blob: 42d2fb94eff1b6716f8a29bf760c3d58b20723ca [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>
48#include <linux/smp_lock.h>
49#include <linux/notifier.h>
50#include <linux/security.h>
51#include <linux/jhash.h>
52#include <linux/jiffies.h>
53#include <linux/random.h>
54#include <linux/bitops.h>
55#include <linux/mm.h>
56#include <linux/types.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010057#include <linux/audit.h>
Steve Grubbe7c34972006-04-03 09:08:13 -040058#include <linux/selinux.h>
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070059#include <linux/mutex.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010060
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)
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;
70 u32 pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 u32 dst_pid;
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;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 void (*data_ready)(struct sock *sk, int bytes);
Patrick McHardy77247bb2005-08-14 19:27:13 -070083 struct module *module;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084};
85
Patrick McHardy77247bb2005-08-14 19:27:13 -070086#define NETLINK_KERNEL_SOCKET 0x1
Patrick McHardy9a4595b2005-08-15 12:32:15 -070087#define NETLINK_RECV_PKTINFO 0x2
Patrick McHardy77247bb2005-08-14 19:27:13 -070088
Linus Torvalds1da177e2005-04-16 15:20:36 -070089static inline struct netlink_sock *nlk_sk(struct sock *sk)
90{
91 return (struct netlink_sock *)sk;
92}
93
94struct nl_pid_hash {
95 struct hlist_head *table;
96 unsigned long rehash_time;
97
98 unsigned int mask;
99 unsigned int shift;
100
101 unsigned int entries;
102 unsigned int max_shift;
103
104 u32 rnd;
105};
106
107struct netlink_table {
108 struct nl_pid_hash hash;
109 struct hlist_head mc_list;
Patrick McHardy4277a082006-03-20 18:52:01 -0800110 unsigned long *listeners;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 unsigned int nl_nonroot;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700112 unsigned int groups;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700113 struct mutex *cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700114 struct module *module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700115 int registered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116};
117
118static struct netlink_table *nl_table;
119
120static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
121
122static int netlink_dump(struct sock *sk);
123static void netlink_destroy_callback(struct netlink_callback *cb);
Adrian Bunk42bad1d2007-04-26 00:57:41 -0700124static void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126static DEFINE_RWLOCK(nl_table_lock);
127static atomic_t nl_table_users = ATOMIC_INIT(0);
128
Alan Sterne041c682006-03-27 01:16:30 -0800129static ATOMIC_NOTIFIER_HEAD(netlink_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Patrick McHardyd629b832005-08-14 19:27:50 -0700131static u32 netlink_group_mask(u32 group)
132{
133 return group ? 1 << (group - 1) : 0;
134}
135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
137{
138 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
139}
140
141static void netlink_sock_destruct(struct sock *sk)
142{
143 skb_queue_purge(&sk->sk_receive_queue);
144
145 if (!sock_flag(sk, SOCK_DEAD)) {
146 printk("Freeing alive netlink socket %p\n", sk);
147 return;
148 }
149 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
150 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
151 BUG_TRAP(!nlk_sk(sk)->cb);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700152 BUG_TRAP(!nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153}
154
155/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
156 * Look, when several writers sleep and reader wakes them up, all but one
157 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
158 * this, _but_ remember, it adds useless work on UP machines.
159 */
160
161static void netlink_table_grab(void)
162{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700163 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
165 if (atomic_read(&nl_table_users)) {
166 DECLARE_WAITQUEUE(wait, current);
167
168 add_wait_queue_exclusive(&nl_table_wait, &wait);
169 for(;;) {
170 set_current_state(TASK_UNINTERRUPTIBLE);
171 if (atomic_read(&nl_table_users) == 0)
172 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700173 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700175 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 }
177
178 __set_current_state(TASK_RUNNING);
179 remove_wait_queue(&nl_table_wait, &wait);
180 }
181}
182
183static __inline__ void netlink_table_ungrab(void)
184{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700185 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 wake_up(&nl_table_wait);
187}
188
189static __inline__ void
190netlink_lock_table(void)
191{
192 /* read_lock() synchronizes us to netlink_table_grab */
193
194 read_lock(&nl_table_lock);
195 atomic_inc(&nl_table_users);
196 read_unlock(&nl_table_lock);
197}
198
199static __inline__ void
200netlink_unlock_table(void)
201{
202 if (atomic_dec_and_test(&nl_table_users))
203 wake_up(&nl_table_wait);
204}
205
206static __inline__ struct sock *netlink_lookup(int protocol, u32 pid)
207{
208 struct nl_pid_hash *hash = &nl_table[protocol].hash;
209 struct hlist_head *head;
210 struct sock *sk;
211 struct hlist_node *node;
212
213 read_lock(&nl_table_lock);
214 head = nl_pid_hashfn(hash, pid);
215 sk_for_each(sk, node, head) {
216 if (nlk_sk(sk)->pid == pid) {
217 sock_hold(sk);
218 goto found;
219 }
220 }
221 sk = NULL;
222found:
223 read_unlock(&nl_table_lock);
224 return sk;
225}
226
227static inline struct hlist_head *nl_pid_hash_alloc(size_t size)
228{
229 if (size <= PAGE_SIZE)
230 return kmalloc(size, GFP_ATOMIC);
231 else
232 return (struct hlist_head *)
233 __get_free_pages(GFP_ATOMIC, get_order(size));
234}
235
236static inline void nl_pid_hash_free(struct hlist_head *table, size_t size)
237{
238 if (size <= PAGE_SIZE)
239 kfree(table);
240 else
241 free_pages((unsigned long)table, get_order(size));
242}
243
244static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
245{
246 unsigned int omask, mask, shift;
247 size_t osize, size;
248 struct hlist_head *otable, *table;
249 int i;
250
251 omask = mask = hash->mask;
252 osize = size = (mask + 1) * sizeof(*table);
253 shift = hash->shift;
254
255 if (grow) {
256 if (++shift > hash->max_shift)
257 return 0;
258 mask = mask * 2 + 1;
259 size *= 2;
260 }
261
262 table = nl_pid_hash_alloc(size);
263 if (!table)
264 return 0;
265
266 memset(table, 0, size);
267 otable = hash->table;
268 hash->table = table;
269 hash->mask = mask;
270 hash->shift = shift;
271 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
272
273 for (i = 0; i <= omask; i++) {
274 struct sock *sk;
275 struct hlist_node *node, *tmp;
276
277 sk_for_each_safe(sk, node, tmp, &otable[i])
278 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
279 }
280
281 nl_pid_hash_free(otable, osize);
282 hash->rehash_time = jiffies + 10 * 60 * HZ;
283 return 1;
284}
285
286static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
287{
288 int avg = hash->entries >> hash->shift;
289
290 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
291 return 1;
292
