blob: 6b178e1247b5c3672e3857486d2d1346903181c2 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NETLINK Kernel-user communication protocol.
3 *
4 * Authors: Alan Cox <alan@redhat.com>
5 * 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>
Steve Grubbe7c34972006-04-03 09:08:13 -040057#include <linux/selinux.h>
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070058#include <linux/mutex.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010059
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020060#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <net/sock.h>
62#include <net/scm.h>
Thomas Graf82ace472005-11-10 02:25:53 +010063#include <net/netlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Patrick McHardyf7fa9b12005-08-15 12:29:13 -070065#define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
Johannes Bergb4ff4f02007-07-18 15:46:06 -070066#define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
68struct netlink_sock {
69 /* struct sock has to be the first member of netlink_sock */
70 struct sock sk;
71 u32 pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 u32 dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -070073 u32 dst_group;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -070074 u32 flags;
75 u32 subscriptions;
76 u32 ngroups;
77 unsigned long *groups;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 unsigned long state;
79 wait_queue_head_t wait;
80 struct netlink_callback *cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070081 struct mutex *cb_mutex;
82 struct mutex cb_def_mutex;
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -070083 void (*netlink_rcv)(struct sk_buff *skb);
Patrick McHardy77247bb2005-08-14 19:27:13 -070084 struct module *module;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085};
86
Patrick McHardy77247bb2005-08-14 19:27:13 -070087#define NETLINK_KERNEL_SOCKET 0x1
Patrick McHardy9a4595b2005-08-15 12:32:15 -070088#define NETLINK_RECV_PKTINFO 0x2
Patrick McHardy77247bb2005-08-14 19:27:13 -070089
Linus Torvalds1da177e2005-04-16 15:20:36 -070090static inline struct netlink_sock *nlk_sk(struct sock *sk)
91{
Denis Cheng32b21e02007-08-28 15:41:11 -070092 return container_of(sk, struct netlink_sock, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093}
94
Denis V. Lunevaed81562007-10-10 21:14:32 -070095static inline int netlink_is_kernel(struct sock *sk)
96{
97 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
98}
99
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100struct nl_pid_hash {
101 struct hlist_head *table;
102 unsigned long rehash_time;
103
104 unsigned int mask;
105 unsigned int shift;
106
107 unsigned int entries;
108 unsigned int max_shift;
109
110 u32 rnd;
111};
112
113struct netlink_table {
114 struct nl_pid_hash hash;
115 struct hlist_head mc_list;
Patrick McHardy4277a082006-03-20 18:52:01 -0800116 unsigned long *listeners;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 unsigned int nl_nonroot;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700118 unsigned int groups;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700119 struct mutex *cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700120 struct module *module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700121 int registered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122};
123
124static struct netlink_table *nl_table;
125
126static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
127
128static int netlink_dump(struct sock *sk);
129static void netlink_destroy_callback(struct netlink_callback *cb);
130
131static DEFINE_RWLOCK(nl_table_lock);
132static atomic_t nl_table_users = ATOMIC_INIT(0);
133
Alan Sterne041c682006-03-27 01:16:30 -0800134static ATOMIC_NOTIFIER_HEAD(netlink_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Patrick McHardyd629b832005-08-14 19:27:50 -0700136static u32 netlink_group_mask(u32 group)
137{
138 return group ? 1 << (group - 1) : 0;
139}
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
142{
143 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
144}
145
146static void netlink_sock_destruct(struct sock *sk)
147{
Herbert Xu3f660d62007-05-03 03:17:14 -0700148 struct netlink_sock *nlk = nlk_sk(sk);
149
Herbert Xu3f660d62007-05-03 03:17:14 -0700150 if (nlk->cb) {
151 if (nlk->cb->done)
152 nlk->cb->done(nlk->cb);
153 netlink_destroy_callback(nlk->cb);
154 }
155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 skb_queue_purge(&sk->sk_receive_queue);
157
158 if (!sock_flag(sk, SOCK_DEAD)) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800159 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 return;
161 }
162 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
163 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700164 BUG_TRAP(!nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165}
166
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800167/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
168 * SMP. Look, when several writers sleep and reader wakes them up, all but one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
170 * this, _but_ remember, it adds useless work on UP machines.
171 */
172
173static void netlink_table_grab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800174 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700176 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178 if (atomic_read(&nl_table_users)) {
179 DECLARE_WAITQUEUE(wait, current);
180
181 add_wait_queue_exclusive(&nl_table_wait, &wait);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800182 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 set_current_state(TASK_UNINTERRUPTIBLE);
184 if (atomic_read(&nl_table_users) == 0)
185 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700186 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700188 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 }
190
191 __set_current_state(TASK_RUNNING);
192 remove_wait_queue(&nl_table_wait, &wait);
193 }
194}
195
Ilpo Järvinen3f252522008-01-12 03:21:50 -0800196static void netlink_table_ungrab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800197 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700199 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 wake_up(&nl_table_wait);
201}
202
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800203static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204netlink_lock_table(void)
205{
206 /* read_lock() synchronizes us to netlink_table_grab */
207
208 read_lock(&nl_table_lock);
209 atomic_inc(&nl_table_users);
210 read_unlock(&nl_table_lock);
211}
212
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800213static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214netlink_unlock_table(void)
215{
216 if (atomic_dec_and_test(&nl_table_users))
217 wake_up(&nl_table_wait);
218}
219
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800220static inline struct sock *netlink_lookup(struct net *net, int protocol,
221 u32 pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
223 struct nl_pid_hash *hash = &nl_table[protocol].hash;
224 struct hlist_head *head;
225 struct sock *sk;
226 struct hlist_node *node;
227
228 read_lock(&nl_table_lock);
229 head = nl_pid_hashfn(hash, pid);
230 sk_for_each(sk, node, head) {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200231 if ((sk->sk_net == net) && (nlk_sk(sk)->pid == pid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 sock_hold(sk);
233 goto found;
234 }
235 }
236 sk = NULL;
237found:
238 read_unlock(&nl_table_lock);
239 return sk;
240}
241
Eric Dumazetea729122007-12-11 02:09:47 -0800242static inline struct hlist_head *nl_pid_hash_zalloc(size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243{
244 if (size <= PAGE_SIZE)
Eric Dumazetea729122007-12-11 02:09:47 -0800245 return kzalloc(size, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 else
247 return (struct hlist_head *)
