blob: 629b06182f3faab89ec771df253575ed467b6a2a [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NETLINK Kernel-user communication protocol.
3 *
Alan Cox113aa832008-10-13 19:01:08 -07004 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +090011 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
13 * added netlink_proto_exit
14 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
15 * use nlk_sk, as sk->protinfo is on a diet 8)
Harald Welte4fdb3bb2005-08-09 19:40:55 -070016 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
17 * - inc module use count of module that owns
18 * the kernel socket in case userspace opens
19 * socket of same protocol
20 * - remove all module support, since netlink is
21 * mandatory if CONFIG_NET=y these days
Linus Torvalds1da177e2005-04-16 15:20:36 -070022 */
23
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/module.h>
25
Randy Dunlap4fc268d2006-01-11 12:17:47 -080026#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/kernel.h>
28#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/signal.h>
30#include <linux/sched.h>
31#include <linux/errno.h>
32#include <linux/string.h>
33#include <linux/stat.h>
34#include <linux/socket.h>
35#include <linux/un.h>
36#include <linux/fcntl.h>
37#include <linux/termios.h>
38#include <linux/sockios.h>
39#include <linux/net.h>
40#include <linux/fs.h>
41#include <linux/slab.h>
42#include <asm/uaccess.h>
43#include <linux/skbuff.h>
44#include <linux/netdevice.h>
45#include <linux/rtnetlink.h>
46#include <linux/proc_fs.h>
47#include <linux/seq_file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/notifier.h>
49#include <linux/security.h>
50#include <linux/jhash.h>
51#include <linux/jiffies.h>
52#include <linux/random.h>
53#include <linux/bitops.h>
54#include <linux/mm.h>
55#include <linux/types.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010056#include <linux/audit.h>
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070057#include <linux/mutex.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010058
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020059#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#include <net/sock.h>
61#include <net/scm.h>
Thomas Graf82ace472005-11-10 02:25:53 +010062#include <net/netlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Patrick McHardyf7fa9b12005-08-15 12:29:13 -070064#define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
Johannes Bergb4ff4f02007-07-18 15:46:06 -070065#define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
67struct netlink_sock {
68 /* struct sock has to be the first member of netlink_sock */
69 struct sock sk;
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;
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -070082 void (*netlink_rcv)(struct sk_buff *skb);
Patrick McHardy77247bb2005-08-14 19:27:13 -070083 struct module *module;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084};
85
Eric Dumazet5c398dc2010-10-24 04:27:10 +000086struct listeners {
87 struct rcu_head rcu;
88 unsigned long masks[0];
Johannes Berg6c04bb12009-07-10 09:51:32 +000089};
90
Patrick McHardy77247bb2005-08-14 19:27:13 -070091#define NETLINK_KERNEL_SOCKET 0x1
Patrick McHardy9a4595b2005-08-15 12:32:15 -070092#define NETLINK_RECV_PKTINFO 0x2
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +000093#define NETLINK_BROADCAST_SEND_ERROR 0x4
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -070094#define NETLINK_RECV_NO_ENOBUFS 0x8
Patrick McHardy77247bb2005-08-14 19:27:13 -070095
Linus Torvalds1da177e2005-04-16 15:20:36 -070096static inline struct netlink_sock *nlk_sk(struct sock *sk)
97{
Denis Cheng32b21e02007-08-28 15:41:11 -070098 return container_of(sk, struct netlink_sock, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099}
100
David S. Miller035c4c12011-12-23 17:33:03 -0500101static inline int netlink_is_kernel(struct sock *sk)
Denis V. Lunevaed81562007-10-10 21:14:32 -0700102{
103 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
104}
105
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106struct nl_pid_hash {
107 struct hlist_head *table;
108 unsigned long rehash_time;
109
110 unsigned int mask;
111 unsigned int shift;
112
113 unsigned int entries;
114 unsigned int max_shift;
115
116 u32 rnd;
117};
118
119struct netlink_table {
120 struct nl_pid_hash hash;
121 struct hlist_head mc_list;
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000122 struct listeners __rcu *listeners;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 unsigned int nl_nonroot;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700124 unsigned int groups;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700125 struct mutex *cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700126 struct module *module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700127 int registered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128};
129
130static struct netlink_table *nl_table;
131
132static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
133