293 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
294 nl_pid_hash_rehash(hash, 0);
295 return 1;
296 }
297
298 return 0;
299}
300
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800301static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
Patrick McHardy4277a082006-03-20 18:52:01 -0800303static void
304netlink_update_listeners(struct sock *sk)
305{
306 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
307 struct hlist_node *node;
308 unsigned long mask;
309 unsigned int i;
310
311 for (i = 0; i < NLGRPSZ(tbl->groups)/sizeof(unsigned long); i++) {
312 mask = 0;
313 sk_for_each_bound(sk, node, &tbl->mc_list)
314 mask |= nlk_sk(sk)->groups[i];
315 tbl->listeners[i] = mask;
316 }
317 /* this function is only called with the netlink table "grabbed", which
318 * makes sure updates are visible before bind or setsockopt return. */
319}
320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321static int netlink_insert(struct sock *sk, u32 pid)
322{
323 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
324 struct hlist_head *head;
325 int err = -EADDRINUSE;
326 struct sock *osk;
327 struct hlist_node *node;
328 int len;
329
330 netlink_table_grab();
331 head = nl_pid_hashfn(hash, pid);
332 len = 0;
333 sk_for_each(osk, node, head) {
334 if (nlk_sk(osk)->pid == pid)
335 break;
336 len++;
337 }
338 if (node)
339 goto err;
340
341 err = -EBUSY;
342 if (nlk_sk(sk)->pid)
343 goto err;
344
345 err = -ENOMEM;
346 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
347 goto err;
348
349 if (len && nl_pid_hash_dilute(hash, len))
350 head = nl_pid_hashfn(hash, pid);
351 hash->entries++;
352 nlk_sk(sk)->pid = pid;
353 sk_add_node(sk, head);
354 err = 0;
355
356err:
357 netlink_table_ungrab();
358 return err;
359}
360
361static void netlink_remove(struct sock *sk)
362{
363 netlink_table_grab();
David S. Millerd470e3b2005-06-26 15:31:51 -0700364 if (sk_del_node_init(sk))
365 nl_table[sk->sk_protocol].hash.entries--;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700366 if (nlk_sk(sk)->subscriptions)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 __sk_del_bind_node(sk);
368 netlink_table_ungrab();
369}
370
371static struct proto netlink_proto = {
372 .name = "NETLINK",
373 .owner = THIS_MODULE,
374 .obj_size = sizeof(struct netlink_sock),
375};
376
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700377static int __netlink_create(struct socket *sock, struct mutex *cb_mutex,
378 int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 struct sock *sk;
381 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700382
383 sock->ops = &netlink_ops;
384
385 sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1);
386 if (!sk)
387 return -ENOMEM;
388
389 sock_init_data(sock, sk);
390
391 nlk = nlk_sk(sk);
Patrick McHardyffa4d722007-04-25 14:01:17 -0700392 if (cb_mutex)
393 nlk->cb_mutex = cb_mutex;
394 else {
395 nlk->cb_mutex = &nlk->cb_def_mutex;
396 mutex_init(nlk->cb_mutex);
397 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700398 init_waitqueue_head(&nlk->wait);
399
400 sk->sk_destruct = netlink_sock_destruct;
401 sk->sk_protocol = protocol;
402 return 0;
403}
404
405static int netlink_create(struct socket *sock, int protocol)
406{
407 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700408 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700409 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700410 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
412 sock->state = SS_UNCONNECTED;
413
414 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
415 return -ESOCKTNOSUPPORT;
416
417 if (protocol<0 || protocol >= MAX_LINKS)
418 return -EPROTONOSUPPORT;
419
Patrick McHardy77247bb2005-08-14 19:27:13 -0700420 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700421#ifdef CONFIG_KMOD
Patrick McHardyab33a172005-08-14 19:31:36 -0700422 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -0700423 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700424 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -0700425 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700426 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700427#endif
428 if (nl_table[protocol].registered &&
429 try_module_get(nl_table[protocol].module))
430 module = nl_table[protocol].module;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700431 cb_mutex = nl_table[protocol].cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700432 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700433
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700434 if ((err = __netlink_create(sock, cb_mutex, protocol)) < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -0700435 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700437 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700438 nlk->module = module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700439out:
440 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Patrick McHardyab33a172005-08-14 19:31:36 -0700442out_module:
443 module_put(module);
444 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445}
446
447static int netlink_release(struct socket *sock)
448{
449 struct sock *sk = sock->sk;
450 struct netlink_sock *nlk;
451
452 if (!sk)
453 return 0;
454
455 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -0700456 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 nlk = nlk_sk(sk);
458
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700459 mutex_lock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 if (nlk->cb) {
Thomas Grafa8f74b22005-11-10 02:25:52 +0100461 if (nlk->cb->done)
462 nlk->cb->done(nlk->cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 netlink_destroy_callback(nlk->cb);
464 nlk->cb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 }
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700466 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
468 /* OK. Socket is unlinked, and, therefore,
469 no new packets will arrive */
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 sock->sk = NULL;
472 wake_up_interruptible_all(&nlk->wait);
473
474 skb_queue_purge(&sk->sk_write_queue);
475
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700476 if (nlk->pid && !nlk->subscriptions) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 struct netlink_notify n = {
478 .protocol = sk->sk_protocol,
479 .pid = nlk->pid,
480 };
Alan Sterne041c682006-03-27 01:16:30 -0800481 atomic_notifier_call_chain(&netlink_chain,
482 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900483 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700484
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -0800485 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700486
Patrick McHardy4277a082006-03-20 18:52:01 -0800487 netlink_table_grab();
Patrick McHardy77247bb2005-08-14 19:27:13 -0700488 if (nlk->flags & NETLINK_KERNEL_SOCKET) {
Patrick McHardy4277a082006-03-20 18:52:01 -0800489 kfree(nl_table[sk->sk_protocol].listeners);
Patrick McHardy77247bb2005-08-14 19:27:13 -0700490 nl_table[sk->sk_protocol].module = NULL;
Patrick McHardyab33a172005-08-14 19:31:36 -0700491 nl_table[sk->sk_protocol].registered = 0;
Patrick McHardy4277a082006-03-20 18:52:01 -0800492 } else if (nlk->subscriptions)
493 netlink_update_listeners(sk);
494 netlink_table_ungrab();
Patrick McHardy77247bb2005-08-14 19:27:13 -0700495
