blob: 4da797fa5ec57641a0fa70b5dd1e6e517884fad4 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NETLINK Kernel-user communication protocol.
3 *
Alan Cox113aa832008-10-13 19:01:08 -07004 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +090011 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
13 * added netlink_proto_exit
14 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
15 * use nlk_sk, as sk->protinfo is on a diet 8)
Harald Welte4fdb3bb2005-08-09 19:40:55 -070016 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
17 * - inc module use count of module that owns
18 * the kernel socket in case userspace opens
19 * socket of same protocol
20 * - remove all module support, since netlink is
21 * mandatory if CONFIG_NET=y these days
Linus Torvalds1da177e2005-04-16 15:20:36 -070022 */
23
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/module.h>
25
Randy Dunlap4fc268d2006-01-11 12:17:47 -080026#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/kernel.h>
28#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/signal.h>
30#include <linux/sched.h>
31#include <linux/errno.h>
32#include <linux/string.h>
33#include <linux/stat.h>
34#include <linux/socket.h>
35#include <linux/un.h>
36#include <linux/fcntl.h>
37#include <linux/termios.h>
38#include <linux/sockios.h>
39#include <linux/net.h>
40#include <linux/fs.h>
41#include <linux/slab.h>
42#include <asm/uaccess.h>
43#include <linux/skbuff.h>
44#include <linux/netdevice.h>
45#include <linux/rtnetlink.h>
46#include <linux/proc_fs.h>
47#include <linux/seq_file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/notifier.h>
49#include <linux/security.h>
50#include <linux/jhash.h>
51#include <linux/jiffies.h>
52#include <linux/random.h>
53#include <linux/bitops.h>
54#include <linux/mm.h>
55#include <linux/types.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010056#include <linux/audit.h>
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070057#include <linux/mutex.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010058
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020059#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#include <net/sock.h>
61#include <net/scm.h>
Thomas Graf82ace472005-11-10 02:25:53 +010062#include <net/netlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Patrick McHardyf7fa9b12005-08-15 12:29:13 -070064#define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
Johannes Bergb4ff4f02007-07-18 15:46:06 -070065#define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67struct netlink_sock {
68 /* struct sock has to be the first member of netlink_sock */
69 struct sock sk;
Eric W. Biederman15e47302012-09-07 20:12:54 +000070 u32 portid;
71 u32 dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -070072 u32 dst_group;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -070073 u32 flags;
74 u32 subscriptions;
75 u32 ngroups;
76 unsigned long *groups;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 unsigned long state;
78 wait_queue_head_t wait;
79 struct netlink_callback *cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070080 struct mutex *cb_mutex;
81 struct mutex cb_def_mutex;
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -070082 void (*netlink_rcv)(struct sk_buff *skb);
Pablo Neira Ayuso03292742012-06-29 06:15:22 +000083 void (*netlink_bind)(int group);
Patrick McHardy77247bb2005-08-14 19:27:13 -070084 struct module *module;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085};
86
Eric Dumazet5c398dc2010-10-24 04:27:10 +000087struct listeners {
88 struct rcu_head rcu;
89 unsigned long masks[0];
Johannes Berg6c04bb12009-07-10 09:51:32 +000090};
91
Patrick McHardy77247bb2005-08-14 19:27:13 -070092#define NETLINK_KERNEL_SOCKET 0x1
Patrick McHardy9a4595b2005-08-15 12:32:15 -070093#define NETLINK_RECV_PKTINFO 0x2
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +000094#define NETLINK_BROADCAST_SEND_ERROR 0x4
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -070095#define NETLINK_RECV_NO_ENOBUFS 0x8
Patrick McHardy77247bb2005-08-14 19:27:13 -070096
Linus Torvalds1da177e2005-04-16 15:20:36 -070097static inline struct netlink_sock *nlk_sk(struct sock *sk)
98{
Denis Cheng32b21e02007-08-28 15:41:11 -070099 return container_of(sk, struct netlink_sock, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100}
101
David S. Miller035c4c12011-12-23 17:33:03 -0500102static inline int netlink_is_kernel(struct sock *sk)
Denis V. Lunevaed81562007-10-10 21:14:32 -0700103{
104 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
105}
106
Eric W. Biederman15e47302012-09-07 20:12:54 +0000107struct nl_portid_hash {
Eric Dumazet658cb352012-04-22 21:30:21 +0000108 struct hlist_head *table;
109 unsigned long rehash_time;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Eric Dumazet658cb352012-04-22 21:30:21 +0000111 unsigned int mask;
112 unsigned int shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Eric Dumazet658cb352012-04-22 21:30:21 +0000114 unsigned int entries;
115 unsigned int max_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Eric Dumazet658cb352012-04-22 21:30:21 +0000117 u32 rnd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118};
119
120struct netlink_table {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000121 struct nl_portid_hash hash;
Eric Dumazet658cb352012-04-22 21:30:21 +0000122 struct hlist_head mc_list;
123 struct listeners __rcu *listeners;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +0000124 unsigned int flags;
Eric Dumazet658cb352012-04-22 21:30:21 +0000125 unsigned int groups;
126 struct mutex *cb_mutex;
127 struct module *module;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +0000128 void (*bind)(int group);
Eric Dumazet658cb352012-04-22 21:30:21 +0000129 int registered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130};
131
132static struct netlink_table *nl_table;
133
134static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
135
136static int netlink_dump(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
138static DEFINE_RWLOCK(nl_table_lock);
139static atomic_t nl_table_users = ATOMIC_INIT(0);
140
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000141#define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
142
Alan Sterne041c682006-03-27 01:16:30 -0800143static ATOMIC_NOTIFIER_HEAD(netlink_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000145static inline u32 netlink_group_mask(u32 group)
Patrick McHardyd629b832005-08-14 19:27:50 -0700146{
147 return group ? 1 << (group - 1) : 0;
148}
149
Eric W. Biederman15e47302012-09-07 20:12:54 +0000150static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151{
Eric W. Biederman15e47302012-09-07 20:12:54 +0000152 return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153}
154
Eric Dumazet658cb352012-04-22 21:30:21 +0000155static void netlink_destroy_callback(struct netlink_callback *cb)
156{
157 kfree_skb(cb->skb);
158 kfree(cb);
159}
160
Eric Dumazetbfb253c2012-04-22 21:30:29 +0000161static void netlink_consume_callback(struct netlink_callback *cb)
162{
163 consume_skb(cb->skb);
164 kfree(cb);
165}
166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167static void netlink_sock_destruct(struct sock *sk)
168{
Herbert Xu3f660d62007-05-03 03:17:14 -0700169 struct netlink_sock *nlk = nlk_sk(sk);
170
Herbert Xu3f660d62007-05-03 03:17:14 -0700171 if (nlk->cb) {
172 if (nlk->cb->done)
173 nlk->cb->done(nlk->cb);
Gao feng6dc878a2012-10-04 20:15:48 +0000174
175 module_put(nlk->cb->module);
Herbert Xu3f660d62007-05-03 03:17:14 -0700176 netlink_destroy_callback(nlk->cb);
177 }
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 skb_queue_purge(&sk->sk_receive_queue);
180
181 if (!sock_flag(sk, SOCK_DEAD)) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800182 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 return;
184 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700185
186 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
187 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
188 WARN_ON(nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189}
190
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800191/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
192 * SMP. Look, when several writers sleep and reader wakes them up, all but one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
194 * this, _but_ remember, it adds useless work on UP machines.