Eric Dumazetea729122007-12-11 02:09:47 -0800248 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
249 get_order(size));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250}
251
252static inline void nl_pid_hash_free(struct hlist_head *table, size_t size)
253{
254 if (size <= PAGE_SIZE)
255 kfree(table);
256 else
257 free_pages((unsigned long)table, get_order(size));
258}
259
260static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
261{
262 unsigned int omask, mask, shift;
263 size_t osize, size;
264 struct hlist_head *otable, *table;
265 int i;
266
267 omask = mask = hash->mask;
268 osize = size = (mask + 1) * sizeof(*table);
269 shift = hash->shift;
270
271 if (grow) {
272 if (++shift > hash->max_shift)
273 return 0;
274 mask = mask * 2 + 1;
275 size *= 2;
276 }
277
Eric Dumazetea729122007-12-11 02:09:47 -0800278 table = nl_pid_hash_zalloc(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 if (!table)
280 return 0;
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 otable = hash->table;
283 hash->table = table;
284 hash->mask = mask;
285 hash->shift = shift;
286 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
287
288 for (i = 0; i <= omask; i++) {
289 struct sock *sk;
290 struct hlist_node *node, *tmp;
291
292 sk_for_each_safe(sk, node, tmp, &otable[i])
293 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
294 }
295
296 nl_pid_hash_free(otable, osize);
297 hash->rehash_time = jiffies + 10 * 60 * HZ;
298 return 1;
299}
300
301static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
302{
303 int avg = hash->entries >> hash->shift;
304
305 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
306 return 1;
307
308 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
309 nl_pid_hash_rehash(hash, 0);
310 return 1;
311 }
312
313 return 0;
314}
315
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800316static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Patrick McHardy4277a082006-03-20 18:52:01 -0800318static void
319netlink_update_listeners(struct sock *sk)
320{
321 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
322 struct hlist_node *node;
323 unsigned long mask;
324 unsigned int i;
325
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700326 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
Patrick McHardy4277a082006-03-20 18:52:01 -0800327 mask = 0;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700328 sk_for_each_bound(sk, node, &tbl->mc_list) {
329 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
330 mask |= nlk_sk(sk)->groups[i];
331 }
Patrick McHardy4277a082006-03-20 18:52:01 -0800332 tbl->listeners[i] = mask;
333 }
334 /* this function is only called with the netlink table "grabbed", which
335 * makes sure updates are visible before bind or setsockopt return. */
336}
337
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200338static int netlink_insert(struct sock *sk, struct net *net, u32 pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339{
340 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
341 struct hlist_head *head;
342 int err = -EADDRINUSE;
343 struct sock *osk;
344 struct hlist_node *node;
345 int len;
346
347 netlink_table_grab();
348 head = nl_pid_hashfn(hash, pid);
349 len = 0;
350 sk_for_each(osk, node, head) {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200351 if ((osk->sk_net == net) && (nlk_sk(osk)->pid == pid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 break;
353 len++;
354 }
355 if (node)
356 goto err;
357
358 err = -EBUSY;
359 if (nlk_sk(sk)->pid)
360 goto err;
361
362 err = -ENOMEM;
363 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
364 goto err;
365
366 if (len && nl_pid_hash_dilute(hash, len))
367 head = nl_pid_hashfn(hash, pid);
368 hash->entries++;
369 nlk_sk(sk)->pid = pid;
370 sk_add_node(sk, head);
371 err = 0;
372
373err:
374 netlink_table_ungrab();
375 return err;
376}
377
378static void netlink_remove(struct sock *sk)
379{
380 netlink_table_grab();
David S. Millerd470e3b2005-06-26 15:31:51 -0700381 if (sk_del_node_init(sk))
382 nl_table[sk->sk_protocol].hash.entries--;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700383 if (nlk_sk(sk)->subscriptions)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 __sk_del_bind_node(sk);
385 netlink_table_ungrab();
386}
387
388static struct proto netlink_proto = {
389 .name = "NETLINK",
390 .owner = THIS_MODULE,
391 .obj_size = sizeof(struct netlink_sock),
392};
393
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700394static int __netlink_create(struct net *net, struct socket *sock,
395 struct mutex *cb_mutex, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396{
397 struct sock *sk;
398 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700399
400 sock->ops = &netlink_ops;
401
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700402 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
Patrick McHardyab33a172005-08-14 19:31:36 -0700403 if (!sk)
404 return -ENOMEM;
405
406 sock_init_data(sock, sk);
407
408 nlk = nlk_sk(sk);
Patrick McHardyffa4d722007-04-25 14:01:17 -0700409 if (cb_mutex)
410 nlk->cb_mutex = cb_mutex;
411 else {
412 nlk->cb_mutex = &nlk->cb_def_mutex;
413 mutex_init(nlk->cb_mutex);
414 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700415 init_waitqueue_head(&nlk->wait);
416
417 sk->sk_destruct = netlink_sock_destruct;
418 sk->sk_protocol = protocol;
419 return 0;
420}
421
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700422static int netlink_create(struct net *net, struct socket *sock, int protocol)
Patrick McHardyab33a172005-08-14 19:31:36 -0700423{
424 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700425 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700426 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700427 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
429 sock->state = SS_UNCONNECTED;
430
431 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
432 return -ESOCKTNOSUPPORT;
433
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800434 if (protocol < 0 || protocol >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 return -EPROTONOSUPPORT;
436
Patrick McHardy77247bb2005-08-14 19:27:13 -0700437 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700438#ifdef CONFIG_KMOD
Patrick McHardyab33a172005-08-14 19:31:36 -0700439 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -0700440 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700441 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -0700442 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700443 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700444#endif
445 if (nl_table[protocol].registered &&
446 try_module_get(nl_table[protocol].module))
447 module = nl_table[protocol].module;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700448 cb_mutex = nl_table[protocol].cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700449 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700450
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800451 err = __netlink_create(net, sock, cb_mutex, protocol);
452 if (err < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -0700453 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700455 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700456 nlk->module = module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700457out:
458 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
Patrick McHardyab33a172005-08-14 19:31:36 -0700460out_module:
461 module_put(module);
462 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463}
464
465static int netlink_release(struct socket *sock)
466{
467 struct sock *sk = sock->sk;
468 struct netlink_sock *nlk;
469
470 if (!sk)
471 return 0;
472
473 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -0700474 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 nlk = nlk_sk(sk);
476
Herbert Xu3f660d62007-05-03 03:17:14 -0700477 /*
478 * OK. Socket is unlinked, any packets that arrive now
479 * will be purged.
