listen,
2814 .shutdown = sock_no_shutdown,
2815 .setsockopt = netlink_setsockopt,
2816 .getsockopt = netlink_getsockopt,
2817 .sendmsg = netlink_sendmsg,
2818 .recvmsg = netlink_recvmsg,
2819 .mmap = sock_no_mmap,
2820 };
2821
2822 static const struct net_proto_family netlink_family_ops = {
2823 .family = PF_NETLINK,
2824 .create = netlink_create,
2825 .owner = THIS_MODULE, /* for consistency 8) */
2826 };
2827
2828 static int __net_init netlink_net_init(struct net *net)
2829 {
2830 #ifdef CONFIG_PROC_FS
2831 if (!proc_create_net("netlink", 0, net->proc_net, &netlink_seq_ops,
2832 sizeof(struct nl_seq_iter)))
2833 return -ENOMEM;
2834 #endif
2835 return 0;
2836 }
2837
2838 static void __net_exit netlink_net_exit(struct net *net)
2839 {
2840 #ifdef CONFIG_PROC_FS
2841 remove_proc_entry("netlink", net->proc_net);
2842 #endif
2843 }
2844
2845 static void __init netlink_add_usersock_entry(void)
2846 {
2847 struct listeners *listeners;
2848 int groups = 32;
2849
2850 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
2851 if (!listeners)
2852 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
2853
2854 netlink_table_grab();
2855
2856 nl_table[NETLINK_USERSOCK].groups = groups;
2857 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
2858 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
2859 nl_table[NETLINK_USERSOCK].registered = 1;
2860 nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
2861
2862 netlink_table_ungrab();
2863 }
2864
2865 static struct pernet_operations __net_initdata netlink_net_ops = {
2866 .init = netlink_net_init,
2867 .exit = netlink_net_exit,
2868 };
2869
2870 static inline u32 netlink_hash(const void *data, u32 len, u32 seed)
2871 {
2872 const struct netlink_sock *nlk = data;
2873 struct netlink_compare_arg arg;
2874
2875 netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid);
2876 return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed);
2877 }
2878
2879 static const struct rhashtable_params netlink_rhashtable_params = {
2880 .head_offset = offsetof(struct netlink_sock, node),
2881 .key_len = netlink_compare_arg_len,
2882 .obj_hashfn = netlink_hash,
2883 .obj_cmpfn = netlink_compare,
2884 .automatic_shrinking = true,
2885 };
2886
2887 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
2888 BTF_ID_LIST(btf_netlink_sock_id)
2889 BTF_ID(struct, netlink_sock)
2890
2891 static const struct bpf_iter_seq_info netlink_seq_info = {
2892 .seq_ops = &netlink_seq_ops,
2893 .init_seq_private = bpf_iter_init_seq_net,
2894 .fini_seq_private = bpf_iter_fini_seq_net,
2895 .seq_priv_size = sizeof(struct nl_seq_iter),
2896 };
2897
2898 static struct bpf_iter_reg netlink_reg_info = {
2899 .target = "netlink",
2900 .ctx_arg_info_size = 1,
2901 .ctx_arg_info = {
2902 { offsetof(struct bpf_iter__netlink, sk),
2903 PTR_TO_BTF_ID_OR_NULL },
2904 },
2905 .seq_info = &netlink_seq_info,
2906 };
2907
2908 static int __init bpf_iter_register(void)
2909 {
2910 netlink_reg_info.ctx_arg_info[0].btf_id = *btf_netlink_sock_id;
2911 return bpf_iter_reg_target(&netlink_reg_info);
2912 }
2913 #endif
2914
2915 static int __init netlink_proto_init(void)
2916 {
2917 int i;
2918 int err = proto_register(&netlink_proto, 0);
2919
2920 if (err != 0)
2921 goto out;
2922
2923 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
2924 err = bpf_iter_register();
2925 if (err)
2926 goto out;
2927 #endif
2928
2929 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof_field(struct sk_buff, cb));
2930
2931 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
2932 if (!nl_table)
2933 goto panic;
2934
2935 for (i = 0; i < MAX_LINKS; i++) {
2936 if (rhashtable_init(&nl_table[i].hash,
2937 &netlink_rhashtable_params) < 0) {
2938 while (--i > 0)
2939 rhashtable_destroy(&nl_table[i].hash);
2940 kfree(nl_table);
2941 goto panic;
2942 }
2943 }
2944
2945 netlink_add_usersock_entry();
2946
2947 sock_register(&netlink_family_ops);
2948 register_pernet_subsys(&netlink_net_ops);
2949 register_pernet_subsys(&netlink_tap_net_ops);
2950 /* The netlink device handler may be needed early. */
2951 rtnetlink_init();
2952 out:
2953 return err;
2954 panic:
2955 panic("netlink_init: Cannot allocate nl_table\n");
2956 }
2957
2958 core_initcall(netlink_proto_init);
2959
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.