134static int netlink_dump(struct sock *sk);
135static void netlink_destroy_callback(struct netlink_callback *cb);
136
137static DEFINE_RWLOCK(nl_table_lock);
138static atomic_t nl_table_users = ATOMIC_INIT(0);
139
Alan Sterne041c682006-03-27 01:16:30 -0800140static ATOMIC_NOTIFIER_HEAD(netlink_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000142static inline u32 netlink_group_mask(u32 group)
Patrick McHardyd629b832005-08-14 19:27:50 -0700143{
144 return group ? 1 << (group - 1) : 0;
145}
146
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000147static inline struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148{
149 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
150}
151
152static void netlink_sock_destruct(struct sock *sk)
153{
Herbert Xu3f660d62007-05-03 03:17:14 -0700154 struct netlink_sock *nlk = nlk_sk(sk);
155
Herbert Xu3f660d62007-05-03 03:17:14 -0700156 if (nlk->cb) {
157 if (nlk->cb->done)
158 nlk->cb->done(nlk->cb);
159 netlink_destroy_callback(nlk->cb);
160 }
161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 skb_queue_purge(&sk->sk_receive_queue);
163
164 if (!sock_flag(sk, SOCK_DEAD)) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800165 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 return;
167 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700168
169 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
170 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
171 WARN_ON(nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172}
173
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800174/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
175 * SMP. Look, when several writers sleep and reader wakes them up, all but one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
177 * this, _but_ remember, it adds useless work on UP machines.
178 */
179
Johannes Bergd136f1b2009-09-12 03:03:15 +0000180void netlink_table_grab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800181 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182{
Johannes Bergd136f1b2009-09-12 03:03:15 +0000183 might_sleep();
184
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700185 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187 if (atomic_read(&nl_table_users)) {
188 DECLARE_WAITQUEUE(wait, current);
189
190 add_wait_queue_exclusive(&nl_table_wait, &wait);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800191 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 set_current_state(TASK_UNINTERRUPTIBLE);
193 if (atomic_read(&nl_table_users) == 0)
194 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700195 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700197 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 }
199
200 __set_current_state(TASK_RUNNING);
201 remove_wait_queue(&nl_table_wait, &wait);
202 }
203}
204
Johannes Bergd136f1b2009-09-12 03:03:15 +0000205void netlink_table_ungrab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800206 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700208 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 wake_up(&nl_table_wait);
210}
211
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800212static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213netlink_lock_table(void)
214{
215 /* read_lock() synchronizes us to netlink_table_grab */
216
217 read_lock(&nl_table_lock);
218 atomic_inc(&nl_table_users);
219 read_unlock(&nl_table_lock);
220}
221
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800222static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223netlink_unlock_table(void)
224{
225 if (atomic_dec_and_test(&nl_table_users))
226 wake_up(&nl_table_wait);
227}
228
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000229static struct sock *netlink_lookup(struct net *net, int protocol, u32 pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230{
231 struct nl_pid_hash *hash = &nl_table[protocol].hash;
232 struct hlist_head *head;
233 struct sock *sk;
234 struct hlist_node *node;
235
236 read_lock(&nl_table_lock);
237 head = nl_pid_hashfn(hash, pid);
238 sk_for_each(sk, node, head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900239 if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->pid == pid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 sock_hold(sk);
241 goto found;
242 }
243 }
244 sk = NULL;
245found:
246 read_unlock(&nl_table_lock);
247 return sk;
248}
249
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000250static struct hlist_head *nl_pid_hash_zalloc(size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251{
252 if (size <= PAGE_SIZE)
Eric Dumazetea729122007-12-11 02:09:47 -0800253 return kzalloc(size, GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 else
255 return (struct hlist_head *)
Eric Dumazetea729122007-12-11 02:09:47 -0800256 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
257 get_order(size));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258}
259
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000260static void nl_pid_hash_free(struct hlist_head *table, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261{