195 */
196
Johannes Bergd136f1b2009-09-12 03:03:15 +0000197void netlink_table_grab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800198 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199{
Johannes Bergd136f1b2009-09-12 03:03:15 +0000200 might_sleep();
201
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700202 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203
204 if (atomic_read(&nl_table_users)) {
205 DECLARE_WAITQUEUE(wait, current);
206
207 add_wait_queue_exclusive(&nl_table_wait, &wait);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800208 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 set_current_state(TASK_UNINTERRUPTIBLE);
210 if (atomic_read(&nl_table_users) == 0)
211 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700212 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700214 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 }
216
217 __set_current_state(TASK_RUNNING);
218 remove_wait_queue(&nl_table_wait, &wait);
219 }
220}
221
Johannes Bergd136f1b2009-09-12 03:03:15 +0000222void netlink_table_ungrab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800223 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700225 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 wake_up(&nl_table_wait);
227}
228
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800229static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230netlink_lock_table(void)
231{
232 /* read_lock() synchronizes us to netlink_table_grab */
233
234 read_lock(&nl_table_lock);
235 atomic_inc(&nl_table_users);
236 read_unlock(&nl_table_lock);
237}
238
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800239static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240netlink_unlock_table(void)
241{
242 if (atomic_dec_and_test(&nl_table_users))
243 wake_up(&nl_table_wait);
244}
245
Eric W. Biederman15e47302012-09-07 20:12:54 +0000246static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247{
Eric W. Biederman15e47302012-09-07 20:12:54 +0000248 struct nl_portid_hash *hash = &nl_table[protocol].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 struct hlist_head *head;
250 struct sock *sk;
251 struct hlist_node *node;
252
253 read_lock(&nl_table_lock);
Eric W. Biederman15e47302012-09-07 20:12:54 +0000254 head = nl_portid_hashfn(hash, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 sk_for_each(sk, node, head) {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000256 if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->portid == portid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 sock_hold(sk);
258 goto found;
259 }
260 }
261 sk = NULL;
262found:
263 read_unlock(&nl_table_lock);
264 return sk;
265}
266
Eric W. Biederman15e47302012-09-07 20:12:54 +0000267static struct hlist_head *nl_portid_hash_zalloc(size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
269 if (size <= PAGE_SIZE)
Eric Dumazetea729122007-12-11 02:09:47 -0800270 return kzalloc(size, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 else
272 return (struct hlist_head *)
Eric Dumazetea729122007-12-11 02:09:47 -0800273 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
274 get_order(size));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275}
276
Eric W. Biederman15e47302012-09-07 20:12:54 +0000277static void nl_portid_hash_free(struct hlist_head *table, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
279 if (size <= PAGE_SIZE)
280 kfree(table);
281 else
282 free_pages((unsigned long)table, get_order(size));
283}
284
Eric W. Biederman15e47302012-09-07 20:12:54 +0000285static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286{
287 unsigned int omask, mask, shift;
288 size_t osize, size;
289 struct hlist_head *otable, *table;
290 int i;
291
292 omask = mask = hash->mask;
293 osize = size = (mask + 1) * sizeof(*table);
294 shift = hash->shift;
295
296 if (grow) {
297 if (++shift > hash->max_shift)
298 return 0;
299 mask = mask * 2 + 1;
300 size *= 2;
301 }
302
Eric W. Biederman15e47302012-09-07 20:12:54 +0000303 table = nl_portid_hash_zalloc(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 if (!table)
305 return 0;
306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 otable = hash->table;
308 hash->table = table;
309 hash->mask = mask;
310 hash->shift = shift;
311 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
312
313 for (i = 0; i <= omask; i++) {
314 struct sock *sk;
315 struct hlist_node *node, *tmp;
316
317 sk_for_each_safe(sk, node, tmp, &otable[i])
Eric W. Biederman15e47302012-09-07 20:12:54 +0000318 __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 }
320
Eric W. Biederman15e47302012-09-07 20:12:54 +0000321 nl_portid_hash_free(otable, osize);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 hash->rehash_time = jiffies + 10 * 60 * HZ;
323 return 1;
324}
325
Eric W. Biederman15e47302012-09-07 20:12:54 +0000326static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327{
328 int avg = hash->entries >> hash->shift;
329
Eric W. Biederman15e47302012-09-07 20:12:54 +0000330 if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 return 1;
332
333 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000334 nl_portid_hash_rehash(hash, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 return 1;
336 }
337
338 return 0;
339}
340
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800341static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
Patrick McHardy4277a082006-03-20 18:52:01 -0800343static void
344netlink_update_listeners(struct sock *sk)
345{
346 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
347 struct hlist_node *node;
348 unsigned long mask;
349 unsigned int i;
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000350 struct listeners *listeners;
351
352 listeners = nl_deref_protected(tbl->listeners);
353 if (!listeners)
354 return;
Patrick McHardy4277a082006-03-20 18:52:01 -0800355
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700356 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
Patrick McHardy4277a082006-03-20 18:52:01 -0800357 mask = 0;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700358 sk_for_each_bound(sk, node, &tbl->mc_list) {
359 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
360 mask |= nlk_sk(sk)->groups[i];
361 }
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000362 listeners->masks[i] = mask;
Patrick McHardy4277a082006-03-20 18:52:01 -0800363 }
364 /* this function is only called with the netlink table "grabbed", which
365 * makes sure updates are visible before bind or setsockopt return. */
366}
367
Eric W. Biederman15e47302012-09-07 20:12:54 +0000368static int netlink_insert(struct sock *sk, struct net *net, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369{
Eric W. Biederman15e47302012-09-07 20:12:54 +0000370 struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 struct hlist_head *head;
372 int err = -EADDRINUSE;
373 struct sock *osk;
374 struct hlist_node *node;
375 int len;
376
377 netlink_table_grab();
Eric W. Biederman15e47302012-09-07 20:12:54 +0000378 head = nl_portid_hashfn(hash, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 len = 0;
380 sk_for_each(osk, node, head) {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000381 if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->portid == portid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 break;
383 len++;
384 }
385 if (node)
386 goto err;
387
388 err = -EBUSY;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000389 if (nlk_sk(sk)->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 goto err;
391
392 err = -ENOMEM;
393 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
394 goto err;
395
Eric W. Biederman15e47302012-09-07 20:12:54 +0000396 if (len && nl_portid_hash_dilute(hash, len))
397 head = nl_portid_hashfn(hash, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 hash->entries++;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000399 nlk_sk(sk)->portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 sk_add_node(sk, head);
401 err = 0;
402
403err:
404 netlink_table_ungrab();
405 return err;
406}
407
408static void netlink_remove(struct sock *sk)
409{
410 netlink_table_grab();
David S. Millerd470e3b2005-06-26 15:31:51 -0700411 if (sk_del_node_init(sk))
412 nl_table[sk->sk_protocol].hash.entries--;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700413 if (nlk_sk(sk)->subscriptions)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 __sk_del_bind_node(sk);
415 netlink_table_ungrab();
416}
417
418static struct proto netlink_proto = {
419 .name = "NETLINK",
420 .owner = THIS_MODULE,
421 .obj_size = sizeof(struct netlink_sock),
422};
423
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700424static int __netlink_create(struct net *net, struct socket *sock,
425 struct mutex *cb_mutex, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426{
427 struct sock *sk;
428 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700429
430 sock->ops = &netlink_ops;
431
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700432 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
Patrick McHardyab33a172005-08-14 19:31:36 -0700433 if (!sk)
434 return -ENOMEM;
435
436 sock_init_data(sock, sk);
437
438 nlk = nlk_sk(sk);
Eric Dumazet658cb352012-04-22 21:30:21 +0000439 if (cb_mutex) {
Patrick McHardyffa4d722007-04-25 14:01:17 -0700440 nlk->cb_mutex = cb_mutex;
Eric Dumazet658cb352012-04-22 21:30:21 +0000441 } else {
Patrick McHardyffa4d722007-04-25 14:01:17 -0700442 nlk->cb_mutex = &nlk->cb_def_mutex;
443 mutex_init(nlk->cb_mutex);
444 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700445 init_waitqueue_head(&nlk->wait);
446
447 sk->sk_destruct = netlink_sock_destruct;
448 sk->sk_protocol = protocol;
449 return 0;
450}
451
Eric Paris3f378b62009-11-05 22:18:14 -0800452static int netlink_create(struct net *net, struct socket *sock, int protocol,
453 int kern)
Patrick McHardyab33a172005-08-14 19:31:36 -0700454{
455 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700456 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700457 struct netlink_sock *nlk;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +0000458 void (*bind)(int group);
Patrick McHardyab33a172005-08-14 19:31:36 -0700459 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
461 sock->state = SS_UNCONNECTED;
462
463 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
464 return -ESOCKTNOSUPPORT;
465
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800466 if (protocol < 0 || protocol >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 return -EPROTONOSUPPORT;
468
Patrick McHardy77247bb2005-08-14 19:27:13 -0700469 netlink_lock_table();
Johannes Berg95a5afc2008-10-16 15:24:51 -0700470#ifdef CONFIG_MODULES
Patrick McHardyab33a172005-08-14 19:31:36 -0700471 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -0700472 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700473 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -0700474 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700475 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700476#endif
477 if (nl_table[protocol].registered &&
478 try_module_get(nl_table[protocol].module))
479 module = nl_table[protocol].module;
Alexey Dobriyan974c37e2010-01-30 10:05:05 +0000480 else
481 err = -EPROTONOSUPPORT;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700482 cb_mutex = nl_table[protocol].cb_mutex;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +0000483 bind = nl_table[protocol].bind;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700484 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700485
Alexey Dobriyan974c37e2010-01-30 10:05:05 +0000486 if (err < 0)
487 goto out;
488
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800489 err = __netlink_create(net, sock, cb_mutex, protocol);
490 if (err < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -0700491 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
David S. Miller6f756a82008-11-23 17:34:03 -0800493 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -0800494 sock_prot_inuse_add(net, &netlink_proto, 1);
David S. Miller6f756a82008-11-23 17:34:03 -0800495 local_bh_enable();
496
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700497 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700498 nlk->module = module;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +0000499 nlk->netlink_bind = bind;
Patrick McHardyab33a172005-08-14 19:31:36 -0700500out:
501 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
Patrick McHardyab33a172005-08-14 19:31:36 -0700503out_module:
504 module_put(module);
505 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506}
507
508static int netlink_release(struct socket *sock)
509{
510 struct sock *sk = sock->sk;
511 struct netlink_sock *nlk;
512
513 if (!sk)
514 return 0;
515
516 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -0700517 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 nlk = nlk_sk(sk);
519
Herbert Xu3f660d62007-05-03 03:17:14 -0700520 /*
521 * OK. Socket is unlinked, any packets that arrive now
522 * will be purged.