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700496 kfree(nlk->groups);
497 nlk->groups = NULL;
498
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 sock_put(sk);
500 return 0;
501}
502
503static int netlink_autobind(struct socket *sock)
504{
505 struct sock *sk = sock->sk;
506 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
507 struct hlist_head *head;
508 struct sock *osk;
509 struct hlist_node *node;
Herbert Xuc27bd492005-11-22 14:41:50 -0800510 s32 pid = current->tgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 int err;
512 static s32 rover = -4097;
513
514retry:
515 cond_resched();
516 netlink_table_grab();
517 head = nl_pid_hashfn(hash, pid);
518 sk_for_each(osk, node, head) {
519 if (nlk_sk(osk)->pid == pid) {
520 /* Bind collision, search negative pid values. */
521 pid = rover--;
522 if (rover > -4097)
523 rover = -4097;
524 netlink_table_ungrab();
525 goto retry;
526 }
527 }
528 netlink_table_ungrab();
529
530 err = netlink_insert(sk, pid);
531 if (err == -EADDRINUSE)
532 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -0700533
534 /* If 2 threads race to autobind, that is fine. */
535 if (err == -EBUSY)
536 err = 0;
537
538 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539}
540
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900541static inline int netlink_capable(struct socket *sock, unsigned int flag)
542{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
544 capable(CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900545}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700547static void
548netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
549{
550 struct netlink_sock *nlk = nlk_sk(sk);
551
552 if (nlk->subscriptions && !subscriptions)
553 __sk_del_bind_node(sk);
554 else if (!nlk->subscriptions && subscriptions)
555 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
556 nlk->subscriptions = subscriptions;
557}
558
Patrick McHardy513c2502005-09-06 15:43:59 -0700559static int netlink_alloc_groups(struct sock *sk)
560{
561 struct netlink_sock *nlk = nlk_sk(sk);
562 unsigned int groups;
563 int err = 0;
564
565 netlink_lock_table();
566 groups = nl_table[sk->sk_protocol].groups;
567 if (!nl_table[sk->sk_protocol].registered)
568 err = -ENOENT;
569 netlink_unlock_table();
570
571 if (err)
572 return err;
573
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700574 nlk->groups = kzalloc(NLGRPSZ(groups), GFP_KERNEL);
Patrick McHardy513c2502005-09-06 15:43:59 -0700575 if (nlk->groups == NULL)
576 return -ENOMEM;
Patrick McHardy513c2502005-09-06 15:43:59 -0700577 nlk->ngroups = groups;
578 return 0;
579}
580
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
582{
583 struct sock *sk = sock->sk;
584 struct netlink_sock *nlk = nlk_sk(sk);
585 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
586 int err;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 if (nladdr->nl_family != AF_NETLINK)
589 return -EINVAL;
590
591 /* Only superuser is allowed to listen multicasts */
Patrick McHardy513c2502005-09-06 15:43:59 -0700592 if (nladdr->nl_groups) {
593 if (!netlink_capable(sock, NL_NONROOT_RECV))
594 return -EPERM;
595 if (nlk->groups == NULL) {
596 err = netlink_alloc_groups(sk);
597 if (err)
598 return err;
599 }
600 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
602 if (nlk->pid) {
603 if (nladdr->nl_pid != nlk->pid)
604 return -EINVAL;
605 } else {
606 err = nladdr->nl_pid ?
607 netlink_insert(sk, nladdr->nl_pid) :
608 netlink_autobind(sock);
609 if (err)
610 return err;
611 }
612
Patrick McHardy513c2502005-09-06 15:43:59 -0700613 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 return 0;
615
616 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700617 netlink_update_subscriptions(sk, nlk->subscriptions +
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900618 hweight32(nladdr->nl_groups) -
619 hweight32(nlk->groups[0]));
620 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
Patrick McHardy4277a082006-03-20 18:52:01 -0800621 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 netlink_table_ungrab();
623
624 return 0;
625}
626
627static int netlink_connect(struct socket *sock, struct sockaddr *addr,
628 int alen, int flags)
629{
630 int err = 0;
631 struct sock *sk = sock->sk;
632 struct netlink_sock *nlk = nlk_sk(sk);
633 struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr;
634
635 if (addr->sa_family == AF_UNSPEC) {
636 sk->sk_state = NETLINK_UNCONNECTED;
637 nlk->dst_pid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -0700638 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 return 0;
640 }
641 if (addr->sa_family != AF_NETLINK)
642 return -EINVAL;
643
644 /* Only superuser is allowed to send multicasts */
645 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
646 return -EPERM;
647
648 if (!nlk->pid)
649 err = netlink_autobind(sock);
650
651 if (err == 0) {
652 sk->sk_state = NETLINK_CONNECTED;
653 nlk->dst_pid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700654 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 }
656
657 return err;
658}
659
660static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer)
661{
662 struct sock *sk = sock->sk;
663 struct netlink_sock *nlk = nlk_sk(sk);
664 struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 nladdr->nl_family = AF_NETLINK;
667 nladdr->nl_pad = 0;
668 *addr_len = sizeof(*nladdr);
669
670 if (peer) {
671 nladdr->nl_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700672 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 } else {
674 nladdr->nl_pid = nlk->pid;
Patrick McHardy513c2502005-09-06 15:43:59 -0700675 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 }
677 return 0;
678}
679
680static void netlink_overrun(struct sock *sk)
681{
682 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
683 sk->sk_err = ENOBUFS;
684 sk->sk_error_report(sk);
685 }
686}
687
688static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
689{
690 int protocol = ssk->sk_protocol;
691 struct sock *sock;
692 struct netlink_sock *nlk;
693
694 sock = netlink_lookup(protocol, pid);
695 if (!sock)
696 return ERR_PTR(-ECONNREFUSED);
697
698 /* Don't bother queuing skb if kernel socket has no input function */
699 nlk = nlk_sk(sock);
700 if ((nlk->pid == 0 && !nlk->data_ready) ||
701 (sock->sk_state == NETLINK_CONNECTED &&
702 nlk->dst_pid != nlk_sk(ssk)->pid)) {
703 sock_put(sock);
704 return ERR_PTR(-ECONNREFUSED);
705 }
706 return sock;
707}
708
709struct sock *netlink_getsockbyfilp(struct file *filp)
710{
Josef Sipek6db5fc52006-12-08 02:37:25 -0800711 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 struct sock *sock;
713
714 if (!S_ISSOCK(inode->i_mode))
715 return ERR_PTR(-ENOTSOCK);
716
717 sock = SOCKET_I(inode)->sk;
718 if (sock->sk_family != AF_NETLINK)
719 return ERR_PTR(-EINVAL);
720
721 sock_hold(sock);
722 return sock;
723}
724
725/*
726 * Attach a skb to a netlink socket.
727 * The caller must hold a reference to the destination socket. On error, the
728 * reference is dropped. The skb is not send to the destination, just all
729 * all error checks are performed and memory in the queue is reserved.
730 * Return values:
731 * < 0: error. skb freed, reference to sock dropped.
732 * 0: continue
733 * 1: repeat lookup - reference dropped while waiting for socket memory.