262 if (size <= PAGE_SIZE)
263 kfree(table);
264 else
265 free_pages((unsigned long)table, get_order(size));
266}
267
268static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
269{
270 unsigned int omask, mask, shift;
271 size_t osize, size;
272 struct hlist_head *otable, *table;
273 int i;
274
275 omask = mask = hash->mask;
276 osize = size = (mask + 1) * sizeof(*table);
277 shift = hash->shift;
278
279 if (grow) {
280 if (++shift > hash->max_shift)
281 return 0;
282 mask = mask * 2 + 1;
283 size *= 2;
284 }
285
Eric Dumazetea729122007-12-11 02:09:47 -0800286 table = nl_pid_hash_zalloc(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 if (!table)
288 return 0;
289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 otable = hash->table;
291 hash->table = table;
292 hash->mask = mask;
293 hash->shift = shift;
294 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
295
296 for (i = 0; i <= omask; i++) {
297 struct sock *sk;
298 struct hlist_node *node, *tmp;
299
300 sk_for_each_safe(sk, node, tmp, &otable[i])
301 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
302 }
303
304 nl_pid_hash_free(otable, osize);
305 hash->rehash_time = jiffies + 10 * 60 * HZ;
306 return 1;
307}
308
309static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
310{
311 int avg = hash->entries >> hash->shift;
312
313 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
314 return 1;
315
316 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
317 nl_pid_hash_rehash(hash, 0);
318 return 1;
319 }
320
321 return 0;
322}
323
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800324static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
Patrick McHardy4277a082006-03-20 18:52:01 -0800326static void
327netlink_update_listeners(struct sock *sk)
328{
329 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
330 struct hlist_node *node;
331 unsigned long mask;
332 unsigned int i;
333
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700334 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
Patrick McHardy4277a082006-03-20 18:52:01 -0800335 mask = 0;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700336 sk_for_each_bound(sk, node, &tbl->mc_list) {
337 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
338 mask |= nlk_sk(sk)->groups[i];
339 }
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000340 tbl->listeners->masks[i] = mask;
Patrick McHardy4277a082006-03-20 18:52:01 -0800341 }
342 /* this function is only called with the netlink table "grabbed", which
343 * makes sure updates are visible before bind or setsockopt return. */
344}
345
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200346static int netlink_insert(struct sock *sk, struct net *net, u32 pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347{
348 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
349 struct hlist_head *head;
350 int err = -EADDRINUSE;
351 struct sock *osk;
352 struct hlist_node *node;
353 int len;
354
355 netlink_table_grab();
356 head = nl_pid_hashfn(hash, pid);
357 len = 0;
358 sk_for_each(osk, node, head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900359 if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->pid == pid))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 break;
361 len++;
362 }
363 if (node)
364 goto err;
365
366 err = -EBUSY;
367 if (nlk_sk(sk)->pid)
368 goto err;
369
370 err = -ENOMEM;
371 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
372 goto err;
373
374 if (len && nl_pid_hash_dilute(hash, len))
375 head = nl_pid_hashfn(hash, pid);
376 hash->entries++;
377 nlk_sk(sk)->pid = pid;
378 sk_add_node(sk, head);
379 err = 0;
380
381err:
382 netlink_table_ungrab();
383 return err;
384}
385
386static void netlink_remove(struct sock *sk)
387{
388 netlink_table_grab();
David S. Millerd470e3b2005-06-26 15:31:51 -0700389 if (sk_del_node_init(sk))
390 nl_table[sk->sk_protocol].hash.entries--;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700391 if (nlk_sk(sk)->subscriptions)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 __sk_del_bind_node(sk);
393 netlink_table_ungrab();
394}
395
396static struct proto netlink_proto = {
397 .name = "NETLINK",
398 .owner = THIS_MODULE,
399 .obj_size = sizeof(struct netlink_sock),
400};
401
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700402static int __netlink_create(struct net *net, struct socket *sock,
403 struct mutex *cb_mutex, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404{
405 struct sock *sk;
406 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700407
408 sock->ops = &netlink_ops;
409
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700410 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
Patrick McHardyab33a172005-08-14 19:31:36 -0700411 if (!sk)
412 return -ENOMEM;
413
414 sock_init_data(sock, sk);
415
416 nlk = nlk_sk(sk);
Patrick McHardyffa4d722007-04-25 14:01:17 -0700417 if (cb_mutex)
418 nlk->cb_mutex = cb_mutex;
419 else {
420 nlk->cb_mutex = &nlk->cb_def_mutex;