480 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 sock->sk = NULL;
483 wake_up_interruptible_all(&nlk->wait);
484
485 skb_queue_purge(&sk->sk_write_queue);
486
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700487 if (nlk->pid && !nlk->subscriptions) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 struct netlink_notify n = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200489 .net = sk->sk_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 .protocol = sk->sk_protocol,
491 .pid = nlk->pid,
492 };
Alan Sterne041c682006-03-27 01:16:30 -0800493 atomic_notifier_call_chain(&netlink_chain,
494 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900495 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700496
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -0800497 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700498
Patrick McHardy4277a082006-03-20 18:52:01 -0800499 netlink_table_grab();
Denis V. Lunevaed81562007-10-10 21:14:32 -0700500 if (netlink_is_kernel(sk)) {
Denis V. Lunev869e58f2008-01-18 23:53:31 -0800501 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
502 if (--nl_table[sk->sk_protocol].registered == 0) {
503 kfree(nl_table[sk->sk_protocol].listeners);
504 nl_table[sk->sk_protocol].module = NULL;
505 nl_table[sk->sk_protocol].registered = 0;
506 }
Patrick McHardy4277a082006-03-20 18:52:01 -0800507 } else if (nlk->subscriptions)
508 netlink_update_listeners(sk);
509 netlink_table_ungrab();
Patrick McHardy77247bb2005-08-14 19:27:13 -0700510
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700511 kfree(nlk->groups);
512 nlk->groups = NULL;
513
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 sock_put(sk);
515 return 0;
516}
517
518static int netlink_autobind(struct socket *sock)
519{
520 struct sock *sk = sock->sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200521 struct net *net = sk->sk_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
523 struct hlist_head *head;
524 struct sock *osk;
525 struct hlist_node *node;
Herbert Xuc27bd492005-11-22 14:41:50 -0800526 s32 pid = current->tgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 int err;
528 static s32 rover = -4097;
529
530retry:
531 cond_resched();
532 netlink_table_grab();
533 head = nl_pid_hashfn(hash, pid);
534 sk_for_each(osk, node, head) {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200535 if ((osk->sk_net != net))
536 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 if (nlk_sk(osk)->pid == pid) {
538 /* Bind collision, search negative pid values. */
539 pid = rover--;
540 if (rover > -4097)
541 rover = -4097;
542 netlink_table_ungrab();
543 goto retry;
544 }
545 }
546 netlink_table_ungrab();
547
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200548 err = netlink_insert(sk, net, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 if (err == -EADDRINUSE)
550 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -0700551
552 /* If 2 threads race to autobind, that is fine. */
553 if (err == -EBUSY)
554 err = 0;
555
556 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557}
558
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900559static inline int netlink_capable(struct socket *sock, unsigned int flag)
560{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
562 capable(CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900563}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700565static void
566netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
567{
568 struct netlink_sock *nlk = nlk_sk(sk);
569
570 if (nlk->subscriptions && !subscriptions)
571 __sk_del_bind_node(sk);
572 else if (!nlk->subscriptions && subscriptions)
573 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
574 nlk->subscriptions = subscriptions;
575}
576
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700577static int netlink_realloc_groups(struct sock *sk)
Patrick McHardy513c2502005-09-06 15:43:59 -0700578{
579 struct netlink_sock *nlk = nlk_sk(sk);
580 unsigned int groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700581 unsigned long *new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700582 int err = 0;
583
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700584 netlink_table_grab();
585
Patrick McHardy513c2502005-09-06 15:43:59 -0700586 groups = nl_table[sk->sk_protocol].groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700587 if (!nl_table[sk->sk_protocol].registered) {
Patrick McHardy513c2502005-09-06 15:43:59 -0700588 err = -ENOENT;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700589 goto out_unlock;
590 }
Patrick McHardy513c2502005-09-06 15:43:59 -0700591
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700592 if (nlk->ngroups >= groups)
593 goto out_unlock;
Patrick McHardy513c2502005-09-06 15:43:59 -0700594
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700595 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
596 if (new_groups == NULL) {
597 err = -ENOMEM;
598 goto out_unlock;
599 }
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800600 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700601 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
602
603 nlk->groups = new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700604 nlk->ngroups = groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700605 out_unlock:
606 netlink_table_ungrab();
607 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700608}
609
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800610static int netlink_bind(struct socket *sock, struct sockaddr *addr,
611 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612{
613 struct sock *sk = sock->sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200614 struct net *net = sk->sk_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 struct netlink_sock *nlk = nlk_sk(sk);
616 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
617 int err;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900618
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 if (nladdr->nl_family != AF_NETLINK)
620 return -EINVAL;
621
622 /* Only superuser is allowed to listen multicasts */
Patrick McHardy513c2502005-09-06 15:43:59 -0700623 if (nladdr->nl_groups) {
624 if (!netlink_capable(sock, NL_NONROOT_RECV))
625 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700626 err = netlink_realloc_groups(sk);
627 if (err)
628 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700629 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
631 if (nlk->pid) {
632 if (nladdr->nl_pid != nlk->pid)
633 return -EINVAL;
634 } else {
635 err = nladdr->nl_pid ?
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200636 netlink_insert(sk, net, nladdr->nl_pid) :
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 netlink_autobind(sock);
638 if (err)
639 return err;
640 }
641
Patrick McHardy513c2502005-09-06 15:43:59 -0700642 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 return 0;
644
645 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700646 netlink_update_subscriptions(sk, nlk->subscriptions +
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900647 hweight32(nladdr->nl_groups) -
648 hweight32(nlk->groups[0]));
649 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
Patrick McHardy4277a082006-03-20 18:52:01 -0800650 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 netlink_table_ungrab();
652
653 return 0;
654}
655
656static int netlink_connect(struct socket *sock, struct sockaddr *addr,
657 int alen, int flags)
658{
659 int err = 0;
660 struct sock *sk = sock->sk;
661 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800662 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
664 if (addr->sa_family == AF_UNSPEC) {
665 sk->sk_state = NETLINK_UNCONNECTED;
666 nlk->dst_pid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -0700667 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 return 0;
669 }
670 if (addr->sa_family != AF_NETLINK)
671 return -EINVAL;
672
673 /* Only superuser is allowed to send multicasts */
674 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
675 return -EPERM;
676
677 if (!nlk->pid)
678 err = netlink_autobind(sock);
679
680 if (err == 0) {
681 sk->sk_state = NETLINK_CONNECTED;
682 nlk->dst_pid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700683 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 }
685
686 return err;
687}
688
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800689static int netlink_getname(struct socket *sock, struct sockaddr *addr,
690 int *addr_len, int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691{
692 struct sock *sk = sock->sk;
693 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800694 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900695
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 nladdr->nl_family = AF_NETLINK;
697 nladdr->nl_pad = 0;
698 *addr_len = sizeof(*nladdr);
699
700 if (peer) {
701 nladdr->nl_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700702 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 } else {
704 nladdr->nl_pid = nlk->pid;
Patrick McHardy513c2502005-09-06 15:43:59 -0700705 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 }
707 return 0;
708}
709
710static void netlink_overrun(struct sock *sk)
711{
712 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
713 sk->sk_err = ENOBUFS;
714 sk->sk_error_report(sk);
715 }
716}
717
718static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
719{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 struct sock *sock;
721 struct netlink_sock *nlk;
722
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700723 sock = netlink_lookup(ssk->sk_net, ssk->sk_protocol, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (!sock)
725 return ERR_PTR(-ECONNREFUSED);
726
727 /* Don't bother queuing skb if kernel socket has no input function */
728 nlk = nlk_sk(sock);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700729 if (sock->sk_state == NETLINK_CONNECTED &&
730 nlk->dst_pid != nlk_sk(ssk)->pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 sock_put(sock);
732 return ERR_PTR(-ECONNREFUSED);
733 }
734 return sock;
735}
736
737struct sock *netlink_getsockbyfilp(struct file *filp)
738{
Josef Sipek6db5fc52006-12-08 02:37:25 -0800739 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 struct sock *sock;
741
742 if (!S_ISSOCK(inode->i_mode))
743 return ERR_PTR(-ENOTSOCK);
744
745 sock = SOCKET_I(inode)->sk;
746 if (sock->sk_family != AF_NETLINK)
747 return ERR_PTR(-EINVAL);
748
749 sock_hold(sock);
750 return sock;
751}
752
753/*
754 * Attach a skb to a netlink socket.
755 * The caller must hold a reference to the destination socket. On error, the
756 * reference is dropped. The skb is not send to the destination, just all
757 * all error checks are performed and memory in the queue is reserved.
758 * Return values:
759 * < 0: error. skb freed, reference to sock dropped.
760 * 0: continue
761 * 1: repeat lookup - reference dropped while waiting for socket memory.