734 */
Alexey Kuznetsova70ea992006-02-09 16:40:11 -0800735int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock,
736 long timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737{
738 struct netlink_sock *nlk;
739
740 nlk = nlk_sk(sk);
741
742 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
743 test_bit(0, &nlk->state)) {
744 DECLARE_WAITQUEUE(wait, current);
745 if (!timeo) {
Alexey Kuznetsova70ea992006-02-09 16:40:11 -0800746 if (!ssk || nlk_sk(ssk)->pid == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 netlink_overrun(sk);
748 sock_put(sk);
749 kfree_skb(skb);
750 return -EAGAIN;
751 }
752
753 __set_current_state(TASK_INTERRUPTIBLE);
754 add_wait_queue(&nlk->wait, &wait);
755
756 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
757 test_bit(0, &nlk->state)) &&
758 !sock_flag(sk, SOCK_DEAD))
759 timeo = schedule_timeout(timeo);
760
761 __set_current_state(TASK_RUNNING);
762 remove_wait_queue(&nlk->wait, &wait);
763 sock_put(sk);
764
765 if (signal_pending(current)) {
766 kfree_skb(skb);
767 return sock_intr_errno(timeo);
768 }
769 return 1;
770 }
771 skb_set_owner_r(skb, sk);
772 return 0;
773}
774
775int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol)
776{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 int len = skb->len;
778
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 skb_queue_tail(&sk->sk_receive_queue, skb);
780 sk->sk_data_ready(sk, len);
781 sock_put(sk);
782 return len;
783}
784
785void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
786{
787 kfree_skb(skb);
788 sock_put(sk);
789}
790
Victor Fusco37da6472005-07-18 13:35:43 -0700791static inline struct sk_buff *netlink_trim(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100792 gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793{
794 int delta;
795
796 skb_orphan(skb);
797
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700798 delta = skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 if (delta * 2 < skb->truesize)
800 return skb;
801
802 if (skb_shared(skb)) {
803 struct sk_buff *nskb = skb_clone(skb, allocation);
804 if (!nskb)
805 return skb;
806 kfree_skb(skb);
807 skb = nskb;
808 }
809
810 if (!pskb_expand_head(skb, 0, -delta, allocation))
811 skb->truesize -= delta;
812
813 return skb;
814}
815
816int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock)
817{
818 struct sock *sk;
819 int err;
820 long timeo;
821
822 skb = netlink_trim(skb, gfp_any());
823
824 timeo = sock_sndtimeo(ssk, nonblock);
825retry:
826 sk = netlink_getsockbypid(ssk, pid);
827 if (IS_ERR(sk)) {
828 kfree_skb(skb);
829 return PTR_ERR(sk);
830 }
Alexey Kuznetsova70ea992006-02-09 16:40:11 -0800831 err = netlink_attachskb(sk, skb, nonblock, timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (err == 1)
833 goto retry;
834 if (err)
835 return err;
836
837 return netlink_sendskb(sk, skb, ssk->sk_protocol);
838}
839
Patrick McHardy4277a082006-03-20 18:52:01 -0800840int netlink_has_listeners(struct sock *sk, unsigned int group)
841{
842 int res = 0;
843
844 BUG_ON(!(nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET));
845 if (group - 1 < nl_table[sk->sk_protocol].groups)
846 res = test_bit(group - 1, nl_table[sk->sk_protocol].listeners);
847 return res;
848}
849EXPORT_SYMBOL_GPL(netlink_has_listeners);
850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
852{
853 struct netlink_sock *nlk = nlk_sk(sk);
854
855 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
856 !test_bit(0, &nlk->state)) {
857 skb_set_owner_r(skb, sk);
858 skb_queue_tail(&sk->sk_receive_queue, skb);
859 sk->sk_data_ready(sk, skb->len);
860 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf;
861 }
862 return -1;
863}
864
865struct netlink_broadcast_data {
866 struct sock *exclude_sk;
867 u32 pid;
868 u32 group;
869 int failure;
870 int congested;
871 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -0400872 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 struct sk_buff *skb, *skb2;
874};
875
876static inline int do_one_broadcast(struct sock *sk,
877 struct netlink_broadcast_data *p)
878{
879 struct netlink_sock *nlk = nlk_sk(sk);
880 int val;
881
882 if (p->exclude_sk == sk)
883 goto out;
884
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700885 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
886 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 goto out;
888
889 if (p->failure) {
890 netlink_overrun(sk);
891 goto out;
892 }
893
894 sock_hold(sk);
895 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -0700896 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 p->skb2 = skb_clone(p->skb, p->allocation);
898 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -0700899 p->skb2 = skb_get(p->skb);
900 /*
901 * skb ownership may have been set when
902 * delivered to a previous socket.