523 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 sock->sk = NULL;
526 wake_up_interruptible_all(&nlk->wait);
527
528 skb_queue_purge(&sk->sk_write_queue);
529
Eric W. Biederman15e47302012-09-07 20:12:54 +0000530 if (nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 struct netlink_notify n = {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900532 .net = sock_net(sk),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 .protocol = sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +0000534 .portid = nlk->portid,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 };
Alan Sterne041c682006-03-27 01:16:30 -0800536 atomic_notifier_call_chain(&netlink_chain,
537 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900538 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700539
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -0800540 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700541
Patrick McHardy4277a082006-03-20 18:52:01 -0800542 netlink_table_grab();
Denis V. Lunevaed81562007-10-10 21:14:32 -0700543 if (netlink_is_kernel(sk)) {
Denis V. Lunev869e58f2008-01-18 23:53:31 -0800544 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
545 if (--nl_table[sk->sk_protocol].registered == 0) {
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000546 struct listeners *old;
547
548 old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
549 RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
550 kfree_rcu(old, rcu);
Denis V. Lunev869e58f2008-01-18 23:53:31 -0800551 nl_table[sk->sk_protocol].module = NULL;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +0000552 nl_table[sk->sk_protocol].bind = NULL;
553 nl_table[sk->sk_protocol].flags = 0;
Denis V. Lunev869e58f2008-01-18 23:53:31 -0800554 nl_table[sk->sk_protocol].registered = 0;
555 }
Eric Dumazet658cb352012-04-22 21:30:21 +0000556 } else if (nlk->subscriptions) {
Patrick McHardy4277a082006-03-20 18:52:01 -0800557 netlink_update_listeners(sk);
Eric Dumazet658cb352012-04-22 21:30:21 +0000558 }
Patrick McHardy4277a082006-03-20 18:52:01 -0800559 netlink_table_ungrab();
Patrick McHardy77247bb2005-08-14 19:27:13 -0700560
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700561 kfree(nlk->groups);
562 nlk->groups = NULL;
563
Eric Dumazet37558102008-11-24 14:05:22 -0800564 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -0800565 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
Eric Dumazet37558102008-11-24 14:05:22 -0800566 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 sock_put(sk);
568 return 0;
569}
570
571static int netlink_autobind(struct socket *sock)
572{
573 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900574 struct net *net = sock_net(sk);
Eric W. Biederman15e47302012-09-07 20:12:54 +0000575 struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 struct hlist_head *head;
577 struct sock *osk;
578 struct hlist_node *node;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000579 s32 portid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 int err;
581 static s32 rover = -4097;
582
583retry:
584 cond_resched();
585 netlink_table_grab();
Eric W. Biederman15e47302012-09-07 20:12:54 +0000586 head = nl_portid_hashfn(hash, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 sk_for_each(osk, node, head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900588 if (!net_eq(sock_net(osk), net))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200589 continue;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000590 if (nlk_sk(osk)->portid == portid) {
591 /* Bind collision, search negative portid values. */
592 portid = rover--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 if (rover > -4097)
594 rover = -4097;
595 netlink_table_ungrab();
596 goto retry;
597 }
598 }
599 netlink_table_ungrab();
600
Eric W. Biederman15e47302012-09-07 20:12:54 +0000601 err = netlink_insert(sk, net, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 if (err == -EADDRINUSE)
603 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -0700604
605 /* If 2 threads race to autobind, that is fine. */
606 if (err == -EBUSY)
607 err = 0;
608
609 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610}
611
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000612static inline int netlink_capable(const struct socket *sock, unsigned int flag)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900613{
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +0000614 return (nl_table[sock->sk->sk_protocol].flags & flag) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 capable(CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900616}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700618static void
619netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
620{
621 struct netlink_sock *nlk = nlk_sk(sk);
622
623 if (nlk->subscriptions && !subscriptions)
624 __sk_del_bind_node(sk);
625 else if (!nlk->subscriptions && subscriptions)
626 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
627 nlk->subscriptions = subscriptions;
628}
629
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700630static int netlink_realloc_groups(struct sock *sk)
Patrick McHardy513c2502005-09-06 15:43:59 -0700631{
632 struct netlink_sock *nlk = nlk_sk(sk);
633 unsigned int groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700634 unsigned long *new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700635 int err = 0;
636
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700637 netlink_table_grab();
638
Patrick McHardy513c2502005-09-06 15:43:59 -0700639 groups = nl_table[sk->sk_protocol].groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700640 if (!nl_table[sk->sk_protocol].registered) {
Patrick McHardy513c2502005-09-06 15:43:59 -0700641 err = -ENOENT;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700642 goto out_unlock;
643 }
Patrick McHardy513c2502005-09-06 15:43:59 -0700644
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700645 if (nlk->ngroups >= groups)
646 goto out_unlock;
Patrick McHardy513c2502005-09-06 15:43:59 -0700647
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700648 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
649 if (new_groups == NULL) {
650 err = -ENOMEM;
651 goto out_unlock;
652 }
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800653 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700654 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
655
656 nlk->groups = new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700657 nlk->ngroups = groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700658 out_unlock:
659 netlink_table_ungrab();
660 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700661}
662
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800663static int netlink_bind(struct socket *sock, struct sockaddr *addr,
664 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665{
666 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900667 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 struct netlink_sock *nlk = nlk_sk(sk);
669 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
670 int err;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900671
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 if (nladdr->nl_family != AF_NETLINK)
673 return -EINVAL;
674
675 /* Only superuser is allowed to listen multicasts */
Patrick McHardy513c2502005-09-06 15:43:59 -0700676 if (nladdr->nl_groups) {
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +0000677 if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy513c2502005-09-06 15:43:59 -0700678 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700679 err = netlink_realloc_groups(sk);
680 if (err)
681 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700682 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
Eric W. Biederman15e47302012-09-07 20:12:54 +0000684 if (nlk->portid) {
685 if (nladdr->nl_pid != nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 return -EINVAL;
687 } else {
688 err = nladdr->nl_pid ?
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200689 netlink_insert(sk, net, nladdr->nl_pid) :
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 netlink_autobind(sock);
691 if (err)
692 return err;
693 }
694
Patrick McHardy513c2502005-09-06 15:43:59 -0700695 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 return 0;
697
698 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700699 netlink_update_subscriptions(sk, nlk->subscriptions +
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900700 hweight32(nladdr->nl_groups) -
701 hweight32(nlk->groups[0]));
702 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
Patrick McHardy4277a082006-03-20 18:52:01 -0800703 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 netlink_table_ungrab();
705
Pablo Neira Ayuso03292742012-06-29 06:15:22 +0000706 if (nlk->netlink_bind && nlk->groups[0]) {
707 int i;
708
709 for (i=0; i<nlk->ngroups; i++) {
710 if (test_bit(i, nlk->groups))
711 nlk->netlink_bind(i);
712 }
713 }
714
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 return 0;
716}
717
718static int netlink_connect(struct socket *sock, struct sockaddr *addr,
719 int alen, int flags)
720{
721 int err = 0;
722 struct sock *sk = sock->sk;
723 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800724 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
Changli Gao6503d962010-03-31 22:58:26 +0000726 if (alen < sizeof(addr->sa_family))
727 return -EINVAL;
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 if (addr->sa_family == AF_UNSPEC) {
730 sk->sk_state = NETLINK_UNCONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000731 nlk->dst_portid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -0700732 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 return 0;
734 }
735 if (addr->sa_family != AF_NETLINK)
736 return -EINVAL;
737
738 /* Only superuser is allowed to send multicasts */
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +0000739 if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return -EPERM;
741
Eric W. Biederman15e47302012-09-07 20:12:54 +0000742 if (!nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 err = netlink_autobind(sock);
744
745 if (err == 0) {
746 sk->sk_state = NETLINK_CONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +0000747 nlk->dst_portid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700748 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 }
750
751 return err;
752}
753
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800754static int netlink_getname(struct socket *sock, struct sockaddr *addr,
755 int *addr_len, int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
757 struct sock *sk = sock->sk;
758 struct netlink_sock *nlk = nlk_sk(sk);
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +0000759 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900760
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 nladdr->nl_family = AF_NETLINK;
762 nladdr->nl_pad = 0;
763 *addr_len = sizeof(*nladdr);
764
765 if (peer) {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000766 nladdr->nl_pid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700767 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +0000769 nladdr->nl_pid = nlk->portid;
Patrick McHardy513c2502005-09-06 15:43:59 -0700770 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 }
772 return 0;
773}
774
775static void netlink_overrun(struct sock *sk)
776{
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -0700777 struct netlink_sock *nlk = nlk_sk(sk);
778
779 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
780 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
781 sk->sk_err = ENOBUFS;
782 sk->sk_error_report(sk);
783 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 }
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -0700785 atomic_inc(&sk->sk_drops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786}
787
Eric W. Biederman15e47302012-09-07 20:12:54 +0000788static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 struct sock *sock;
791 struct netlink_sock *nlk;
792
Eric W. Biederman15e47302012-09-07 20:12:54 +0000793 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 if (!sock)
795 return ERR_PTR(-ECONNREFUSED);
796
797 /* Don't bother queuing skb if kernel socket has no input function */
798 nlk = nlk_sk(sock);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700799 if (sock->sk_state == NETLINK_CONNECTED &&
Eric W. Biederman15e47302012-09-07 20:12:54 +0000800 nlk->dst_portid != nlk_sk(ssk)->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 sock_put(sock);
802 return ERR_PTR(-ECONNREFUSED);
803 }
804 return sock;
805}
806
807struct sock *netlink_getsockbyfilp(struct file *filp)
808{
Josef Sipek6db5fc52006-12-08 02:37:25 -0800809 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 struct sock *sock;
811
812 if (!S_ISSOCK(inode->i_mode))
813 return ERR_PTR(-ENOTSOCK);
814
815 sock = SOCKET_I(inode)->sk;
816 if (sock->sk_family != AF_NETLINK)
817 return ERR_PTR(-EINVAL);
818
819 sock_hold(sock);
820 return sock;
821}
822
823/*
824 * Attach a skb to a netlink socket.