762 */
Alexey Kuznetsova70ea992006-02-09 16:40:11 -0800763int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock,
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800764 long *timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765{
766 struct netlink_sock *nlk;
767
768 nlk = nlk_sk(sk);
769
770 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
771 test_bit(0, &nlk->state)) {
772 DECLARE_WAITQUEUE(wait, current);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800773 if (!*timeo) {
Denis V. Lunevaed81562007-10-10 21:14:32 -0700774 if (!ssk || netlink_is_kernel(ssk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 netlink_overrun(sk);
776 sock_put(sk);
777 kfree_skb(skb);
778 return -EAGAIN;
779 }
780
781 __set_current_state(TASK_INTERRUPTIBLE);
782 add_wait_queue(&nlk->wait, &wait);
783
784 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
785 test_bit(0, &nlk->state)) &&
786 !sock_flag(sk, SOCK_DEAD))
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800787 *timeo = schedule_timeout(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789 __set_current_state(TASK_RUNNING);
790 remove_wait_queue(&nlk->wait, &wait);
791 sock_put(sk);
792
793 if (signal_pending(current)) {
794 kfree_skb(skb);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800795 return sock_intr_errno(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 }
797 return 1;
798 }
799 skb_set_owner_r(skb, sk);
800 return 0;
801}
802
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700803int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 int len = skb->len;
806
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 skb_queue_tail(&sk->sk_receive_queue, skb);
808 sk->sk_data_ready(sk, len);
809 sock_put(sk);
810 return len;
811}
812
813void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
814{
815 kfree_skb(skb);
816 sock_put(sk);
817}
818
Victor Fusco37da6472005-07-18 13:35:43 -0700819static inline struct sk_buff *netlink_trim(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100820 gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821{
822 int delta;
823
824 skb_orphan(skb);
825
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700826 delta = skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 if (delta * 2 < skb->truesize)
828 return skb;
829
830 if (skb_shared(skb)) {
831 struct sk_buff *nskb = skb_clone(skb, allocation);
832 if (!nskb)
833 return skb;
834 kfree_skb(skb);
835 skb = nskb;
836 }
837
838 if (!pskb_expand_head(skb, 0, -delta, allocation))
839 skb->truesize -= delta;
840
841 return skb;
842}
843
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700844static inline void netlink_rcv_wake(struct sock *sk)
845{
846 struct netlink_sock *nlk = nlk_sk(sk);
847
848 if (skb_queue_empty(&sk->sk_receive_queue))
849 clear_bit(0, &nlk->state);
850 if (!test_bit(0, &nlk->state))
851 wake_up_interruptible(&nlk->wait);
852}
853
854static inline int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb)
855{
856 int ret;
857 struct netlink_sock *nlk = nlk_sk(sk);
858
859 ret = -ECONNREFUSED;
860 if (nlk->netlink_rcv != NULL) {
861 ret = skb->len;
862 skb_set_owner_r(skb, sk);
863 nlk->netlink_rcv(skb);
864 }
865 kfree_skb(skb);
866 sock_put(sk);
867 return ret;
868}
869
870int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
871 u32 pid, int nonblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
873 struct sock *sk;
874 int err;
875 long timeo;
876
877 skb = netlink_trim(skb, gfp_any());
878
879 timeo = sock_sndtimeo(ssk, nonblock);
880retry:
881 sk = netlink_getsockbypid(ssk, pid);
882 if (IS_ERR(sk)) {
883 kfree_skb(skb);
884 return PTR_ERR(sk);
885 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700886 if (netlink_is_kernel(sk))
887 return netlink_unicast_kernel(sk, skb);
888
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800889 err = netlink_attachskb(sk, skb, nonblock, &timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 if (err == 1)
891 goto retry;
892 if (err)
893 return err;
894
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700895 return netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896}
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800897EXPORT_SYMBOL(netlink_unicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
Patrick McHardy4277a082006-03-20 18:52:01 -0800899int netlink_has_listeners(struct sock *sk, unsigned int group)
900{
901 int res = 0;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700902 unsigned long *listeners;
Patrick McHardy4277a082006-03-20 18:52:01 -0800903
Denis V. Lunevaed81562007-10-10 21:14:32 -0700904 BUG_ON(!netlink_is_kernel(sk));
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700905
906 rcu_read_lock();
907 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
908
Patrick McHardy4277a082006-03-20 18:52:01 -0800909 if (group - 1 < nl_table[sk->sk_protocol].groups)
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700910 res = test_bit(group - 1, listeners);
911
912 rcu_read_unlock();
913
Patrick McHardy4277a082006-03-20 18:52:01 -0800914 return res;
915}
916EXPORT_SYMBOL_GPL(netlink_has_listeners);
917
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800918static inline int netlink_broadcast_deliver(struct sock *sk,
919 struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
921 struct netlink_sock *nlk = nlk_sk(sk);
922
923 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
924 !test_bit(0, &nlk->state)) {
925 skb_set_owner_r(skb, sk);
926 skb_queue_tail(&sk->sk_receive_queue, skb);
927 sk->sk_data_ready(sk, skb->len);
928 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf;
929 }
930 return -1;
931}
932
933struct netlink_broadcast_data {
934 struct sock *exclude_sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200935 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 u32 pid;
937 u32 group;
938 int failure;
939 int congested;
940 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -0400941 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 struct sk_buff *skb, *skb2;
943};
944
945static inline int do_one_broadcast(struct sock *sk,
946 struct netlink_broadcast_data *p)
947{
948 struct netlink_sock *nlk = nlk_sk(sk);
949 int val;
950
951 if (p->exclude_sk == sk)
952 goto out;
953
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700954 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
955 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 goto out;
957
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200958 if ((sk->sk_net != p->net))
959 goto out;
960
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 if (p->failure) {
962 netlink_overrun(sk);
963 goto out;
964 }
965
966 sock_hold(sk);
967 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -0700968 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 p->skb2 = skb_clone(p->skb, p->allocation);
970 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -0700971 p->skb2 = skb_get(p->skb);
972 /*
973 * skb ownership may have been set when
974 * delivered to a previous socket.