903 */
904 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 }
906 }
907 if (p->skb2 == NULL) {
908 netlink_overrun(sk);
909 /* Clone failed. Notify ALL listeners. */
910 p->failure = 1;
911 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
912 netlink_overrun(sk);
913 } else {
914 p->congested |= val;
915 p->delivered = 1;
916 p->skb2 = NULL;
917 }
918 sock_put(sk);
919
920out:
921 return 0;
922}
923
924int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
Al Virodd0fc662005-10-07 07:46:04 +0100925 u32 group, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926{
927 struct netlink_broadcast_data info;
928 struct hlist_node *node;
929 struct sock *sk;
930
931 skb = netlink_trim(skb, allocation);
932
933 info.exclude_sk = ssk;
934 info.pid = pid;
935 info.group = group;
936 info.failure = 0;
937 info.congested = 0;
938 info.delivered = 0;
939 info.allocation = allocation;
940 info.skb = skb;
941 info.skb2 = NULL;
942
943 /* While we sleep in clone, do not allow to change socket list */
944
945 netlink_lock_table();
946
947 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
948 do_one_broadcast(sk, &info);
949
Tommy S. Christensenaa1c6a62005-05-19 13:07:32 -0700950 kfree_skb(skb);
951
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 netlink_unlock_table();
953
954 if (info.skb2)
955 kfree_skb(info.skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956
957 if (info.delivered) {
958 if (info.congested && (allocation & __GFP_WAIT))
959 yield();
960 return 0;
961 }
962 if (info.failure)
963 return -ENOBUFS;
964 return -ESRCH;
965}
966
967struct netlink_set_err_data {
968 struct sock *exclude_sk;
969 u32 pid;
970 u32 group;
971 int code;
972};
973
974static inline int do_one_set_err(struct sock *sk,
975 struct netlink_set_err_data *p)
976{
977 struct netlink_sock *nlk = nlk_sk(sk);
978
979 if (sk == p->exclude_sk)
980 goto out;
981
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700982 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
983 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 goto out;
985
986 sk->sk_err = p->code;
987 sk->sk_error_report(sk);
988out:
989 return 0;
990}
991
992void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
993{
994 struct netlink_set_err_data info;
995 struct hlist_node *node;
996 struct sock *sk;
997
998 info.exclude_sk = ssk;
999 info.pid = pid;
1000 info.group = group;
1001 info.code = code;
1002
1003 read_lock(&nl_table_lock);
1004
1005 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1006 do_one_set_err(sk, &info);
1007
1008 read_unlock(&nl_table_lock);
1009}
1010
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001011static int netlink_setsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001012 char __user *optval, int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001013{
1014 struct sock *sk = sock->sk;
1015 struct netlink_sock *nlk = nlk_sk(sk);
1016 int val = 0, err;
1017
1018 if (level != SOL_NETLINK)
1019 return -ENOPROTOOPT;
1020
1021 if (optlen >= sizeof(int) &&
1022 get_user(val, (int __user *)optval))
1023 return -EFAULT;
1024
1025 switch (optname) {
1026 case NETLINK_PKTINFO:
1027 if (val)
1028 nlk->flags |= NETLINK_RECV_PKTINFO;
1029 else
1030 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1031 err = 0;
1032 break;
1033 case NETLINK_ADD_MEMBERSHIP:
1034 case NETLINK_DROP_MEMBERSHIP: {
1035 unsigned int subscriptions;
1036 int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0;
1037
1038 if (!netlink_capable(sock, NL_NONROOT_RECV))
1039 return -EPERM;
Patrick McHardy513c2502005-09-06 15:43:59 -07001040 if (nlk->groups == NULL) {
1041 err = netlink_alloc_groups(sk);
1042 if (err)
1043 return err;
1044 }
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001045 if (!val || val - 1 >= nlk->ngroups)
1046 return -EINVAL;
1047 netlink_table_grab();
1048 old = test_bit(val - 1, nlk->groups);
1049 subscriptions = nlk->subscriptions - old + new;
1050 if (new)
1051 __set_bit(val - 1, nlk->groups);
1052 else
1053 __clear_bit(val - 1, nlk->groups);
1054 netlink_update_subscriptions(sk, subscriptions);
Patrick McHardy4277a082006-03-20 18:52:01 -08001055 netlink_update_listeners(sk);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001056 netlink_table_ungrab();
1057 err = 0;
1058 break;
1059 }
1060 default:
1061 err = -ENOPROTOOPT;
1062 }
1063 return err;
1064}
1065
1066static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001067 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001068{
1069 struct sock *sk = sock->sk;
1070 struct netlink_sock *nlk = nlk_sk(sk);
1071 int len, val, err;
1072
1073 if (level != SOL_NETLINK)
1074 return -ENOPROTOOPT;
1075
1076 if (get_user(len, optlen))
1077 return -EFAULT;
1078 if (len < 0)
1079 return -EINVAL;
1080
1081 switch (optname) {
1082 case NETLINK_PKTINFO:
1083 if (len < sizeof(int))
1084 return -EINVAL;
1085 len = sizeof(int);
1086 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08001087 if (put_user(len, optlen) ||
1088 put_user(val, optval))
1089 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001090 err = 0;
1091 break;
1092 default:
1093 err = -ENOPROTOOPT;
1094 }
1095 return err;
1096}
1097
1098static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1099{
1100 struct nl_pktinfo info;
1101
1102 info.group = NETLINK_CB(skb).dst_group;
1103 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1104}
1105
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106static inline void netlink_rcv_wake(struct sock *sk)
1107{
1108 struct netlink_sock *nlk = nlk_sk(sk);
1109
David S. Millerb03efcf2005-07-08 14:57:23 -07001110 if (skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 clear_bit(0, &nlk->state);
1112 if (!test_bit(0, &nlk->state))
1113 wake_up_interruptible(&nlk->wait);
1114}
1115
1116static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1117 struct msghdr *msg, size_t len)
1118{
1119 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1120 struct sock *sk = sock->sk;
1121 struct netlink_sock *nlk = nlk_sk(sk);
1122 struct sockaddr_nl *addr=msg->msg_name;
1123 u32 dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001124 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 struct sk_buff *skb;
1126 int err;
1127 struct scm_cookie scm;
1128
1129 if (msg->msg_flags&MSG_OOB)
1130 return -EOPNOTSUPP;
1131
1132 if (NULL == siocb->scm)
1133 siocb->scm = &scm;
1134 err = scm_send(sock, msg, siocb->scm);
1135 if (err < 0)
1136 return err;
1137
1138 if (msg->msg_namelen) {
1139 if (addr->nl_family != AF_NETLINK)
1140 return -EINVAL;
1141 dst_pid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001142 dst_group = ffs(addr->nl_groups);
1143 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 return -EPERM;
1145 } else {
1146 dst_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001147 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 }
1149
1150 if (!nlk->pid) {
1151 err = netlink_autobind(sock);
1152 if (err)
1153 goto out;
1154 }
1155
1156 err = -EMSGSIZE;
1157 if (len > sk->sk_sndbuf - 32)
1158 goto out;
1159 err = -ENOBUFS;
Thomas Graf339bf982006-11-10 14:10:15 -08001160 skb = alloc_skb(len, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 if (skb==NULL)
1162 goto out;
1163
1164 NETLINK_CB(skb).pid = nlk->pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001165 NETLINK_CB(skb).dst_group = dst_group;
Serge Hallync94c2572005-04-29 16:27:17 +01001166 NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context);
Steve Grubbe7c34972006-04-03 09:08:13 -04001167 selinux_get_task_sid(current, &(NETLINK_CB(skb).sid));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1169
1170 /* What can I do? Netlink is asynchronous, so that
1171 we will have to save current capabilities to