825 * The caller must hold a reference to the destination socket. On error, the
826 * reference is dropped. The skb is not send to the destination, just all
827 * all error checks are performed and memory in the queue is reserved.
828 * Return values:
829 * < 0: error. skb freed, reference to sock dropped.
830 * 0: continue
831 * 1: repeat lookup - reference dropped while waiting for socket memory.
832 */
Denis V. Lunev9457afe2008-06-05 11:23:39 -0700833int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800834 long *timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
836 struct netlink_sock *nlk;
837
838 nlk = nlk_sk(sk);
839
840 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
841 test_bit(0, &nlk->state)) {
842 DECLARE_WAITQUEUE(wait, current);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800843 if (!*timeo) {
Denis V. Lunevaed81562007-10-10 21:14:32 -0700844 if (!ssk || netlink_is_kernel(ssk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 netlink_overrun(sk);
846 sock_put(sk);
847 kfree_skb(skb);
848 return -EAGAIN;
849 }
850
851 __set_current_state(TASK_INTERRUPTIBLE);
852 add_wait_queue(&nlk->wait, &wait);
853
854 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
855 test_bit(0, &nlk->state)) &&
856 !sock_flag(sk, SOCK_DEAD))
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800857 *timeo = schedule_timeout(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
859 __set_current_state(TASK_RUNNING);
860 remove_wait_queue(&nlk->wait, &wait);
861 sock_put(sk);
862
863 if (signal_pending(current)) {
864 kfree_skb(skb);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800865 return sock_intr_errno(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 }
867 return 1;
868 }
869 skb_set_owner_r(skb, sk);
870 return 0;
871}
872
Eric Dumazet4a7e7c22012-04-05 22:17:46 +0000873static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 int len = skb->len;
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 skb_queue_tail(&sk->sk_receive_queue, skb);
878 sk->sk_data_ready(sk, len);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +0000879 return len;
880}
881
882int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
883{
884 int len = __netlink_sendskb(sk, skb);
885
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 sock_put(sk);
887 return len;
888}
889
890void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
891{
892 kfree_skb(skb);
893 sock_put(sk);
894}
895
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000896static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897{
898 int delta;
899
900 skb_orphan(skb);
901
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700902 delta = skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 if (delta * 2 < skb->truesize)
904 return skb;
905
906 if (skb_shared(skb)) {
907 struct sk_buff *nskb = skb_clone(skb, allocation);
908 if (!nskb)
909 return skb;
Eric Dumazet8460c002012-04-19 02:24:28 +0000910 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 skb = nskb;
912 }
913
914 if (!pskb_expand_head(skb, 0, -delta, allocation))
915 skb->truesize -= delta;
916
917 return skb;
918}
919
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000920static void netlink_rcv_wake(struct sock *sk)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700921{
922 struct netlink_sock *nlk = nlk_sk(sk);
923
924 if (skb_queue_empty(&sk->sk_receive_queue))
925 clear_bit(0, &nlk->state);
926 if (!test_bit(0, &nlk->state))
927 wake_up_interruptible(&nlk->wait);
928}
929
Eric W. Biederman3fbc2902012-05-24 17:21:27 -0600930static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
931 struct sock *ssk)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700932{
933 int ret;
934 struct netlink_sock *nlk = nlk_sk(sk);
935
936 ret = -ECONNREFUSED;
937 if (nlk->netlink_rcv != NULL) {
938 ret = skb->len;
939 skb_set_owner_r(skb, sk);
Eric W. Biederman3fbc2902012-05-24 17:21:27 -0600940 NETLINK_CB(skb).ssk = ssk;
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700941 nlk->netlink_rcv(skb);
Eric Dumazetbfb253c2012-04-22 21:30:29 +0000942 consume_skb(skb);
943 } else {
944 kfree_skb(skb);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700945 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700946 sock_put(sk);
947 return ret;
948}
949
950int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
Eric W. Biederman15e47302012-09-07 20:12:54 +0000951 u32 portid, int nonblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952{
953 struct sock *sk;
954 int err;
955 long timeo;
956
957 skb = netlink_trim(skb, gfp_any());
958
959 timeo = sock_sndtimeo(ssk, nonblock);
960retry:
Eric W. Biederman15e47302012-09-07 20:12:54 +0000961 sk = netlink_getsockbyportid(ssk, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 if (IS_ERR(sk)) {
963 kfree_skb(skb);
964 return PTR_ERR(sk);
965 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700966 if (netlink_is_kernel(sk))
Eric W. Biederman3fbc2902012-05-24 17:21:27 -0600967 return netlink_unicast_kernel(sk, skb, ssk);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700968
Stephen Hemmingerb1153f22008-03-21 15:46:12 -0700969 if (sk_filter(sk, skb)) {
Wang Chen84874602008-07-01 19:55:09 -0700970 err = skb->len;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -0700971 kfree_skb(skb);
972 sock_put(sk);
973 return err;
974 }
975
Denis V. Lunev9457afe2008-06-05 11:23:39 -0700976 err = netlink_attachskb(sk, skb, &timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 if (err == 1)
978 goto retry;
979 if (err)
980 return err;
981
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700982 return netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983}
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800984EXPORT_SYMBOL(netlink_unicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Patrick McHardy4277a082006-03-20 18:52:01 -0800986int netlink_has_listeners(struct sock *sk, unsigned int group)
987{
988 int res = 0;
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000989 struct listeners *listeners;
Patrick McHardy4277a082006-03-20 18:52:01 -0800990
Denis V. Lunevaed81562007-10-10 21:14:32 -0700991 BUG_ON(!netlink_is_kernel(sk));
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700992
993 rcu_read_lock();
994 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
995
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000996 if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000997 res = test_bit(group - 1, listeners->masks);
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700998
999 rcu_read_unlock();
1000
Patrick McHardy4277a082006-03-20 18:52:01 -08001001 return res;
1002}
1003EXPORT_SYMBOL_GPL(netlink_has_listeners);
1004
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001005static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006{
1007 struct netlink_sock *nlk = nlk_sk(sk);
1008
1009 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
1010 !test_bit(0, &nlk->state)) {
1011 skb_set_owner_r(skb, sk);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001012 __netlink_sendskb(sk, skb);
stephen hemminger2c6458002011-12-22 08:52:03 +00001013 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 }
1015 return -1;
1016}
1017
1018struct netlink_broadcast_data {
1019 struct sock *exclude_sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001020 struct net *net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001021 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 u32 group;
1023 int failure;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001024 int delivery_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 int congested;
1026 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -04001027 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 struct sk_buff *skb, *skb2;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001029 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
1030 void *tx_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031};
1032
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001033static int do_one_broadcast(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 struct netlink_broadcast_data *p)
1035{
1036 struct netlink_sock *nlk = nlk_sk(sk);
1037 int val;
1038
1039 if (p->exclude_sk == sk)
1040 goto out;
1041
Eric W. Biederman15e47302012-09-07 20:12:54 +00001042 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001043 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 goto out;
1045
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001046 if (!net_eq(sock_net(sk), p->net))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001047 goto out;
1048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 if (p->failure) {
1050 netlink_overrun(sk);
1051 goto out;
1052 }
1053
1054 sock_hold(sk);
1055 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001056 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 p->skb2 = skb_clone(p->skb, p->allocation);
1058 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001059 p->skb2 = skb_get(p->skb);
1060 /*
1061 * skb ownership may have been set when
1062 * delivered to a previous socket.
1063 */
1064 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 }
1066 }
1067 if (p->skb2 == NULL) {
1068 netlink_overrun(sk);
1069 /* Clone failed. Notify ALL listeners. */
1070 p->failure = 1;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001071 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1072 p->delivery_failure = 1;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001073 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1074 kfree_skb(p->skb2);
1075 p->skb2 = NULL;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001076 } else if (sk_filter(sk, p->skb2)) {
1077 kfree_skb(p->skb2);
1078 p->skb2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1080 netlink_overrun(sk);
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001081 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1082 p->delivery_failure = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 } else {
1084 p->congested |= val;
1085 p->delivered = 1;
1086 p->skb2 = NULL;
1087 }
1088 sock_put(sk);
1089
1090out:
1091 return 0;
1092}
1093
Eric W. Biederman15e47302012-09-07 20:12:54 +00001094int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001095 u32 group, gfp_t allocation,
1096 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1097 void *filter_data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001099 struct net *net = sock_net(ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 struct netlink_broadcast_data info;
1101 struct hlist_node *node;
1102 struct sock *sk;
1103
1104 skb = netlink_trim(skb, allocation);
1105
1106 info.exclude_sk = ssk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001107 info.net = net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001108 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 info.group = group;
1110 info.failure = 0;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001111 info.delivery_failure = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 info.congested = 0;
1113 info.delivered = 0;
1114 info.allocation = allocation;
1115 info.skb = skb;
1116 info.skb2 = NULL;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001117 info.tx_filter = filter;
1118 info.tx_data = filter_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
1120 /* While we sleep in clone, do not allow to change socket list */
1121
1122 netlink_lock_table();
1123
1124 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1125 do_one_broadcast(sk, &info);
1126
Neil Horman70d4bf62010-07-20 06:45:56 +00001127 consume_skb(skb);
Tommy S. Christensenaa1c6a6f2005-05-19 13:07:32 -07001128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 netlink_unlock_table();
1130
Neil Horman70d4bf62010-07-20 06:45:56 +00001131 if (info.delivery_failure) {
1132 kfree_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001133 return -ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00001134 }
1135 consume_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001136
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 if (info.delivered) {
1138 if (info.congested && (allocation & __GFP_WAIT))
1139 yield();
1140 return 0;
1141 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 return -ESRCH;
1143}
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001144EXPORT_SYMBOL(netlink_broadcast_filtered);
1145
Eric W. Biederman15e47302012-09-07 20:12:54 +00001146int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001147 u32 group, gfp_t allocation)
1148{
Eric W. Biederman15e47302012-09-07 20:12:54 +00001149 return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001150 NULL, NULL);
1151}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001152EXPORT_SYMBOL(netlink_broadcast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154struct netlink_set_err_data {
1155 struct sock *exclude_sk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001156 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 u32 group;
1158 int code;
1159};
1160
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001161static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162{
1163 struct netlink_sock *nlk = nlk_sk(sk);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001164 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165
1166 if (sk == p->exclude_sk)
1167 goto out;
1168
Octavian Purdila09ad9bc2009-11-25 15:14:13 -08001169 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001170 goto out;
1171
Eric W. Biederman15e47302012-09-07 20:12:54 +00001172 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001173 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 goto out;
1175
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001176 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
1177 ret = 1;
1178 goto out;
1179 }
1180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 sk->sk_err = p->code;
1182 sk->sk_error_report(sk);
1183out:
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001184 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185}
1186
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001187/**
1188 * netlink_set_err - report error to broadcast listeners
1189 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
Eric W. Biederman15e47302012-09-07 20:12:54 +00001190 * @portid: the PORTID of a process that we want to skip (if any)
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001191 * @groups: the broadcast group that will notice the error
1192 * @code: error code, must be negative (as usual in kernelspace)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001193 *
1194 * This function returns the number of broadcast listeners that have set the
1195 * NETLINK_RECV_NO_ENOBUFS socket option.