975 */
976 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 }
978 }
979 if (p->skb2 == NULL) {
980 netlink_overrun(sk);
981 /* Clone failed. Notify ALL listeners. */
982 p->failure = 1;
983 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
984 netlink_overrun(sk);
985 } else {
986 p->congested |= val;
987 p->delivered = 1;
988 p->skb2 = NULL;
989 }
990 sock_put(sk);
991
992out:
993 return 0;
994}
995
996int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
Al Virodd0fc662005-10-07 07:46:04 +0100997 u32 group, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998{
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200999 struct net *net = ssk->sk_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 struct netlink_broadcast_data info;
1001 struct hlist_node *node;
1002 struct sock *sk;
1003
1004 skb = netlink_trim(skb, allocation);
1005
1006 info.exclude_sk = ssk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001007 info.net = net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 info.pid = pid;
1009 info.group = group;
1010 info.failure = 0;
1011 info.congested = 0;
1012 info.delivered = 0;
1013 info.allocation = allocation;
1014 info.skb = skb;
1015 info.skb2 = NULL;
1016
1017 /* While we sleep in clone, do not allow to change socket list */
1018
1019 netlink_lock_table();
1020
1021 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1022 do_one_broadcast(sk, &info);
1023
Tommy S. Christensenaa1c6a62005-05-19 13:07:32 -07001024 kfree_skb(skb);
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 netlink_unlock_table();
1027
1028 if (info.skb2)
1029 kfree_skb(info.skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031 if (info.delivered) {
1032 if (info.congested && (allocation & __GFP_WAIT))
1033 yield();
1034 return 0;
1035 }
1036 if (info.failure)
1037 return -ENOBUFS;
1038 return -ESRCH;
1039}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001040EXPORT_SYMBOL(netlink_broadcast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041
1042struct netlink_set_err_data {
1043 struct sock *exclude_sk;
1044 u32 pid;
1045 u32 group;
1046 int code;
1047};
1048
1049static inline int do_one_set_err(struct sock *sk,
1050 struct netlink_set_err_data *p)
1051{
1052 struct netlink_sock *nlk = nlk_sk(sk);
1053
1054 if (sk == p->exclude_sk)
1055 goto out;
1056
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001057 if (sk->sk_net != p->exclude_sk->sk_net)
1058 goto out;
1059
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001060 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
1061 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 goto out;
1063
1064 sk->sk_err = p->code;
1065 sk->sk_error_report(sk);
1066out:
1067 return 0;
1068}
1069
1070void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
1071{
1072 struct netlink_set_err_data info;
1073 struct hlist_node *node;
1074 struct sock *sk;
1075
1076 info.exclude_sk = ssk;
1077 info.pid = pid;
1078 info.group = group;
1079 info.code = code;
1080
1081 read_lock(&nl_table_lock);
1082
1083 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1084 do_one_set_err(sk, &info);
1085
1086 read_unlock(&nl_table_lock);
1087}
1088
Johannes Berg84659eb2007-07-18 15:47:05 -07001089/* must be called with netlink table grabbed */
1090static void netlink_update_socket_mc(struct netlink_sock *nlk,
1091 unsigned int group,
1092 int is_new)
1093{
1094 int old, new = !!is_new, subscriptions;
1095
1096 old = test_bit(group - 1, nlk->groups);
1097 subscriptions = nlk->subscriptions - old + new;
1098 if (new)
1099 __set_bit(group - 1, nlk->groups);
1100 else
1101 __clear_bit(group - 1, nlk->groups);
1102 netlink_update_subscriptions(&nlk->sk, subscriptions);
1103 netlink_update_listeners(&nlk->sk);
1104}
1105
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001106static int netlink_setsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001107 char __user *optval, int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001108{
1109 struct sock *sk = sock->sk;
1110 struct netlink_sock *nlk = nlk_sk(sk);
Johannes Bergeb496532007-07-18 02:07:51 -07001111 unsigned int val = 0;
1112 int err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001113
1114 if (level != SOL_NETLINK)
1115 return -ENOPROTOOPT;
1116
1117 if (optlen >= sizeof(int) &&
Johannes Bergeb496532007-07-18 02:07:51 -07001118 get_user(val, (unsigned int __user *)optval))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001119 return -EFAULT;
1120
1121 switch (optname) {
1122 case NETLINK_PKTINFO:
1123 if (val)
1124 nlk->flags |= NETLINK_RECV_PKTINFO;
1125 else
1126 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1127 err = 0;
1128 break;
1129 case NETLINK_ADD_MEMBERSHIP:
1130 case NETLINK_DROP_MEMBERSHIP: {
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001131 if (!netlink_capable(sock, NL_NONROOT_RECV))
1132 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001133 err = netlink_realloc_groups(sk);
1134 if (err)
1135 return err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001136 if (!val || val - 1 >= nlk->ngroups)
1137 return -EINVAL;
1138 netlink_table_grab();
Johannes Berg84659eb2007-07-18 15:47:05 -07001139 netlink_update_socket_mc(nlk, val,
1140 optname == NETLINK_ADD_MEMBERSHIP);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001141 netlink_table_ungrab();
1142 err = 0;
1143 break;
1144 }
1145 default:
1146 err = -ENOPROTOOPT;
1147 }
1148 return err;
1149}
1150
1151static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001152 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001153{
1154 struct sock *sk = sock->sk;
1155 struct netlink_sock *nlk = nlk_sk(sk);
1156 int len, val, err;
1157
1158 if (level != SOL_NETLINK)
1159 return -ENOPROTOOPT;
1160
1161 if (get_user(len, optlen))
1162 return -EFAULT;
1163 if (len < 0)
1164 return -EINVAL;
1165
1166 switch (optname) {
1167 case NETLINK_PKTINFO:
1168 if (len < sizeof(int))
1169 return -EINVAL;
1170 len = sizeof(int);
1171 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08001172 if (put_user(len, optlen) ||
1173 put_user(val, optval))
1174 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001175 err = 0;
1176 break;
1177 default:
1178 err = -ENOPROTOOPT;
1179 }
1180 return err;
1181}
1182
1183static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1184{
1185 struct nl_pktinfo info;
1186
1187 info.group = NETLINK_CB(skb).dst_group;
1188 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1189}
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1192 struct msghdr *msg, size_t len)
1193{
1194 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1195 struct sock *sk = sock->sk;
1196 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001197 struct sockaddr_nl *addr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 u32 dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001199 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 struct sk_buff *skb;
1201 int err;
1202 struct scm_cookie scm;
1203
1204 if (msg->msg_flags&MSG_OOB)
1205 return -EOPNOTSUPP;
1206
1207 if (NULL == siocb->scm)
1208 siocb->scm = &scm;
1209 err = scm_send(sock, msg, siocb->scm);
1210 if (err < 0)
1211 return err;
1212
1213 if (msg->msg_namelen) {
1214 if (addr->nl_family != AF_NETLINK)
1215 return -EINVAL;
1216 dst_pid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001217 dst_group = ffs(addr->nl_groups);
1218 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 return -EPERM;
1220 } else {
1221 dst_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001222 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 }
1224
1225 if (!nlk->pid) {
1226 err = netlink_autobind(sock);
1227 if (err)
1228 goto out;
1229 }
1230
1231 err = -EMSGSIZE;
1232 if (len > sk->sk_sndbuf - 32)
1233 goto out;
1234 err = -ENOBUFS;
Thomas Graf339bf982006-11-10 14:10:15 -08001235 skb = alloc_skb(len, GFP_KERNEL);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001236 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 goto out;
1238
1239 NETLINK_CB(skb).pid = nlk->pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001240 NETLINK_CB(skb).dst_group = dst_group;
Serge Hallync94c2572005-04-29 16:27:17 +01001241 NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context);
Steve Grubbe7c34972006-04-03 09:08:13 -04001242 selinux_get_task_sid(current, &(NETLINK_CB(skb).sid));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1244
1245 /* What can I do? Netlink is asynchronous, so that
1246 we will have to save current capabilities to
1247 check them, when this message will be delivered
1248 to corresponding kernel module. --ANK (980802)
1249 */
1250
1251 err = -EFAULT;
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001252 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 kfree_skb(skb);
1254 goto out;
1255 }
1256
1257 err = security_netlink_send(sk, skb);
1258 if (err) {
1259 kfree_skb(skb);
1260 goto out;
1261 }
1262
Patrick McHardyd629b832005-08-14 19:27:50 -07001263 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 atomic_inc(&skb->users);
Patrick McHardyd629b832005-08-14 19:27:50 -07001265 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 }
1267 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1268
1269out:
1270 return err;
1271}
1272
1273static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1274 struct msghdr *msg, size_t len,
1275 int flags)
1276{
1277 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1278 struct scm_cookie scm;
1279 struct sock *sk = sock->sk;
1280 struct netlink_sock *nlk = nlk_sk(sk);
1281 int noblock = flags&MSG_DONTWAIT;
1282 size_t copied;
1283 struct sk_buff *skb;
1284 int err;
1285
1286 if (flags&MSG_OOB)
1287 return -EOPNOTSUPP;
1288
1289 copied = 0;
1290
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001291 skb = skb_recv_datagram(sk, flags, noblock, &err);
1292 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 goto out;
1294
1295 msg->msg_namelen = 0;
1296
1297 copied = skb->len;
1298 if (len < copied) {
1299 msg->msg_flags |= MSG_TRUNC;
1300 copied = len;
1301 }
1302
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001303 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1305
1306 if (msg->msg_name) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001307 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 addr->nl_family = AF_NETLINK;
1309 addr->nl_pad = 0;
1310 addr->nl_pid = NETLINK_CB(skb).pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001311 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 msg->msg_namelen = sizeof(*addr);
1313 }
1314
Patrick McHardycc9a06c2006-03-12 20:34:27 -08001315 if (nlk->flags & NETLINK_RECV_PKTINFO)
1316 netlink_cmsg_recv_pktinfo(msg, skb);
1317
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 if (NULL == siocb->scm) {
1319 memset(&scm, 0, sizeof(scm));
1320 siocb->scm = &scm;
1321 }
1322 siocb->scm->creds = *NETLINK_CREDS(skb);
Patrick McHardy188ccb52007-05-03 03:27:01 -07001323 if (flags & MSG_TRUNC)
1324 copied = skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 skb_free_datagram(sk, skb);
1326
1327 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2)
1328 netlink_dump(sk);
1329
1330 scm_recv(sock, msg, siocb->scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331out:
1332 netlink_rcv_wake(sk);
1333 return err ? : copied;
1334}
1335
1336static void netlink_data_ready(struct sock *sk, int len)
1337{
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001338 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339}
1340
1341/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001342 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 * complete set of kernel non-blocking support for message
1344 * queueing.