1172 check them, when this message will be delivered
1173 to corresponding kernel module. --ANK (980802)
1174 */
1175
1176 err = -EFAULT;
1177 if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) {
1178 kfree_skb(skb);
1179 goto out;
1180 }
1181
1182 err = security_netlink_send(sk, skb);
1183 if (err) {
1184 kfree_skb(skb);
1185 goto out;
1186 }
1187
Patrick McHardyd629b832005-08-14 19:27:50 -07001188 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 atomic_inc(&skb->users);
Patrick McHardyd629b832005-08-14 19:27:50 -07001190 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 }
1192 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1193
1194out:
1195 return err;
1196}
1197
1198static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1199 struct msghdr *msg, size_t len,
1200 int flags)
1201{
1202 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1203 struct scm_cookie scm;
1204 struct sock *sk = sock->sk;
1205 struct netlink_sock *nlk = nlk_sk(sk);
1206 int noblock = flags&MSG_DONTWAIT;
1207 size_t copied;
1208 struct sk_buff *skb;
1209 int err;
1210
1211 if (flags&MSG_OOB)
1212 return -EOPNOTSUPP;
1213
1214 copied = 0;
1215
1216 skb = skb_recv_datagram(sk,flags,noblock,&err);
1217 if (skb==NULL)
1218 goto out;
1219
1220 msg->msg_namelen = 0;
1221
1222 copied = skb->len;
1223 if (len < copied) {
1224 msg->msg_flags |= MSG_TRUNC;
1225 copied = len;
1226 }
1227
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001228 skb_reset_transport_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1230
1231 if (msg->msg_name) {
1232 struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name;
1233 addr->nl_family = AF_NETLINK;
1234 addr->nl_pad = 0;
1235 addr->nl_pid = NETLINK_CB(skb).pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001236 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 msg->msg_namelen = sizeof(*addr);
1238 }
1239
Patrick McHardycc9a06c2006-03-12 20:34:27 -08001240 if (nlk->flags & NETLINK_RECV_PKTINFO)
1241 netlink_cmsg_recv_pktinfo(msg, skb);
1242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 if (NULL == siocb->scm) {
1244 memset(&scm, 0, sizeof(scm));
1245 siocb->scm = &scm;
1246 }
1247 siocb->scm->creds = *NETLINK_CREDS(skb);
1248 skb_free_datagram(sk, skb);
1249
1250 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2)
1251 netlink_dump(sk);
1252
1253 scm_recv(sock, msg, siocb->scm, flags);
1254
David S. Millerb558ff72007-03-06 17:02:35 -08001255 if (flags & MSG_TRUNC)
1256 copied = skb->len;
1257
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258out:
1259 netlink_rcv_wake(sk);
1260 return err ? : copied;
1261}
1262
1263static void netlink_data_ready(struct sock *sk, int len)
1264{
1265 struct netlink_sock *nlk = nlk_sk(sk);
1266
1267 if (nlk->data_ready)
1268 nlk->data_ready(sk, len);
1269 netlink_rcv_wake(sk);
1270}
1271
1272/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001273 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 * complete set of kernel non-blocking support for message
1275 * queueing.
1276 */
1277
1278struct sock *
Patrick McHardy06628602005-08-15 12:33:26 -07001279netlink_kernel_create(int unit, unsigned int groups,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001280 void (*input)(struct sock *sk, int len),
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001281 struct mutex *cb_mutex, struct module *module)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282{
1283 struct socket *sock;
1284 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001285 struct netlink_sock *nlk;
Patrick McHardy4277a082006-03-20 18:52:01 -08001286 unsigned long *listeners = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001288 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
1290 if (unit<0 || unit>=MAX_LINKS)
1291 return NULL;
1292
1293 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1294 return NULL;
1295
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001296 if (__netlink_create(sock, cb_mutex, unit) < 0)
Patrick McHardy77247bb2005-08-14 19:27:13 -07001297 goto out_sock_release;
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001298
Patrick McHardy4277a082006-03-20 18:52:01 -08001299 if (groups < 32)
1300 groups = 32;
1301
1302 listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL);
1303 if (!listeners)
1304 goto out_sock_release;
1305
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 sk = sock->sk;
1307 sk->sk_data_ready = netlink_data_ready;
1308 if (input)
1309 nlk_sk(sk)->data_ready = input;
1310
Patrick McHardy77247bb2005-08-14 19:27:13 -07001311 if (netlink_insert(sk, 0))
1312 goto out_sock_release;
1313
1314 nlk = nlk_sk(sk);
1315 nlk->flags |= NETLINK_KERNEL_SOCKET;
1316
1317 netlink_table_grab();
Patrick McHardy4277a082006-03-20 18:52:01 -08001318 nl_table[unit].groups = groups;
1319 nl_table[unit].listeners = listeners;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001320 nl_table[unit].cb_mutex = cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001321 nl_table[unit].module = module;
Patrick McHardyab33a172005-08-14 19:31:36 -07001322 nl_table[unit].registered = 1;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001323 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001324
1325 return sk;
1326
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001327out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08001328 kfree(listeners);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001329 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001330 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331}
1332
1333void netlink_set_nonroot(int protocol, unsigned int flags)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001334{
1335 if ((unsigned int)protocol < MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 nl_table[protocol].nl_nonroot = flags;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001337}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
1339static void netlink_destroy_callback(struct netlink_callback *cb)
1340{
1341 if (cb->skb)
1342 kfree_skb(cb->skb);
1343 kfree(cb);
1344}
1345
1346/*
1347 * It looks a bit ugly.
1348 * It would be better to create kernel thread.
1349 */
1350
1351static int netlink_dump(struct sock *sk)
1352{
1353 struct netlink_sock *nlk = nlk_sk(sk);
1354 struct netlink_callback *cb;
1355 struct sk_buff *skb;
1356 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001357 int len, err = -ENOBUFS;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001358
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL);
1360 if (!skb)
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001361 goto errout;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001363 mutex_lock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
1365 cb = nlk->cb;
1366 if (cb == NULL) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001367 err = -EINVAL;
1368 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 }
1370
1371 len = cb->dump(skb, cb);
1372
1373 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001374 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 skb_queue_tail(&sk->sk_receive_queue, skb);
1376 sk->sk_data_ready(sk, len);
1377 return 0;
1378 }
1379
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001380 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1381 if (!nlh)
1382 goto errout_skb;
1383
1384 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1385
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 skb_queue_tail(&sk->sk_receive_queue, skb);
1387 sk->sk_data_ready(sk, skb->len);
1388
Thomas Grafa8f74b22005-11-10 02:25:52 +01001389 if (cb->done)
1390 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 nlk->cb = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001392 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