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001196 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001197int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198{
1199 struct netlink_set_err_data info;
1200 struct hlist_node *node;
1201 struct sock *sk;
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001202 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
1204 info.exclude_sk = ssk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001205 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 info.group = group;
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001207 /* sk->sk_err wants a positive error value */
1208 info.code = -code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
1210 read_lock(&nl_table_lock);
1211
1212 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001213 ret += do_one_set_err(sk, &info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214
1215 read_unlock(&nl_table_lock);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001216 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217}
Pablo Neira Ayusodd5b6ce2009-03-23 13:21:06 +01001218EXPORT_SYMBOL(netlink_set_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219
Johannes Berg84659eb2007-07-18 15:47:05 -07001220/* must be called with netlink table grabbed */
1221static void netlink_update_socket_mc(struct netlink_sock *nlk,
1222 unsigned int group,
1223 int is_new)
1224{
1225 int old, new = !!is_new, subscriptions;
1226
1227 old = test_bit(group - 1, nlk->groups);
1228 subscriptions = nlk->subscriptions - old + new;
1229 if (new)
1230 __set_bit(group - 1, nlk->groups);
1231 else
1232 __clear_bit(group - 1, nlk->groups);
1233 netlink_update_subscriptions(&nlk->sk, subscriptions);
1234 netlink_update_listeners(&nlk->sk);
1235}
1236
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001237static int netlink_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001238 char __user *optval, unsigned int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001239{
1240 struct sock *sk = sock->sk;
1241 struct netlink_sock *nlk = nlk_sk(sk);
Johannes Bergeb496532007-07-18 02:07:51 -07001242 unsigned int val = 0;
1243 int err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001244
1245 if (level != SOL_NETLINK)
1246 return -ENOPROTOOPT;
1247
1248 if (optlen >= sizeof(int) &&
Johannes Bergeb496532007-07-18 02:07:51 -07001249 get_user(val, (unsigned int __user *)optval))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001250 return -EFAULT;
1251
1252 switch (optname) {
1253 case NETLINK_PKTINFO:
1254 if (val)
1255 nlk->flags |= NETLINK_RECV_PKTINFO;
1256 else
1257 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1258 err = 0;
1259 break;
1260 case NETLINK_ADD_MEMBERSHIP:
1261 case NETLINK_DROP_MEMBERSHIP: {
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001262 if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001263 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001264 err = netlink_realloc_groups(sk);
1265 if (err)
1266 return err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001267 if (!val || val - 1 >= nlk->ngroups)
1268 return -EINVAL;
1269 netlink_table_grab();
Johannes Berg84659eb2007-07-18 15:47:05 -07001270 netlink_update_socket_mc(nlk, val,
1271 optname == NETLINK_ADD_MEMBERSHIP);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001272 netlink_table_ungrab();
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00001273
1274 if (nlk->netlink_bind)
1275 nlk->netlink_bind(val);
1276
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001277 err = 0;
1278 break;
1279 }
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001280 case NETLINK_BROADCAST_ERROR:
1281 if (val)
1282 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
1283 else
1284 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
1285 err = 0;
1286 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001287 case NETLINK_NO_ENOBUFS:
1288 if (val) {
1289 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
1290 clear_bit(0, &nlk->state);
1291 wake_up_interruptible(&nlk->wait);
Eric Dumazet658cb352012-04-22 21:30:21 +00001292 } else {
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001293 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00001294 }
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001295 err = 0;
1296 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001297 default:
1298 err = -ENOPROTOOPT;
1299 }
1300 return err;
1301}
1302
1303static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001304 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001305{
1306 struct sock *sk = sock->sk;
1307 struct netlink_sock *nlk = nlk_sk(sk);
1308 int len, val, err;
1309
1310 if (level != SOL_NETLINK)
1311 return -ENOPROTOOPT;
1312
1313 if (get_user(len, optlen))
1314 return -EFAULT;
1315 if (len < 0)
1316 return -EINVAL;
1317
1318 switch (optname) {
1319 case NETLINK_PKTINFO:
1320 if (len < sizeof(int))
1321 return -EINVAL;
1322 len = sizeof(int);
1323 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08001324 if (put_user(len, optlen) ||
1325 put_user(val, optval))
1326 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001327 err = 0;
1328 break;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001329 case NETLINK_BROADCAST_ERROR:
1330 if (len < sizeof(int))
1331 return -EINVAL;
1332 len = sizeof(int);
1333 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
1334 if (put_user(len, optlen) ||
1335 put_user(val, optval))
1336 return -EFAULT;
1337 err = 0;
1338 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001339 case NETLINK_NO_ENOBUFS:
1340 if (len < sizeof(int))
1341 return -EINVAL;
1342 len = sizeof(int);
1343 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
1344 if (put_user(len, optlen) ||
1345 put_user(val, optval))
1346 return -EFAULT;
1347 err = 0;
1348 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001349 default:
1350 err = -ENOPROTOOPT;
1351 }
1352 return err;
1353}
1354
1355static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1356{
1357 struct nl_pktinfo info;
1358
1359 info.group = NETLINK_CB(skb).dst_group;
1360 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1361}
1362
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1364 struct msghdr *msg, size_t len)
1365{
1366 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1367 struct sock *sk = sock->sk;
1368 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001369 struct sockaddr_nl *addr = msg->msg_name;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001370 u32 dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001371 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 struct sk_buff *skb;
1373 int err;
1374 struct scm_cookie scm;
1375
1376 if (msg->msg_flags&MSG_OOB)
1377 return -EOPNOTSUPP;
1378
Eric Dumazet16e57262011-09-19 05:52:27 +00001379 if (NULL == siocb->scm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 siocb->scm = &scm;
Eric Dumazet16e57262011-09-19 05:52:27 +00001381
Eric Dumazete0e3cea2012-08-21 06:21:17 +00001382 err = scm_send(sock, msg, siocb->scm, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 if (err < 0)
1384 return err;
1385
1386 if (msg->msg_namelen) {
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001387 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 if (addr->nl_family != AF_NETLINK)
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001389 goto out;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001390 dst_portid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001391 dst_group = ffs(addr->nl_groups);
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001392 err = -EPERM;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001393 if ((dst_group || dst_portid) &&
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001394 !netlink_capable(sock, NL_CFG_F_NONROOT_SEND))
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001395 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001397 dst_portid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001398 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 }
1400
Eric W. Biederman15e47302012-09-07 20:12:54 +00001401 if (!nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 err = netlink_autobind(sock);
1403 if (err)
1404 goto out;
1405 }
1406
1407 err = -EMSGSIZE;
1408 if (len > sk->sk_sndbuf - 32)
1409 goto out;
1410 err = -ENOBUFS;
Thomas Graf339bf982006-11-10 14:10:15 -08001411 skb = alloc_skb(len, GFP_KERNEL);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001412 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 goto out;
1414
Eric W. Biederman15e47302012-09-07 20:12:54 +00001415 NETLINK_CB(skb).portid = nlk->portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001416 NETLINK_CB(skb).dst_group = dst_group;
Eric W. Biedermandbe9a412012-09-06 18:20:01 +00001417 NETLINK_CB(skb).creds = siocb->scm->creds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 err = -EFAULT;
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001420 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 kfree_skb(skb);
1422 goto out;
1423 }
1424
1425 err = security_netlink_send(sk, skb);
1426 if (err) {
1427 kfree_skb(skb);
1428 goto out;
1429 }
1430
Patrick McHardyd629b832005-08-14 19:27:50 -07001431 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00001433 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 }
Eric W. Biederman15e47302012-09-07 20:12:54 +00001435 err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436
1437out:
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001438 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 return err;
1440}
1441
1442static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1443 struct msghdr *msg, size_t len,
1444 int flags)
1445{
1446 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1447 struct scm_cookie scm;
1448 struct sock *sk = sock->sk;
1449 struct netlink_sock *nlk = nlk_sk(sk);
1450 int noblock = flags&MSG_DONTWAIT;
1451 size_t copied;
Johannes Berg68d6ac62010-08-15 21:20:44 +00001452 struct sk_buff *skb, *data_skb;
Andrey Vaginb44d2112011-02-21 02:40:47 +00001453 int err, ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454
1455 if (flags&MSG_OOB)
1456 return -EOPNOTSUPP;
1457
1458 copied = 0;
1459
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001460 skb = skb_recv_datagram(sk, flags, noblock, &err);
1461 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 goto out;
1463
Johannes Berg68d6ac62010-08-15 21:20:44 +00001464 data_skb = skb;
1465
Johannes Berg1dacc762009-07-01 11:26:02 +00001466#ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1467 if (unlikely(skb_shinfo(skb)->frag_list)) {
Johannes Berg1dacc762009-07-01 11:26:02 +00001468 /*
Johannes Berg68d6ac62010-08-15 21:20:44 +00001469 * If this skb has a frag_list, then here that means that we
1470 * will have to use the frag_list skb's data for compat tasks
1471 * and the regular skb's data for normal (non-compat) tasks.