1345 */
1346
1347struct sock *
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001348netlink_kernel_create(struct net *net, int unit, unsigned int groups,
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001349 void (*input)(struct sk_buff *skb),
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001350 struct mutex *cb_mutex, struct module *module)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351{
1352 struct socket *sock;
1353 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001354 struct netlink_sock *nlk;
Patrick McHardy4277a082006-03-20 18:52:01 -08001355 unsigned long *listeners = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001357 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001359 if (unit < 0 || unit >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 return NULL;
1361
1362 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1363 return NULL;
1364
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001365 if (__netlink_create(net, sock, cb_mutex, unit) < 0)
Patrick McHardy77247bb2005-08-14 19:27:13 -07001366 goto out_sock_release;
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001367
Patrick McHardy4277a082006-03-20 18:52:01 -08001368 if (groups < 32)
1369 groups = 32;
1370
1371 listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL);
1372 if (!listeners)
1373 goto out_sock_release;
1374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 sk = sock->sk;
1376 sk->sk_data_ready = netlink_data_ready;
1377 if (input)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001378 nlk_sk(sk)->netlink_rcv = input;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001380 if (netlink_insert(sk, net, 0))
Patrick McHardy77247bb2005-08-14 19:27:13 -07001381 goto out_sock_release;
1382
1383 nlk = nlk_sk(sk);
1384 nlk->flags |= NETLINK_KERNEL_SOCKET;
1385
1386 netlink_table_grab();
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001387 if (!nl_table[unit].registered) {
1388 nl_table[unit].groups = groups;
1389 nl_table[unit].listeners = listeners;
1390 nl_table[unit].cb_mutex = cb_mutex;
1391 nl_table[unit].module = module;
1392 nl_table[unit].registered = 1;
Jesper Juhlf937f1f462007-10-15 01:39:12 -07001393 } else {
1394 kfree(listeners);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001395 nl_table[unit].registered++;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001396 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07001397 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001398
Denis V. Lunev775516b2008-01-18 23:55:19 -08001399 /* Do not hold an extra referrence to a namespace as this socket is
1400 * internal to a namespace and does not prevent it to stop. */
1401 put_net(net);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001402 return sk;
1403
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001404out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08001405 kfree(listeners);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001406 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001407 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001409EXPORT_SYMBOL(netlink_kernel_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001411
1412void
1413netlink_kernel_release(struct sock *sk)
1414{
1415 if (sk == NULL || sk->sk_socket == NULL)
1416 return;
Denis V. Lunev775516b2008-01-18 23:55:19 -08001417
1418 /*
1419 * Last sock_put should drop referrence to sk->sk_net. It has already
1420 * been dropped in netlink_kernel_create. Taking referrence to stopping
1421 * namespace is not an option.
1422 * Take referrence to a socket to remove it from netlink lookup table
1423 * _alive_ and after that destroy it in the context of init_net.
1424 */
1425 sock_hold(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001426 sock_release(sk->sk_socket);
Denis V. Lunev775516b2008-01-18 23:55:19 -08001427
1428 sk->sk_net = get_net(&init_net);
1429 sock_put(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001430}
1431EXPORT_SYMBOL(netlink_kernel_release);
1432
1433
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001434/**
1435 * netlink_change_ngroups - change number of multicast groups
1436 *
1437 * This changes the number of multicast groups that are available
1438 * on a certain netlink family. Note that it is not possible to
Johannes Berg84659eb2007-07-18 15:47:05 -07001439 * change the number of groups to below 32. Also note that it does
1440 * not implicitly call netlink_clear_multicast_users() when the
1441 * number of groups is reduced.
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001442 *
1443 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1444 * @groups: The new number of groups.
1445 */
1446int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1447{
1448 unsigned long *listeners, *old = NULL;
1449 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
1450 int err = 0;
1451
1452 if (groups < 32)
1453 groups = 32;
1454
1455 netlink_table_grab();
1456 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
1457 listeners = kzalloc(NLGRPSZ(groups), GFP_ATOMIC);
1458 if (!listeners) {
1459 err = -ENOMEM;
1460 goto out_ungrab;
1461 }
1462 old = tbl->listeners;
1463 memcpy(listeners, old, NLGRPSZ(tbl->groups));
1464 rcu_assign_pointer(tbl->listeners, listeners);
1465 }
1466 tbl->groups = groups;
1467
1468 out_ungrab:
1469 netlink_table_ungrab();
1470 synchronize_rcu();
1471 kfree(old);
1472 return err;
1473}
1474EXPORT_SYMBOL(netlink_change_ngroups);
1475
Johannes Berg84659eb2007-07-18 15:47:05 -07001476/**
1477 * netlink_clear_multicast_users - kick off multicast listeners
1478 *
1479 * This function removes all listeners from the given group.
1480 * @ksk: The kernel netlink socket, as returned by
1481 * netlink_kernel_create().
1482 * @group: The multicast group to clear.
1483 */
1484void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1485{
1486 struct sock *sk;
1487 struct hlist_node *node;
1488 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1489
1490 netlink_table_grab();
1491
1492 sk_for_each_bound(sk, node, &tbl->mc_list)
1493 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1494
1495 netlink_table_ungrab();
1496}
1497EXPORT_SYMBOL(netlink_clear_multicast_users);
1498
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499void netlink_set_nonroot(int protocol, unsigned int flags)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001500{
1501 if ((unsigned int)protocol < MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 nl_table[protocol].nl_nonroot = flags;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001503}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001504EXPORT_SYMBOL(netlink_set_nonroot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505
1506static void netlink_destroy_callback(struct netlink_callback *cb)
1507{
1508 if (cb->skb)
1509 kfree_skb(cb->skb);
1510 kfree(cb);
1511}
1512
1513/*
1514 * It looks a bit ugly.