1394 netlink_destroy_callback(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07001396
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001397errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001398 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001399 kfree_skb(skb);
1400errout:
1401 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402}
1403
1404int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1405 struct nlmsghdr *nlh,
1406 int (*dump)(struct sk_buff *skb, struct netlink_callback*),
1407 int (*done)(struct netlink_callback*))
1408{
1409 struct netlink_callback *cb;
1410 struct sock *sk;
1411 struct netlink_sock *nlk;
1412
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001413 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 if (cb == NULL)
1415 return -ENOBUFS;
1416
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 cb->dump = dump;
1418 cb->done = done;
1419 cb->nlh = nlh;
1420 atomic_inc(&skb->users);
1421 cb->skb = skb;
1422
1423 sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid);
1424 if (sk == NULL) {
1425 netlink_destroy_callback(cb);
1426 return -ECONNREFUSED;
1427 }
1428 nlk = nlk_sk(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -07001429 /* A dump or destruction is in progress... */
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001430 mutex_lock(nlk->cb_mutex);
Denis Lunevac57b3a2007-04-18 17:05:58 -07001431 if (nlk->cb || sock_flag(sk, SOCK_DEAD)) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001432 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 netlink_destroy_callback(cb);
1434 sock_put(sk);
1435 return -EBUSY;
1436 }
1437 nlk->cb = cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001438 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439
1440 netlink_dump(sk);
1441 sock_put(sk);
Thomas Grafc702e802007-03-22 23:30:55 -07001442
1443 /* We successfully started a dump, by returning -EINTR we
1444 * signal the queue mangement to interrupt processing of
1445 * any netlink messages so userspace gets a chance to read
1446 * the results. */
1447 return -EINTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448}
1449
1450void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1451{
1452 struct sk_buff *skb;
1453 struct nlmsghdr *rep;
1454 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08001455 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
Thomas Graf339bf982006-11-10 14:10:15 -08001457 /* error messages get the original request appened */
1458 if (err)
1459 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
Thomas Graf339bf982006-11-10 14:10:15 -08001461 skb = nlmsg_new(payload, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (!skb) {
1463 struct sock *sk;
1464
1465 sk = netlink_lookup(in_skb->sk->sk_protocol,
1466 NETLINK_CB(in_skb).pid);
1467 if (sk) {
1468 sk->sk_err = ENOBUFS;
1469 sk->sk_error_report(sk);
1470 sock_put(sk);
1471 }
1472 return;
1473 }
1474
1475 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
Thomas Graf17977542005-06-18 22:53:48 -07001476 NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001477 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001479 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1481}
1482
Thomas Graf82ace472005-11-10 02:25:53 +01001483static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001484 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01001485{
Thomas Graf82ace472005-11-10 02:25:53 +01001486 struct nlmsghdr *nlh;
1487 int err;
1488
1489 while (skb->len >= nlmsg_total_size(0)) {
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001490 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07001491 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01001492
Martin Murrayad8e4b72006-01-10 13:02:29 -08001493 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01001494 return 0;
1495
Thomas Grafd35b6852007-03-22 23:28:46 -07001496 /* Only requests are handled by the kernel */
1497 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
1498 goto skip;
1499
Thomas Graf45e7ae72007-03-22 23:29:10 -07001500 /* Skip control messages */
1501 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
1502 goto skip;
1503
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001504 err = cb(skb, nlh);
1505 if (err == -EINTR) {
1506 /* Not an error, but we interrupt processing */
1507 netlink_queue_skip(nlh, skb);
1508 return err;
Thomas Grafd35b6852007-03-22 23:28:46 -07001509 }
1510skip:
1511 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01001512 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01001513
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001514 netlink_queue_skip(nlh, skb);
Thomas Graf82ace472005-11-10 02:25:53 +01001515 }
1516
1517 return 0;
1518}
1519
1520/**
1521 * nelink_run_queue - Process netlink receive queue.
1522 * @sk: Netlink socket containing the queue
1523 * @qlen: Place to store queue length upon entry
1524 * @cb: Callback function invoked for each netlink message found
1525 *
1526 * Processes as much as there was in the queue upon entry and invokes
1527 * a callback function for each netlink message found. The callback
1528 * function may refuse a message by returning a negative error code
1529 * but setting the error pointer to 0 in which case this function
1530 * returns with a qlen != 0.
1531 *
1532 * qlen must be initialized to 0 before the initial entry, afterwards
1533 * the function may be called repeatedly until qlen reaches 0.
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001534 *
1535 * The callback function may return -EINTR to signal that processing
1536 * of netlink messages shall be interrupted. In this case the message
1537 * currently being processed will NOT be requeued onto the receive
1538 * queue.
Thomas Graf82ace472005-11-10 02:25:53 +01001539 */
1540void netlink_run_queue(struct sock *sk, unsigned int *qlen,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001541 int (*cb)(struct sk_buff *, struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01001542{
1543 struct sk_buff *skb;
1544
1545 if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue))
1546 *qlen = skb_queue_len(&sk->sk_receive_queue);
1547
1548 for (; *qlen; (*qlen)--) {
1549 skb = skb_dequeue(&sk->sk_receive_queue);
1550 if (netlink_rcv_skb(skb, cb)) {
1551 if (skb->len)
1552 skb_queue_head(&sk->sk_receive_queue, skb);
1553 else {
1554 kfree_skb(skb);
1555 (*qlen)--;
1556 }
1557 break;
1558 }
1559
1560 kfree_skb(skb);
1561 }
1562}
1563
1564/**
1565 * netlink_queue_skip - Skip netlink message while processing queue.
1566 * @nlh: Netlink message to be skipped
1567 * @skb: Socket buffer containing the netlink messages.
1568 *
1569 * Pulls the given netlink message off the socket buffer so the next
1570 * call to netlink_queue_run() will not reconsider the message.
1571 */
Adrian Bunk42bad1d2007-04-26 00:57:41 -07001572static void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb)
Thomas Graf82ace472005-11-10 02:25:53 +01001573{
1574 int msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1575
1576 if (msglen > skb->len)
1577 msglen = skb->len;
1578
1579 skb_pull(skb, msglen);
1580}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581
Thomas Grafd387f6a2006-08-15 00:31:06 -07001582/**
1583 * nlmsg_notify - send a notification netlink message
1584 * @sk: netlink socket to use
1585 * @skb: notification message
1586 * @pid: destination netlink pid for reports or 0
1587 * @group: destination multicast group or 0
1588 * @report: 1 to report back, 0 to disable
1589 * @flags: allocation flags
1590 */
1591int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid,