421 mutex_init(nlk->cb_mutex);
422 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700423 init_waitqueue_head(&nlk->wait);
424
425 sk->sk_destruct = netlink_sock_destruct;
426 sk->sk_protocol = protocol;
427 return 0;
428}
429
Eric Paris3f378b62009-11-05 22:18:14 -0800430static int netlink_create(struct net *net, struct socket *sock, int protocol,
431 int kern)
Patrick McHardyab33a172005-08-14 19:31:36 -0700432{
433 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700434 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700435 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -0700436 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 sock->state = SS_UNCONNECTED;
439
440 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
441 return -ESOCKTNOSUPPORT;
442
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800443 if (protocol < 0 || protocol >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 return -EPROTONOSUPPORT;
445
Patrick McHardy77247bb2005-08-14 19:27:13 -0700446 netlink_lock_table();
Johannes Berg95a5afc2008-10-16 15:24:51 -0700447#ifdef CONFIG_MODULES
Patrick McHardyab33a172005-08-14 19:31:36 -0700448 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -0700449 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700450 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -0700451 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700452 }
Patrick McHardyab33a172005-08-14 19:31:36 -0700453#endif
454 if (nl_table[protocol].registered &&
455 try_module_get(nl_table[protocol].module))
456 module = nl_table[protocol].module;
Alexey Dobriyan974c37e2010-01-30 10:05:05 +0000457 else
458 err = -EPROTONOSUPPORT;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -0700459 cb_mutex = nl_table[protocol].cb_mutex;
Patrick McHardy77247bb2005-08-14 19:27:13 -0700460 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700461
Alexey Dobriyan974c37e2010-01-30 10:05:05 +0000462 if (err < 0)
463 goto out;
464
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800465 err = __netlink_create(net, sock, cb_mutex, protocol);
466 if (err < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -0700467 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
David S. Miller6f756a82008-11-23 17:34:03 -0800469 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -0800470 sock_prot_inuse_add(net, &netlink_proto, 1);
David S. Miller6f756a82008-11-23 17:34:03 -0800471 local_bh_enable();
472
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700473 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700474 nlk->module = module;
Patrick McHardyab33a172005-08-14 19:31:36 -0700475out:
476 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Patrick McHardyab33a172005-08-14 19:31:36 -0700478out_module:
479 module_put(module);
480 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481}
482
483static int netlink_release(struct socket *sock)
484{
485 struct sock *sk = sock->sk;
486 struct netlink_sock *nlk;
487
488 if (!sk)
489 return 0;
490
491 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -0700492 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 nlk = nlk_sk(sk);
494
Herbert Xu3f660d62007-05-03 03:17:14 -0700495 /*
496 * OK. Socket is unlinked, any packets that arrive now
497 * will be purged.
498 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 sock->sk = NULL;
501 wake_up_interruptible_all(&nlk->wait);
502
503 skb_queue_purge(&sk->sk_write_queue);
504
Johannes Berg649300b2009-11-16 12:05:34 +0000505 if (nlk->pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 struct netlink_notify n = {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900507 .net = sock_net(sk),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 .protocol = sk->sk_protocol,
509 .pid = nlk->pid,
510 };
Alan Sterne041c682006-03-27 01:16:30 -0800511 atomic_notifier_call_chain(&netlink_chain,
512 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900513 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700514
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -0800515 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -0700516
Patrick McHardy4277a082006-03-20 18:52:01 -0800517 netlink_table_grab();
Denis V. Lunevaed81562007-10-10 21:14:32 -0700518 if (netlink_is_kernel(sk)) {
Denis V. Lunev869e58f2008-01-18 23:53:31 -0800519 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
520 if (--nl_table[sk->sk_protocol].registered == 0) {
521 kfree(nl_table[sk->sk_protocol].listeners);
522 nl_table[sk->sk_protocol].module = NULL;
523 nl_table[sk->sk_protocol].registered = 0;
524 }
Patrick McHardy4277a082006-03-20 18:52:01 -0800525 } else if (nlk->subscriptions)
526 netlink_update_listeners(sk);
527 netlink_table_ungrab();
Patrick McHardy77247bb2005-08-14 19:27:13 -0700528
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700529 kfree(nlk->groups);
530 nlk->groups = NULL;
531
Eric Dumazet37558102008-11-24 14:05:22 -0800532 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -0800533 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