1515 * It would be better to create kernel thread.
1516 */
1517
1518static int netlink_dump(struct sock *sk)
1519{
1520 struct netlink_sock *nlk = nlk_sk(sk);
1521 struct netlink_callback *cb;
1522 struct sk_buff *skb;
1523 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001524 int len, err = -ENOBUFS;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001525
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL);
1527 if (!skb)
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001528 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001530 mutex_lock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
1532 cb = nlk->cb;
1533 if (cb == NULL) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001534 err = -EINVAL;
1535 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 }
1537
1538 len = cb->dump(skb, cb);
1539
1540 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001541 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 skb_queue_tail(&sk->sk_receive_queue, skb);
1543 sk->sk_data_ready(sk, len);
1544 return 0;
1545 }
1546
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001547 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1548 if (!nlh)
1549 goto errout_skb;
1550
1551 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1552
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553 skb_queue_tail(&sk->sk_receive_queue, skb);
1554 sk->sk_data_ready(sk, skb->len);
1555
Thomas Grafa8f74b22005-11-10 02:25:52 +01001556 if (cb->done)
1557 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 nlk->cb = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001559 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
1561 netlink_destroy_callback(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07001563
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001564errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001565 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001566 kfree_skb(skb);
1567errout:
1568 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569}
1570
1571int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1572 struct nlmsghdr *nlh,
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001573 int (*dump)(struct sk_buff *skb,
1574 struct netlink_callback *),
1575 int (*done)(struct netlink_callback *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576{
1577 struct netlink_callback *cb;
1578 struct sock *sk;
1579 struct netlink_sock *nlk;
1580
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001581 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (cb == NULL)
1583 return -ENOBUFS;
1584
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 cb->dump = dump;
1586 cb->done = done;
1587 cb->nlh = nlh;
1588 atomic_inc(&skb->users);
1589 cb->skb = skb;
1590
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001591 sk = netlink_lookup(ssk->sk_net, ssk->sk_protocol, NETLINK_CB(skb).pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 if (sk == NULL) {
1593 netlink_destroy_callback(cb);
1594 return -ECONNREFUSED;
1595 }
1596 nlk = nlk_sk(sk);
Herbert Xu3f660d62007-05-03 03:17:14 -07001597 /* A dump is in progress... */
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001598 mutex_lock(nlk->cb_mutex);
Herbert Xu3f660d62007-05-03 03:17:14 -07001599 if (nlk->cb) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001600 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 netlink_destroy_callback(cb);
1602 sock_put(sk);
1603 return -EBUSY;
1604 }
1605 nlk->cb = cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001606 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607
1608 netlink_dump(sk);
1609 sock_put(sk);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001610
1611 /* We successfully started a dump, by returning -EINTR we
1612 * signal not to send ACK even if it was requested.
1613 */
1614 return -EINTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001616EXPORT_SYMBOL(netlink_dump_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617
1618void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1619{
1620 struct sk_buff *skb;
1621 struct nlmsghdr *rep;
1622 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08001623 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624
Thomas Graf339bf982006-11-10 14:10:15 -08001625 /* error messages get the original request appened */
1626 if (err)
1627 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628
Thomas Graf339bf982006-11-10 14:10:15 -08001629 skb = nlmsg_new(payload, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 if (!skb) {
1631 struct sock *sk;
1632
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001633 sk = netlink_lookup(in_skb->sk->sk_net,
1634 in_skb->sk->sk_protocol,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 NETLINK_CB(in_skb).pid);
1636 if (sk) {
1637 sk->sk_err = ENOBUFS;
1638 sk->sk_error_report(sk);
1639 sock_put(sk);
1640 }
1641 return;
1642 }
1643
1644 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
Thomas Graf17977542005-06-18 22:53:48 -07001645 NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001646 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001648 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1650}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001651EXPORT_SYMBOL(netlink_ack);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001653int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001654 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01001655{
Thomas Graf82ace472005-11-10 02:25:53 +01001656 struct nlmsghdr *nlh;
1657 int err;
1658
1659 while (skb->len >= nlmsg_total_size(0)) {
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001660 int msglen;
1661
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001662 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07001663 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01001664
Martin Murrayad8e4b72006-01-10 13:02:29 -08001665 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01001666 return 0;
1667
Thomas Grafd35b6852007-03-22 23:28:46 -07001668 /* Only requests are handled by the kernel */
1669 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
Denis V. Lunev5c582982007-10-23 20:29:25 -07001670 goto ack;
Thomas Grafd35b6852007-03-22 23:28:46 -07001671
Thomas Graf45e7ae72007-03-22 23:29:10 -07001672 /* Skip control messages */
1673 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
Denis V. Lunev5c582982007-10-23 20:29:25 -07001674 goto ack;
Thomas Graf45e7ae72007-03-22 23:29:10 -07001675
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001676 err = cb(skb, nlh);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001677 if (err == -EINTR)
1678 goto skip;
1679
1680ack:
Thomas Grafd35b6852007-03-22 23:28:46 -07001681 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01001682 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01001683
Denis V. Lunev5c582982007-10-23 20:29:25 -07001684skip:
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001685 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001686 if (msglen > skb->len)
1687 msglen = skb->len;
1688 skb_pull(skb, msglen);
Thomas Graf82ace472005-11-10 02:25:53 +01001689 }
1690
1691 return 0;
1692}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001693EXPORT_SYMBOL(netlink_rcv_skb);
Thomas Graf82ace472005-11-10 02:25:53 +01001694
1695/**
Thomas Grafd387f6a2006-08-15 00:31:06 -07001696 * nlmsg_notify - send a notification netlink message
1697 * @sk: netlink socket to use
1698 * @skb: notification message
1699 * @pid: destination netlink pid for reports or 0
1700 * @group: destination multicast group or 0
1701 * @report: 1 to report back, 0 to disable
1702 * @flags: allocation flags
1703 */
1704int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid,
1705 unsigned int group, int report, gfp_t flags)
1706{
1707 int err = 0;
1708
1709 if (group) {
1710 int exclude_pid = 0;
1711
1712 if (report) {
1713 atomic_inc(&skb->users);
1714 exclude_pid = pid;
1715 }
1716
1717 /* errors reported via destination sk->sk_err */
1718 nlmsg_multicast(sk, skb, exclude_pid, group, flags);
1719 }
1720
1721 if (report)
1722 err = nlmsg_unicast(sk, skb, pid);
1723
1724 return err;
1725}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001726EXPORT_SYMBOL(nlmsg_notify);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001727
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728#ifdef CONFIG_PROC_FS
1729struct nl_seq_iter {
Denis V. Luneve372c412007-11-19 22:31:54 -08001730 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 int link;
1732 int hash_idx;