Johannes Berg1dacc762009-07-01 11:26:02 +00001472 *
Johannes Berg68d6ac62010-08-15 21:20:44 +00001473 * If we need to send the compat skb, assign it to the
1474 * 'data_skb' variable so that it will be used below for data
1475 * copying. We keep 'skb' for everything else, including
1476 * freeing both later.
Johannes Berg1dacc762009-07-01 11:26:02 +00001477 */
Johannes Berg68d6ac62010-08-15 21:20:44 +00001478 if (flags & MSG_CMSG_COMPAT)
1479 data_skb = skb_shinfo(skb)->frag_list;
Johannes Berg1dacc762009-07-01 11:26:02 +00001480 }
1481#endif
1482
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 msg->msg_namelen = 0;
1484
Johannes Berg68d6ac62010-08-15 21:20:44 +00001485 copied = data_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 if (len < copied) {
1487 msg->msg_flags |= MSG_TRUNC;
1488 copied = len;
1489 }
1490
Johannes Berg68d6ac62010-08-15 21:20:44 +00001491 skb_reset_transport_header(data_skb);
1492 err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
1494 if (msg->msg_name) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001495 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 addr->nl_family = AF_NETLINK;
1497 addr->nl_pad = 0;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001498 addr->nl_pid = NETLINK_CB(skb).portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001499 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 msg->msg_namelen = sizeof(*addr);
1501 }
1502
Patrick McHardycc9a06c2006-03-12 20:34:27 -08001503 if (nlk->flags & NETLINK_RECV_PKTINFO)
1504 netlink_cmsg_recv_pktinfo(msg, skb);
1505
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 if (NULL == siocb->scm) {
1507 memset(&scm, 0, sizeof(scm));
1508 siocb->scm = &scm;
1509 }
1510 siocb->scm->creds = *NETLINK_CREDS(skb);
Patrick McHardy188ccb52007-05-03 03:27:01 -07001511 if (flags & MSG_TRUNC)
Johannes Berg68d6ac62010-08-15 21:20:44 +00001512 copied = data_skb->len;
David S. Millerdaa37662010-08-15 23:21:50 -07001513
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 skb_free_datagram(sk, skb);
1515
Andrey Vaginb44d2112011-02-21 02:40:47 +00001516 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
1517 ret = netlink_dump(sk);
1518 if (ret) {
1519 sk->sk_err = ret;
1520 sk->sk_error_report(sk);
1521 }
1522 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
1524 scm_recv(sock, msg, siocb->scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525out:
1526 netlink_rcv_wake(sk);
1527 return err ? : copied;
1528}
1529
1530static void netlink_data_ready(struct sock *sk, int len)
1531{
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001532 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533}
1534
1535/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001536 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 * complete set of kernel non-blocking support for message
1538 * queueing.
1539 */
1540
1541struct sock *
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00001542__netlink_kernel_create(struct net *net, int unit, struct module *module,
1543 struct netlink_kernel_cfg *cfg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544{
1545 struct socket *sock;
1546 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001547 struct netlink_sock *nlk;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001548 struct listeners *listeners = NULL;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001549 struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
1550 unsigned int groups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001552 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001554 if (unit < 0 || unit >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 return NULL;
1556
1557 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1558 return NULL;
1559
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08001560 /*
1561 * We have to just have a reference on the net from sk, but don't
1562 * get_net it. Besides, we cannot get and then put the net here.
1563 * So we create one inside init_net and the move it to net.
1564 */
1565
1566 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
1567 goto out_sock_release_nosk;
1568
1569 sk = sock->sk;
Denis V. Lunevedf02082008-02-29 11:18:32 -08001570 sk_change_net(sk, net);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001571
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001572 if (!cfg || cfg->groups < 32)
Patrick McHardy4277a082006-03-20 18:52:01 -08001573 groups = 32;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001574 else
1575 groups = cfg->groups;
Patrick McHardy4277a082006-03-20 18:52:01 -08001576
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001577 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
Patrick McHardy4277a082006-03-20 18:52:01 -08001578 if (!listeners)
1579 goto out_sock_release;
1580
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 sk->sk_data_ready = netlink_data_ready;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00001582 if (cfg && cfg->input)
1583 nlk_sk(sk)->netlink_rcv = cfg->input;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001585 if (netlink_insert(sk, net, 0))
Patrick McHardy77247bb2005-08-14 19:27:13 -07001586 goto out_sock_release;
1587
1588 nlk = nlk_sk(sk);
1589 nlk->flags |= NETLINK_KERNEL_SOCKET;
1590
1591 netlink_table_grab();
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001592 if (!nl_table[unit].registered) {
1593 nl_table[unit].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001594 rcu_assign_pointer(nl_table[unit].listeners, listeners);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001595 nl_table[unit].cb_mutex = cb_mutex;
1596 nl_table[unit].module = module;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001597 if (cfg) {
1598 nl_table[unit].bind = cfg->bind;
1599 nl_table[unit].flags = cfg->flags;
1600 }
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001601 nl_table[unit].registered = 1;
Jesper Juhlf937f1f462007-10-15 01:39:12 -07001602 } else {
1603 kfree(listeners);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001604 nl_table[unit].registered++;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001605 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07001606 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001607 return sk;
1608
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001609out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08001610 kfree(listeners);
Denis V. Lunev9dfbec12008-02-29 11:17:56 -08001611 netlink_kernel_release(sk);
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08001612 return NULL;
1613
1614out_sock_release_nosk:
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001615 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001616 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617}
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00001618EXPORT_SYMBOL(__netlink_kernel_create);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001619
1620void
1621netlink_kernel_release(struct sock *sk)
1622{
Denis V. Lunevedf02082008-02-29 11:18:32 -08001623 sk_release_kernel(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001624}
1625EXPORT_SYMBOL(netlink_kernel_release);
1626
Johannes Bergd136f1b2009-09-12 03:03:15 +00001627int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001628{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001629 struct listeners *new, *old;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001630 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001631
1632 if (groups < 32)
1633 groups = 32;
1634
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001635 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001636 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
1637 if (!new)
Johannes Bergd136f1b2009-09-12 03:03:15 +00001638 return -ENOMEM;
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001639 old = nl_deref_protected(tbl->listeners);
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001640 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
1641 rcu_assign_pointer(tbl->listeners, new);
1642
Lai Jiangshan37b6b932011-03-15 18:01:42 +08001643 kfree_rcu(old, rcu);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001644 }
1645 tbl->groups = groups;
1646
Johannes Bergd136f1b2009-09-12 03:03:15 +00001647 return 0;
1648}
1649
1650/**
1651 * netlink_change_ngroups - change number of multicast groups
1652 *
1653 * This changes the number of multicast groups that are available
1654 * on a certain netlink family. Note that it is not possible to
1655 * change the number of groups to below 32. Also note that it does
1656 * not implicitly call netlink_clear_multicast_users() when the
1657 * number of groups is reduced.
1658 *
1659 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1660 * @groups: The new number of groups.
1661 */
1662int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1663{
1664 int err;
1665
1666 netlink_table_grab();
1667 err = __netlink_change_ngroups(sk, groups);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001668 netlink_table_ungrab();
Johannes Bergd136f1b2009-09-12 03:03:15 +00001669
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001670 return err;
1671}
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001672
Johannes Bergb8273572009-09-24 15:44:05 -07001673void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1674{
1675 struct sock *sk;
1676 struct hlist_node *node;
1677 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1678
1679 sk_for_each_bound(sk, node, &tbl->mc_list)
1680 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1681}
1682
Johannes Berg84659eb2007-07-18 15:47:05 -07001683/**
1684 * netlink_clear_multicast_users - kick off multicast listeners
1685 *
1686 * This function removes all listeners from the given group.
1687 * @ksk: The kernel netlink socket, as returned by
1688 * netlink_kernel_create().
1689 * @group: The multicast group to clear.