Eric Dumazet37558102008-11-24 14:05:22 -0800534 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 sock_put(sk);
536 return 0;
537}
538
539static int netlink_autobind(struct socket *sock)
540{
541 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900542 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
544 struct hlist_head *head;
545 struct sock *osk;
546 struct hlist_node *node;
Tom Goff66aa4a52010-03-19 15:38:50 +0000547 s32 pid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 int err;
549 static s32 rover = -4097;
550
551retry:
552 cond_resched();
553 netlink_table_grab();
554 head = nl_pid_hashfn(hash, pid);
555 sk_for_each(osk, node, head) {
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900556 if (!net_eq(sock_net(osk), net))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200557 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 if (nlk_sk(osk)->pid == pid) {
559 /* Bind collision, search negative pid values. */
560 pid = rover--;
561 if (rover > -4097)
562 rover = -4097;
563 netlink_table_ungrab();
564 goto retry;
565 }
566 }
567 netlink_table_ungrab();
568
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200569 err = netlink_insert(sk, net, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 if (err == -EADDRINUSE)
571 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -0700572
573 /* If 2 threads race to autobind, that is fine. */
574 if (err == -EBUSY)
575 err = 0;
576
577 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578}
579
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000580static inline int netlink_capable(const struct socket *sock, unsigned int flag)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900581{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
583 capable(CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900584}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700586static void
587netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
588{
589 struct netlink_sock *nlk = nlk_sk(sk);
590
591 if (nlk->subscriptions && !subscriptions)
592 __sk_del_bind_node(sk);
593 else if (!nlk->subscriptions && subscriptions)
594 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
595 nlk->subscriptions = subscriptions;
596}
597
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700598static int netlink_realloc_groups(struct sock *sk)
Patrick McHardy513c2502005-09-06 15:43:59 -0700599{
600 struct netlink_sock *nlk = nlk_sk(sk);
601 unsigned int groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700602 unsigned long *new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700603 int err = 0;
604
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700605 netlink_table_grab();
606
Patrick McHardy513c2502005-09-06 15:43:59 -0700607 groups = nl_table[sk->sk_protocol].groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700608 if (!nl_table[sk->sk_protocol].registered) {
Patrick McHardy513c2502005-09-06 15:43:59 -0700609 err = -ENOENT;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700610 goto out_unlock;
611 }
Patrick McHardy513c2502005-09-06 15:43:59 -0700612
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700613 if (nlk->ngroups >= groups)
614 goto out_unlock;
Patrick McHardy513c2502005-09-06 15:43:59 -0700615
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700616 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
617 if (new_groups == NULL) {
618 err = -ENOMEM;
619 goto out_unlock;
620 }
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800621 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700622 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
623
624 nlk->groups = new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -0700625 nlk->ngroups = groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700626 out_unlock:
627 netlink_table_ungrab();
628 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700629}
630
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800631static int netlink_bind(struct socket *sock, struct sockaddr *addr,
632 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
634 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900635 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 struct netlink_sock *nlk = nlk_sk(sk);
637 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
638 int err;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900639
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 if (nladdr->nl_family != AF_NETLINK)
641 return -EINVAL;
642
643 /* Only superuser is allowed to listen multicasts */
Patrick McHardy513c2502005-09-06 15:43:59 -0700644 if (nladdr->nl_groups) {
645 if (!netlink_capable(sock, NL_NONROOT_RECV))
646 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700647 err = netlink_realloc_groups(sk);
648 if (err)
649 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -0700650 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652 if (nlk->pid) {
653 if (nladdr->nl_pid != nlk->pid)
654 return -EINVAL;
655 } else {
656 err = nladdr->nl_pid ?