1733};
1734
1735static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1736{
1737 struct nl_seq_iter *iter = seq->private;
1738 int i, j;
1739 struct sock *s;
1740 struct hlist_node *node;
1741 loff_t off = 0;
1742
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001743 for (i = 0; i < MAX_LINKS; i++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 struct nl_pid_hash *hash = &nl_table[i].hash;
1745
1746 for (j = 0; j <= hash->mask; j++) {
1747 sk_for_each(s, node, &hash->table[j]) {
Denis V. Luneve372c412007-11-19 22:31:54 -08001748 if (iter->p.net != s->sk_net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001749 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (off == pos) {
1751 iter->link = i;
1752 iter->hash_idx = j;
1753 return s;
1754 }
1755 ++off;
1756 }
1757 }
1758 }
1759 return NULL;
1760}
1761
1762static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001763 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764{
1765 read_lock(&nl_table_lock);
1766 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1767}
1768
1769static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1770{
1771 struct sock *s;
1772 struct nl_seq_iter *iter;
1773 int i, j;
1774
1775 ++*pos;
1776
1777 if (v == SEQ_START_TOKEN)
1778 return netlink_seq_socket_idx(seq, 0);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001779
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001780 iter = seq->private;
1781 s = v;
1782 do {
1783 s = sk_next(s);
Denis V. Luneve372c412007-11-19 22:31:54 -08001784 } while (s && (iter->p.net != s->sk_net));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 if (s)
1786 return s;
1787
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 i = iter->link;
1789 j = iter->hash_idx + 1;
1790
1791 do {
1792 struct nl_pid_hash *hash = &nl_table[i].hash;
1793
1794 for (; j <= hash->mask; j++) {
1795 s = sk_head(&hash->table[j]);
Denis V. Luneve372c412007-11-19 22:31:54 -08001796 while (s && (iter->p.net != s->sk_net))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001797 s = sk_next(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 if (s) {
1799 iter->link = i;
1800 iter->hash_idx = j;
1801 return s;
1802 }
1803 }
1804
1805 j = 0;
1806 } while (++i < MAX_LINKS);
1807
1808 return NULL;
1809}
1810
1811static void netlink_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001812 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813{
1814 read_unlock(&nl_table_lock);
1815}
1816
1817
1818static int netlink_seq_show(struct seq_file *seq, void *v)
1819{
1820 if (v == SEQ_START_TOKEN)
1821 seq_puts(seq,
1822 "sk Eth Pid Groups "
1823 "Rmem Wmem Dump Locks\n");
1824 else {
1825 struct sock *s = v;
1826 struct netlink_sock *nlk = nlk_sk(s);
1827
1828 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1829 s,
1830 s->sk_protocol,
1831 nlk->pid,
Patrick McHardy513c2502005-09-06 15:43:59 -07001832 nlk->groups ? (u32)nlk->groups[0] : 0,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 atomic_read(&s->sk_rmem_alloc),
1834 atomic_read(&s->sk_wmem_alloc),
1835 nlk->cb,
1836 atomic_read(&s->sk_refcnt)
1837 );
1838
1839 }
1840 return 0;
1841}
1842
Philippe De Muyter56b3d972007-07-10 23:07:31 -07001843static const struct seq_operations netlink_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 .start = netlink_seq_start,
1845 .next = netlink_seq_next,
1846 .stop = netlink_seq_stop,
1847 .show = netlink_seq_show,
1848};
1849
1850
1851static int netlink_seq_open(struct inode *inode, struct file *file)
1852{
Denis V. Luneve372c412007-11-19 22:31:54 -08001853 return seq_open_net(inode, file, &netlink_seq_ops,
1854 sizeof(struct nl_seq_iter));
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001855}
1856
Arjan van de Venda7071d2007-02-12 00:55:36 -08001857static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 .owner = THIS_MODULE,
1859 .open = netlink_seq_open,
1860 .read = seq_read,
1861 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08001862 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863};
1864
1865#endif
1866
1867int netlink_register_notifier(struct notifier_block *nb)
1868{
Alan Sterne041c682006-03-27 01:16:30 -08001869 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001871EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872
1873int netlink_unregister_notifier(struct notifier_block *nb)
1874{
Alan Sterne041c682006-03-27 01:16:30 -08001875 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001877EXPORT_SYMBOL(netlink_unregister_notifier);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001878
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001879static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 .family = PF_NETLINK,
1881 .owner = THIS_MODULE,
1882 .release = netlink_release,
1883 .bind = netlink_bind,
1884 .connect = netlink_connect,
1885 .socketpair = sock_no_socketpair,
1886 .accept = sock_no_accept,
1887 .getname = netlink_getname,
1888 .poll = datagram_poll,
1889 .ioctl = sock_no_ioctl,
1890 .listen = sock_no_listen,
1891 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001892 .setsockopt = netlink_setsockopt,
1893 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 .sendmsg = netlink_sendmsg,
1895 .recvmsg = netlink_recvmsg,
1896 .mmap = sock_no_mmap,
1897 .sendpage = sock_no_sendpage,
1898};
1899
1900static struct net_proto_family netlink_family_ops = {
1901 .family = PF_NETLINK,
1902 .create = netlink_create,
1903 .owner = THIS_MODULE, /* for consistency 8) */
1904};
1905
Pavel Emelyanov46650792007-10-08 20:38:39 -07001906static int __net_init netlink_net_init(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001907{
1908#ifdef CONFIG_PROC_FS
1909 if (!proc_net_fops_create(net, "netlink", 0, &netlink_seq_fops))
1910 return -ENOMEM;
1911#endif
1912 return 0;
1913}
1914
Pavel Emelyanov46650792007-10-08 20:38:39 -07001915static void __net_exit netlink_net_exit(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001916{
1917#ifdef CONFIG_PROC_FS
1918 proc_net_remove(net, "netlink");
1919#endif
1920}
1921
Denis V. Lunev022cbae2007-11-13 03:23:50 -08001922static struct pernet_operations __net_initdata netlink_net_ops = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001923 .init = netlink_net_init,
1924 .exit = netlink_net_exit,
1925};
1926
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927static int __init netlink_proto_init(void)
1928{
1929 struct sk_buff *dummy_skb;
1930 int i;
Denis Cheng26ff5dd2007-09-16 16:36:02 -07001931 unsigned long limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 unsigned int order;
1933 int err = proto_register(&netlink_proto, 0);
1934
1935 if (err != 0)
1936 goto out;
1937
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001938 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001940 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001941 if (!nl_table)
1942 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 if (num_physpages >= (128 * 1024))
Denis Cheng26ff5dd2007-09-16 16:36:02 -07001945 limit = num_physpages >> (21 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 else
Denis Cheng26ff5dd2007-09-16 16:36:02 -07001947 limit = num_physpages >> (23 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948
Denis Cheng26ff5dd2007-09-16 16:36:02 -07001949 order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
1950 limit = (1UL << order) / sizeof(struct hlist_head);
1951 order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952
1953 for (i = 0; i < MAX_LINKS; i++) {
1954 struct nl_pid_hash *hash = &nl_table[i].hash;
1955
Eric Dumazetea729122007-12-11 02:09:47 -08001956 hash->table = nl_pid_hash_zalloc(1 * sizeof(*hash->table));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 if (!hash->table) {
1958 while (i-- > 0)
1959 nl_pid_hash_free(nl_table[i].hash.table,
1960 1 * sizeof(*hash->table));
1961 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001962 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 hash->max_shift = order;
1965 hash->shift = 0;
1966 hash->mask = 0;
1967 hash->rehash_time = jiffies;
1968 }
1969
1970 sock_register(&netlink_family_ops);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001971 register_pernet_subsys(&netlink_net_ops);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001972 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 rtnetlink_init();
1974out:
1975 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001976panic:
1977 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978}
1979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980core_initcall(netlink_proto_init);