1690 */
1691void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1692{
Johannes Berg84659eb2007-07-18 15:47:05 -07001693 netlink_table_grab();
Johannes Bergb8273572009-09-24 15:44:05 -07001694 __netlink_clear_multicast_users(ksk, group);
Johannes Berg84659eb2007-07-18 15:47:05 -07001695 netlink_table_ungrab();
1696}
Johannes Berg84659eb2007-07-18 15:47:05 -07001697
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05001698struct nlmsghdr *
Eric W. Biederman15e47302012-09-07 20:12:54 +00001699__nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05001700{
1701 struct nlmsghdr *nlh;
1702 int size = NLMSG_LENGTH(len);
1703
1704 nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size));
1705 nlh->nlmsg_type = type;
1706 nlh->nlmsg_len = size;
1707 nlh->nlmsg_flags = flags;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001708 nlh->nlmsg_pid = portid;
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05001709 nlh->nlmsg_seq = seq;
1710 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
1711 memset(NLMSG_DATA(nlh) + len, 0, NLMSG_ALIGN(size) - size);
1712 return nlh;
1713}
1714EXPORT_SYMBOL(__nlmsg_put);
1715
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716/*
1717 * It looks a bit ugly.
1718 * It would be better to create kernel thread.
1719 */
1720
1721static int netlink_dump(struct sock *sk)
1722{
1723 struct netlink_sock *nlk = nlk_sk(sk);
1724 struct netlink_callback *cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00001725 struct sk_buff *skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001727 int len, err = -ENOBUFS;
Greg Rosec7ac8672011-06-10 01:27:09 +00001728 int alloc_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001730 mutex_lock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732 cb = nlk->cb;
1733 if (cb == NULL) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001734 err = -EINVAL;
1735 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 }
1737
Greg Rosec7ac8672011-06-10 01:27:09 +00001738 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
1739
1740 skb = sock_rmalloc(sk, alloc_size, 0, GFP_KERNEL);
1741 if (!skb)
Dan Carpenterc63d6ea2011-06-15 03:11:42 +00001742 goto errout_skb;
Greg Rosec7ac8672011-06-10 01:27:09 +00001743
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 len = cb->dump(skb, cb);
1745
1746 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001747 mutex_unlock(nlk->cb_mutex);
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001748
1749 if (sk_filter(sk, skb))
1750 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001751 else
1752 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 return 0;
1754 }
1755
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001756 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1757 if (!nlh)
1758 goto errout_skb;
1759
Johannes Berg670dc282011-06-20 13:40:46 +02001760 nl_dump_check_consistent(cb, nlh);
1761
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001762 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1763
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001764 if (sk_filter(sk, skb))
1765 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001766 else
1767 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
Thomas Grafa8f74b22005-11-10 02:25:52 +01001769 if (cb->done)
1770 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 nlk->cb = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001772 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
Gao feng6dc878a2012-10-04 20:15:48 +00001774 module_put(cb->module);
Eric Dumazetbfb253c2012-04-22 21:30:29 +00001775 netlink_consume_callback(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07001777
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001778errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001779 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001780 kfree_skb(skb);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001781 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782}
1783
Gao feng6dc878a2012-10-04 20:15:48 +00001784int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1785 const struct nlmsghdr *nlh,
1786 struct netlink_dump_control *control)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787{
1788 struct netlink_callback *cb;
1789 struct sock *sk;
1790 struct netlink_sock *nlk;
Andrey Vaginb44d2112011-02-21 02:40:47 +00001791 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001793 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 if (cb == NULL)
1795 return -ENOBUFS;
1796
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00001797 cb->dump = control->dump;
1798 cb->done = control->done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 cb->nlh = nlh;
Pablo Neira Ayuso7175c882012-02-24 14:30:16 +00001800 cb->data = control->data;
Gao feng6dc878a2012-10-04 20:15:48 +00001801 cb->module = control->module;
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00001802 cb->min_dump_alloc = control->min_dump_alloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 atomic_inc(&skb->users);
1804 cb->skb = skb;
1805
Eric W. Biederman15e47302012-09-07 20:12:54 +00001806 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 if (sk == NULL) {
1808 netlink_destroy_callback(cb);
1809 return -ECONNREFUSED;
1810 }
1811 nlk = nlk_sk(sk);
Gao feng6dc878a2012-10-04 20:15:48 +00001812
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001813 mutex_lock(nlk->cb_mutex);
Gao feng6dc878a2012-10-04 20:15:48 +00001814 /* A dump is in progress... */
Herbert Xu3f660d62007-05-03 03:17:14 -07001815 if (nlk->cb) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001816 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 netlink_destroy_callback(cb);
Gao feng6dc878a2012-10-04 20:15:48 +00001818 ret = -EBUSY;
1819 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 }
Gao feng6dc878a2012-10-04 20:15:48 +00001821 /* add reference of module which cb->dump belongs to */
1822 if (!try_module_get(cb->module)) {
1823 mutex_unlock(nlk->cb_mutex);
1824 netlink_destroy_callback(cb);
1825 ret = -EPROTONOSUPPORT;
1826 goto out;
1827 }
1828
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 nlk->cb = cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001830 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831
Andrey Vaginb44d2112011-02-21 02:40:47 +00001832 ret = netlink_dump(sk);
Gao feng6dc878a2012-10-04 20:15:48 +00001833out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 sock_put(sk);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001835
Andrey Vaginb44d2112011-02-21 02:40:47 +00001836 if (ret)
1837 return ret;
1838
Denis V. Lunev5c582982007-10-23 20:29:25 -07001839 /* We successfully started a dump, by returning -EINTR we
1840 * signal not to send ACK even if it was requested.
1841 */
1842 return -EINTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843}
Gao feng6dc878a2012-10-04 20:15:48 +00001844EXPORT_SYMBOL(__netlink_dump_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845
1846void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1847{
1848 struct sk_buff *skb;
1849 struct nlmsghdr *rep;
1850 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08001851 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
Thomas Graf339bf982006-11-10 14:10:15 -08001853 /* error messages get the original request appened */
1854 if (err)
1855 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
Thomas Graf339bf982006-11-10 14:10:15 -08001857 skb = nlmsg_new(payload, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 if (!skb) {
1859 struct sock *sk;
1860
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001861 sk = netlink_lookup(sock_net(in_skb->sk),
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001862 in_skb->sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00001863 NETLINK_CB(in_skb).portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 if (sk) {
1865 sk->sk_err = ENOBUFS;
1866 sk->sk_error_report(sk);
1867 sock_put(sk);
1868 }
1869 return;
1870 }
1871
Eric W. Biederman15e47302012-09-07 20:12:54 +00001872 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
John Fastabend5dba93a2009-09-25 13:11:44 +00001873 NLMSG_ERROR, payload, 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001874 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001876 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Eric W. Biederman15e47302012-09-07 20:12:54 +00001877 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001879EXPORT_SYMBOL(netlink_ack);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001881int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001882 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01001883{
Thomas Graf82ace472005-11-10 02:25:53 +01001884 struct nlmsghdr *nlh;
1885 int err;
1886
1887 while (skb->len >= nlmsg_total_size(0)) {
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001888 int msglen;
1889
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001890 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07001891 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01001892
Martin Murrayad8e4b72006-01-10 13:02:29 -08001893 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01001894 return 0;
1895
Thomas Grafd35b6852007-03-22 23:28:46 -07001896 /* Only requests are handled by the kernel */
1897 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
Denis V. Lunev5c582982007-10-23 20:29:25 -07001898 goto ack;
Thomas Grafd35b6852007-03-22 23:28:46 -07001899
Thomas Graf45e7ae72007-03-22 23:29:10 -07001900 /* Skip control messages */
1901 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
Denis V. Lunev5c582982007-10-23 20:29:25 -07001902 goto ack;
Thomas Graf45e7ae72007-03-22 23:29:10 -07001903
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001904 err = cb(skb, nlh);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001905 if (err == -EINTR)
1906 goto skip;
1907
1908ack:
Thomas Grafd35b6852007-03-22 23:28:46 -07001909 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01001910 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01001911
Denis V. Lunev5c582982007-10-23 20:29:25 -07001912skip:
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001913 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001914 if (msglen > skb->len)
1915 msglen = skb->len;
1916 skb_pull(skb, msglen);
Thomas Graf82ace472005-11-10 02:25:53 +01001917 }
1918
1919 return 0;
1920}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001921EXPORT_SYMBOL(netlink_rcv_skb);
Thomas Graf82ace472005-11-10 02:25:53 +01001922
1923/**
Thomas Grafd387f6a2006-08-15 00:31:06 -07001924 * nlmsg_notify - send a notification netlink message
1925 * @sk: netlink socket to use
1926 * @skb: notification message
Eric W. Biederman15e47302012-09-07 20:12:54 +00001927 * @portid: destination netlink portid for reports or 0
Thomas Grafd387f6a2006-08-15 00:31:06 -07001928 * @group: destination multicast group or 0
1929 * @report: 1 to report back, 0 to disable
1930 * @flags: allocation flags
1931 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001932int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
Thomas Grafd387f6a2006-08-15 00:31:06 -07001933 unsigned int group, int report, gfp_t flags)
1934{
1935 int err = 0;
1936
1937 if (group) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001938 int exclude_portid = 0;
Thomas Grafd387f6a2006-08-15 00:31:06 -07001939
1940 if (report) {
1941 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00001942 exclude_portid = portid;
Thomas Grafd387f6a2006-08-15 00:31:06 -07001943 }
1944
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08001945 /* errors reported via destination sk->sk_err, but propagate
1946 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
Eric W. Biederman15e47302012-09-07 20:12:54 +00001947 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001948 }