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200657 netlink_insert(sk, net, nladdr->nl_pid) :
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 netlink_autobind(sock);
659 if (err)
660 return err;
661 }
662
Patrick McHardy513c2502005-09-06 15:43:59 -0700663 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 return 0;
665
666 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700667 netlink_update_subscriptions(sk, nlk->subscriptions +
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900668 hweight32(nladdr->nl_groups) -
669 hweight32(nlk->groups[0]));
670 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
Patrick McHardy4277a082006-03-20 18:52:01 -0800671 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 netlink_table_ungrab();
673
674 return 0;
675}
676
677static int netlink_connect(struct socket *sock, struct sockaddr *addr,
678 int alen, int flags)
679{
680 int err = 0;
681 struct sock *sk = sock->sk;
682 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800683 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
Changli Gao6503d962010-03-31 22:58:26 +0000685 if (alen < sizeof(addr->sa_family))
686 return -EINVAL;
687
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 if (addr->sa_family == AF_UNSPEC) {
689 sk->sk_state = NETLINK_UNCONNECTED;
690 nlk->dst_pid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -0700691 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 return 0;
693 }
694 if (addr->sa_family != AF_NETLINK)
695 return -EINVAL;
696
697 /* Only superuser is allowed to send multicasts */
698 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
699 return -EPERM;
700
701 if (!nlk->pid)
702 err = netlink_autobind(sock);
703
704 if (err == 0) {
705 sk->sk_state = NETLINK_CONNECTED;
706 nlk->dst_pid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700707 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709
710 return err;
711}
712
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800713static int netlink_getname(struct socket *sock, struct sockaddr *addr,
714 int *addr_len, int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715{
716 struct sock *sk = sock->sk;
717 struct netlink_sock *nlk = nlk_sk(sk);
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +0000718 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +0900719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 nladdr->nl_family = AF_NETLINK;
721 nladdr->nl_pad = 0;
722 *addr_len = sizeof(*nladdr);
723
724 if (peer) {
725 nladdr->nl_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -0700726 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 } else {
728 nladdr->nl_pid = nlk->pid;
Patrick McHardy513c2502005-09-06 15:43:59 -0700729 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 }
731 return 0;
732}
733
734static void netlink_overrun(struct sock *sk)
735{
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -0700736 struct netlink_sock *nlk = nlk_sk(sk);
737
738 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
739 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
740 sk->sk_err = ENOBUFS;
741 sk->sk_error_report(sk);
742 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 }
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -0700744 atomic_inc(&sk->sk_drops);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745}
746
747static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
748{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 struct sock *sock;
750 struct netlink_sock *nlk;
751
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900752 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 if (!sock)
754 return ERR_PTR(-ECONNREFUSED);
755
756 /* Don't bother queuing skb if kernel socket has no input function */
757 nlk = nlk_sk(sock);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700758 if (sock->sk_state == NETLINK_CONNECTED &&
759 nlk->dst_pid != nlk_sk(ssk)->pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 sock_put(sock);
761 return ERR_PTR(-ECONNREFUSED);
762 }
763 return sock;
764}
765
766struct sock *netlink_getsockbyfilp(struct file *filp)
767{
Josef Sipek6db5fc52006-12-08 02:37:25 -0800768 struct inode *inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 struct sock *sock;
770
771 if (!S_ISSOCK(inode->i_mode))
772 return ERR_PTR(-ENOTSOCK);
773
774 sock = SOCKET_I(inode)->sk;
775 if (sock->sk_family != AF_NETLINK)
776 return ERR_PTR(-EINVAL);
777
778 sock_hold(sock);
779 return sock;
780}
781
782/*
783 * Attach a skb to a netlink socket.
784 * The caller must hold a reference to the destination socket. On error, the
785 * reference is dropped. The skb is not send to the destination, just all
786 * all error checks are performed and memory in the queue is reserved.
787 * Return values:
788 * < 0: error. skb freed, reference to sock dropped.
789 * 0: continue
790 * 1: repeat lookup - reference dropped while waiting for socket memory.
791 */
Denis V. Lunev9457afe2008-06-05 11:23:39 -0700792int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800793 long *timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794{
795 struct netlink_sock *nlk;
796
797 nlk = nlk_sk(sk);
798
799 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
800 test_bit(0, &nlk->state)) {
801 DECLARE_WAITQUEUE(wait, current);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800802 if (!*timeo) {
Denis V. Lunevaed81562007-10-10 21:14:32 -0700803 if (!ssk || netlink_is_kernel(ssk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 netlink_overrun(sk);
805 sock_put(sk);
806 kfree_skb(skb);
807 return -EAGAIN;
808 }
809
810 __set_current_state(TASK_INTERRUPTIBLE);
811 add_wait_queue(&nlk->wait, &wait);
812
813 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
814 test_bit(0, &nlk->state)) &&
815 !sock_flag(sk, SOCK_DEAD))