1592 unsigned int group, int report, gfp_t flags)
1593{
1594 int err = 0;
1595
1596 if (group) {
1597 int exclude_pid = 0;
1598
1599 if (report) {
1600 atomic_inc(&skb->users);
1601 exclude_pid = pid;
1602 }
1603
1604 /* errors reported via destination sk->sk_err */
1605 nlmsg_multicast(sk, skb, exclude_pid, group, flags);
1606 }
1607
1608 if (report)
1609 err = nlmsg_unicast(sk, skb, pid);
1610
1611 return err;
1612}
1613
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614#ifdef CONFIG_PROC_FS
1615struct nl_seq_iter {
1616 int link;
1617 int hash_idx;
1618};
1619
1620static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1621{
1622 struct nl_seq_iter *iter = seq->private;
1623 int i, j;
1624 struct sock *s;
1625 struct hlist_node *node;
1626 loff_t off = 0;
1627
1628 for (i=0; i<MAX_LINKS; i++) {
1629 struct nl_pid_hash *hash = &nl_table[i].hash;
1630
1631 for (j = 0; j <= hash->mask; j++) {
1632 sk_for_each(s, node, &hash->table[j]) {
1633 if (off == pos) {
1634 iter->link = i;
1635 iter->hash_idx = j;
1636 return s;
1637 }
1638 ++off;
1639 }
1640 }
1641 }
1642 return NULL;
1643}
1644
1645static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
1646{
1647 read_lock(&nl_table_lock);
1648 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1649}
1650
1651static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1652{
1653 struct sock *s;
1654 struct nl_seq_iter *iter;
1655 int i, j;
1656
1657 ++*pos;
1658
1659 if (v == SEQ_START_TOKEN)
1660 return netlink_seq_socket_idx(seq, 0);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 s = sk_next(v);
1663 if (s)
1664 return s;
1665
1666 iter = seq->private;
1667 i = iter->link;
1668 j = iter->hash_idx + 1;
1669
1670 do {
1671 struct nl_pid_hash *hash = &nl_table[i].hash;
1672
1673 for (; j <= hash->mask; j++) {
1674 s = sk_head(&hash->table[j]);
1675 if (s) {
1676 iter->link = i;
1677 iter->hash_idx = j;
1678 return s;
1679 }
1680 }
1681
1682 j = 0;
1683 } while (++i < MAX_LINKS);
1684
1685 return NULL;
1686}
1687
1688static void netlink_seq_stop(struct seq_file *seq, void *v)
1689{
1690 read_unlock(&nl_table_lock);
1691}
1692
1693
1694static int netlink_seq_show(struct seq_file *seq, void *v)
1695{
1696 if (v == SEQ_START_TOKEN)
1697 seq_puts(seq,
1698 "sk Eth Pid Groups "
1699 "Rmem Wmem Dump Locks\n");
1700 else {
1701 struct sock *s = v;
1702 struct netlink_sock *nlk = nlk_sk(s);
1703
1704 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1705 s,
1706 s->sk_protocol,
1707 nlk->pid,
Patrick McHardy513c2502005-09-06 15:43:59 -07001708 nlk->groups ? (u32)nlk->groups[0] : 0,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 atomic_read(&s->sk_rmem_alloc),
1710 atomic_read(&s->sk_wmem_alloc),
1711 nlk->cb,
1712 atomic_read(&s->sk_refcnt)
1713 );
1714
1715 }
1716 return 0;
1717}
1718
1719static struct seq_operations netlink_seq_ops = {
1720 .start = netlink_seq_start,
1721 .next = netlink_seq_next,
1722 .stop = netlink_seq_stop,
1723 .show = netlink_seq_show,
1724};
1725
1726
1727static int netlink_seq_open(struct inode *inode, struct file *file)
1728{
1729 struct seq_file *seq;
1730 struct nl_seq_iter *iter;
1731 int err;
1732
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001733 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 if (!iter)
1735 return -ENOMEM;
1736
1737 err = seq_open(file, &netlink_seq_ops);
1738 if (err) {
1739 kfree(iter);
1740 return err;
1741 }
1742
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 seq = file->private_data;
1744 seq->private = iter;
1745 return 0;
1746}
1747
Arjan van de Venda7071d2007-02-12 00:55:36 -08001748static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 .owner = THIS_MODULE,
1750 .open = netlink_seq_open,
1751 .read = seq_read,
1752 .llseek = seq_lseek,
1753 .release = seq_release_private,
1754};
1755
1756#endif
1757
1758int netlink_register_notifier(struct notifier_block *nb)
1759{
Alan Sterne041c682006-03-27 01:16:30 -08001760 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761}
1762
1763int netlink_unregister_notifier(struct notifier_block *nb)
1764{
Alan Sterne041c682006-03-27 01:16:30 -08001765 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766}
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001767
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001768static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 .family = PF_NETLINK,
1770 .owner = THIS_MODULE,
1771 .release = netlink_release,
1772 .bind = netlink_bind,
1773 .connect = netlink_connect,
1774 .socketpair = sock_no_socketpair,
1775 .accept = sock_no_accept,
1776 .getname = netlink_getname,
1777 .poll = datagram_poll,
1778 .ioctl = sock_no_ioctl,
1779 .listen = sock_no_listen,
1780 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001781 .setsockopt = netlink_setsockopt,
1782 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 .sendmsg = netlink_sendmsg,
1784 .recvmsg = netlink_recvmsg,
1785 .mmap = sock_no_mmap,
1786 .sendpage = sock_no_sendpage,
1787};
1788
1789static struct net_proto_family netlink_family_ops = {
1790 .family = PF_NETLINK,
1791 .create = netlink_create,
1792 .owner = THIS_MODULE, /* for consistency 8) */
1793};
1794
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795static int __init netlink_proto_init(void)
1796{
1797 struct sk_buff *dummy_skb;
1798 int i;
1799 unsigned long max;
1800 unsigned int order;
1801 int err = proto_register(&netlink_proto, 0);
1802
1803 if (err != 0)
1804 goto out;
1805
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07001806 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001808 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001809 if (!nl_table)
1810 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 if (num_physpages >= (128 * 1024))
1813 max = num_physpages >> (21 - PAGE_SHIFT);
1814 else
1815 max = num_physpages >> (23 - PAGE_SHIFT);
1816
1817 order = get_bitmask_order(max) - 1 + PAGE_SHIFT;
1818 max = (1UL << order) / sizeof(struct hlist_head);
1819 order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1;
1820
1821 for (i = 0; i < MAX_LINKS; i++) {
1822 struct nl_pid_hash *hash = &nl_table[i].hash;
1823
1824 hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table));
1825 if (!hash->table) {
1826 while (i-- > 0)
1827 nl_pid_hash_free(nl_table[i].hash.table,
1828 1 * sizeof(*hash->table));
1829 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001830 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 }
1832 memset(hash->table, 0, 1 * sizeof(*hash->table));
1833 hash->max_shift = order;
1834 hash->shift = 0;
1835 hash->mask = 0;
1836 hash->rehash_time = jiffies;
1837 }
1838
1839 sock_register(&netlink_family_ops);
1840#ifdef CONFIG_PROC_FS
1841 proc_net_fops_create("netlink", 0, &netlink_seq_fops);
1842#endif
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001843 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 rtnetlink_init();
1845out:
1846 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001847panic:
1848 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849}
1850
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851core_initcall(netlink_proto_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
1853EXPORT_SYMBOL(netlink_ack);
Thomas Graf82ace472005-11-10 02:25:53 +01001854EXPORT_SYMBOL(netlink_run_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855EXPORT_SYMBOL(netlink_broadcast);
1856EXPORT_SYMBOL(netlink_dump_start);
1857EXPORT_SYMBOL(netlink_kernel_create);
1858EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859EXPORT_SYMBOL(netlink_set_nonroot);
1860EXPORT_SYMBOL(netlink_unicast);
1861EXPORT_SYMBOL(netlink_unregister_notifier);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001862EXPORT_SYMBOL(nlmsg_notify);