1949
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08001950 if (report) {
1951 int err2;
1952
Eric W. Biederman15e47302012-09-07 20:12:54 +00001953 err2 = nlmsg_unicast(sk, skb, portid);
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08001954 if (!err || err == -ESRCH)
1955 err = err2;
1956 }
Thomas Grafd387f6a2006-08-15 00:31:06 -07001957
1958 return err;
1959}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001960EXPORT_SYMBOL(nlmsg_notify);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001961
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962#ifdef CONFIG_PROC_FS
1963struct nl_seq_iter {
Denis V. Luneve372c412007-11-19 22:31:54 -08001964 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 int link;
1966 int hash_idx;
1967};
1968
1969static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1970{
1971 struct nl_seq_iter *iter = seq->private;
1972 int i, j;
1973 struct sock *s;
1974 struct hlist_node *node;
1975 loff_t off = 0;
1976
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001977 for (i = 0; i < MAX_LINKS; i++) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001978 struct nl_portid_hash *hash = &nl_table[i].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979
1980 for (j = 0; j <= hash->mask; j++) {
1981 sk_for_each(s, node, &hash->table[j]) {
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09001982 if (sock_net(s) != seq_file_net(seq))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001983 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 if (off == pos) {
1985 iter->link = i;
1986 iter->hash_idx = j;
1987 return s;
1988 }
1989 ++off;
1990 }
1991 }
1992 }
1993 return NULL;
1994}
1995
1996static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001997 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998{
1999 read_lock(&nl_table_lock);
2000 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2001}
2002
2003static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2004{
2005 struct sock *s;
2006 struct nl_seq_iter *iter;
2007 int i, j;
2008
2009 ++*pos;
2010
2011 if (v == SEQ_START_TOKEN)
2012 return netlink_seq_socket_idx(seq, 0);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002013
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002014 iter = seq->private;
2015 s = v;
2016 do {
2017 s = sk_next(s);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002018 } while (s && sock_net(s) != seq_file_net(seq));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 if (s)
2020 return s;
2021
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 i = iter->link;
2023 j = iter->hash_idx + 1;
2024
2025 do {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002026 struct nl_portid_hash *hash = &nl_table[i].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027
2028 for (; j <= hash->mask; j++) {
2029 s = sk_head(&hash->table[j]);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09002030 while (s && sock_net(s) != seq_file_net(seq))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002031 s = sk_next(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 if (s) {
2033 iter->link = i;
2034 iter->hash_idx = j;
2035 return s;
2036 }
2037 }
2038
2039 j = 0;
2040 } while (++i < MAX_LINKS);
2041
2042 return NULL;
2043}
2044
2045static void netlink_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002046 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047{
2048 read_unlock(&nl_table_lock);
2049}
2050
2051
2052static int netlink_seq_show(struct seq_file *seq, void *v)
2053{
Eric Dumazet658cb352012-04-22 21:30:21 +00002054 if (v == SEQ_START_TOKEN) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 seq_puts(seq,
2056 "sk Eth Pid Groups "
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00002057 "Rmem Wmem Dump Locks Drops Inode\n");
Eric Dumazet658cb352012-04-22 21:30:21 +00002058 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 struct sock *s = v;
2060 struct netlink_sock *nlk = nlk_sk(s);
2061
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002062 seq_printf(seq, "%pK %-3d %-6d %08x %-8d %-8d %pK %-8d %-8d %-8lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 s,
2064 s->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00002065 nlk->portid,
Patrick McHardy513c2502005-09-06 15:43:59 -07002066 nlk->groups ? (u32)nlk->groups[0] : 0,
Eric Dumazet31e6d362009-06-17 19:05:41 -07002067 sk_rmem_alloc_get(s),
2068 sk_wmem_alloc_get(s),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069 nlk->cb,
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002070 atomic_read(&s->sk_refcnt),
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00002071 atomic_read(&s->sk_drops),
2072 sock_i_ino(s)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 );
2074
2075 }
2076 return 0;
2077}
2078
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002079static const struct seq_operations netlink_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 .start = netlink_seq_start,
2081 .next = netlink_seq_next,
2082 .stop = netlink_seq_stop,
2083 .show = netlink_seq_show,
2084};
2085
2086
2087static int netlink_seq_open(struct inode *inode, struct file *file)
2088{
Denis V. Luneve372c412007-11-19 22:31:54 -08002089 return seq_open_net(inode, file, &netlink_seq_ops,
2090 sizeof(struct nl_seq_iter));
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002091}
2092
Arjan van de Venda7071d2007-02-12 00:55:36 -08002093static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 .owner = THIS_MODULE,
2095 .open = netlink_seq_open,
2096 .read = seq_read,
2097 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002098 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099};
2100
2101#endif
2102
2103int netlink_register_notifier(struct notifier_block *nb)
2104{
Alan Sterne041c682006-03-27 01:16:30 -08002105 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002107EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108
2109int netlink_unregister_notifier(struct notifier_block *nb)
2110{
Alan Sterne041c682006-03-27 01:16:30 -08002111 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002113EXPORT_SYMBOL(netlink_unregister_notifier);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002114
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08002115static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 .family = PF_NETLINK,
2117 .owner = THIS_MODULE,
2118 .release = netlink_release,
2119 .bind = netlink_bind,
2120 .connect = netlink_connect,
2121 .socketpair = sock_no_socketpair,
2122 .accept = sock_no_accept,
2123 .getname = netlink_getname,
2124 .poll = datagram_poll,
2125 .ioctl = sock_no_ioctl,
2126 .listen = sock_no_listen,
2127 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002128 .setsockopt = netlink_setsockopt,
2129 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 .sendmsg = netlink_sendmsg,
2131 .recvmsg = netlink_recvmsg,
2132 .mmap = sock_no_mmap,
2133 .sendpage = sock_no_sendpage,
2134};
2135
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002136static const struct net_proto_family netlink_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 .family = PF_NETLINK,
2138 .create = netlink_create,
2139 .owner = THIS_MODULE, /* for consistency 8) */
2140};
2141
Pavel Emelyanov46650792007-10-08 20:38:39 -07002142static int __net_init netlink_net_init(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002143{
2144#ifdef CONFIG_PROC_FS
2145 if (!proc_net_fops_create(net, "netlink", 0, &netlink_seq_fops))
2146 return -ENOMEM;
2147#endif
2148 return 0;
2149}
2150
Pavel Emelyanov46650792007-10-08 20:38:39 -07002151static void __net_exit netlink_net_exit(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002152{
2153#ifdef CONFIG_PROC_FS
2154 proc_net_remove(net, "netlink");
2155#endif
2156}
2157
David S. Millerb963ea82010-08-30 19:08:01 -07002158static void __init netlink_add_usersock_entry(void)
2159{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002160 struct listeners *listeners;
David S. Millerb963ea82010-08-30 19:08:01 -07002161 int groups = 32;
2162
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002163 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
David S. Millerb963ea82010-08-30 19:08:01 -07002164 if (!listeners)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002165 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
David S. Millerb963ea82010-08-30 19:08:01 -07002166
2167 netlink_table_grab();
2168
2169 nl_table[NETLINK_USERSOCK].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002170 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
David S. Millerb963ea82010-08-30 19:08:01 -07002171 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
2172 nl_table[NETLINK_USERSOCK].registered = 1;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002173 nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
David S. Millerb963ea82010-08-30 19:08:01 -07002174
2175 netlink_table_ungrab();
2176}
2177
Denis V. Lunev022cbae2007-11-13 03:23:50 -08002178static struct pernet_operations __net_initdata netlink_net_ops = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002179 .init = netlink_net_init,
2180 .exit = netlink_net_exit,
2181};
2182
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183static int __init netlink_proto_init(void)
2184{
2185 struct sk_buff *dummy_skb;
2186 int i;
Denis Cheng26ff5dd2007-09-16 16:36:02 -07002187 unsigned long limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 unsigned int order;
2189 int err = proto_register(&netlink_proto, 0);
2190
2191 if (err != 0)
2192 goto out;
2193
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07002194 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07002196 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002197 if (!nl_table)
2198 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199
Jan Beulich44813742009-09-21 17:03:05 -07002200 if (totalram_pages >= (128 * 1024))
2201 limit = totalram_pages >> (21 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 else
Jan Beulich44813742009-09-21 17:03:05 -07002203 limit = totalram_pages >> (23 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
Denis Cheng26ff5dd2007-09-16 16:36:02 -07002205 order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
2206 limit = (1UL << order) / sizeof(struct hlist_head);
2207 order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208
2209 for (i = 0; i < MAX_LINKS; i++) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002210 struct nl_portid_hash *hash = &nl_table[i].hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211
Eric W. Biederman15e47302012-09-07 20:12:54 +00002212 hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 if (!hash->table) {
2214 while (i-- > 0)
Eric W. Biederman15e47302012-09-07 20:12:54 +00002215 nl_portid_hash_free(nl_table[i].hash.table,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 1 * sizeof(*hash->table));
2217 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002218 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 hash->max_shift = order;
2221 hash->shift = 0;
2222 hash->mask = 0;
2223 hash->rehash_time = jiffies;
2224 }
2225
David S. Millerb963ea82010-08-30 19:08:01 -07002226 netlink_add_usersock_entry();
2227
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228 sock_register(&netlink_family_ops);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002229 register_pernet_subsys(&netlink_net_ops);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002230 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 rtnetlink_init();
2232out:
2233 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002234panic:
2235 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236}
2237
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238core_initcall(netlink_proto_init);