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800816 *timeo = schedule_timeout(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
818 __set_current_state(TASK_RUNNING);
819 remove_wait_queue(&nlk->wait, &wait);
820 sock_put(sk);
821
822 if (signal_pending(current)) {
823 kfree_skb(skb);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -0800824 return sock_intr_errno(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 }
826 return 1;
827 }
828 skb_set_owner_r(skb, sk);
829 return 0;
830}
831
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700832int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 int len = skb->len;
835
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 skb_queue_tail(&sk->sk_receive_queue, skb);
837 sk->sk_data_ready(sk, len);
838 sock_put(sk);
839 return len;
840}
841
842void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
843{
844 kfree_skb(skb);
845 sock_put(sk);
846}
847
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000848static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849{
850 int delta;
851
852 skb_orphan(skb);
853
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700854 delta = skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 if (delta * 2 < skb->truesize)
856 return skb;
857
858 if (skb_shared(skb)) {
859 struct sk_buff *nskb = skb_clone(skb, allocation);
860 if (!nskb)
861 return skb;
862 kfree_skb(skb);
863 skb = nskb;
864 }
865
866 if (!pskb_expand_head(skb, 0, -delta, allocation))
867 skb->truesize -= delta;
868
869 return skb;
870}
871
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000872static void netlink_rcv_wake(struct sock *sk)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700873{
874 struct netlink_sock *nlk = nlk_sk(sk);
875
876 if (skb_queue_empty(&sk->sk_receive_queue))
877 clear_bit(0, &nlk->state);
878 if (!test_bit(0, &nlk->state))
879 wake_up_interruptible(&nlk->wait);
880}
881
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000882static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700883{
884 int ret;
885 struct netlink_sock *nlk = nlk_sk(sk);
886
887 ret = -ECONNREFUSED;
888 if (nlk->netlink_rcv != NULL) {
889 ret = skb->len;
890 skb_set_owner_r(skb, sk);
891 nlk->netlink_rcv(skb);
892 }
893 kfree_skb(skb);
894 sock_put(sk);
895 return ret;
896}
897
898int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
899 u32 pid, int nonblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900{
901 struct sock *sk;
902 int err;
903 long timeo;
904
905 skb = netlink_trim(skb, gfp_any());
906
907 timeo = sock_sndtimeo(ssk, nonblock);
908retry:
909 sk = netlink_getsockbypid(ssk, pid);
910 if (IS_ERR(sk)) {
911 kfree_skb(skb);
912 return PTR_ERR(sk);
913 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -0700914 if (netlink_is_kernel(sk))
915 return netlink_unicast_kernel(sk, skb);
916
Stephen Hemmingerb1153f22008-03-21 15:46:12 -0700917 if (sk_filter(sk, skb)) {
Wang Chen84874602008-07-01 19:55:09 -0700918 err = skb->len;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -0700919 kfree_skb(skb);
920 sock_put(sk);
921 return err;
922 }
923
Denis V. Lunev9457afe2008-06-05 11:23:39 -0700924 err = netlink_attachskb(sk, skb, &timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 if (err == 1)
926 goto retry;
927 if (err)
928 return err;
929
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700930 return netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931}
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800932EXPORT_SYMBOL(netlink_unicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933
Patrick McHardy4277a082006-03-20 18:52:01 -0800934int netlink_has_listeners(struct sock *sk, unsigned int group)
935{
936 int res = 0;
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000937 struct listeners *listeners;
Patrick McHardy4277a082006-03-20 18:52:01 -0800938
Denis V. Lunevaed81562007-10-10 21:14:32 -0700939 BUG_ON(!netlink_is_kernel(sk));
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700940
941 rcu_read_lock();
942 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
943
Patrick McHardy4277a082006-03-20 18:52:01 -0800944 if (group - 1 < nl_table[sk->sk_protocol].groups)
Eric Dumazet5c398dc2010-10-24 04:27:10 +0000945 res = test_bit(group - 1, listeners->masks);
Johannes Bergb4ff4f02007-07-18 15:46:06 -0700946
947 rcu_read_unlock();
948
Patrick McHardy4277a082006-03-20 18:52:01 -0800949 return res;
950}
951EXPORT_SYMBOL_GPL(netlink_has_listeners);
952
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000953static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954{
955 struct netlink_sock *nlk = nlk_sk(sk);
956
957 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
958 !test_bit(0, &nlk->state)) {
959 skb_set_owner_r(skb, sk);
960 skb_queue_tail(&sk->sk_receive_queue, skb);
961 sk->sk_data_ready(sk, skb->len);
stephen hemminger2c6458002011-12-22 08:52:03 +0000962 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 }
964 return -1;
965}
966
967struct netlink_broadcast_data {
968 struct sock *exclude_sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200969 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 u32 pid;
971 u32 group;
972 int failure;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -0800973 int delivery_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 int congested;
975 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -0400976 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 struct sk_buff *skb, *skb2;
Eric W. Biederman910a7e92010-05-04 17:36:46 -0700978 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
979 void *tx_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980};
981
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000982static int do_one_broadcast(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 struct netlink_broadcast_data *p)
984{
985 struct netlink_sock *nlk = nlk_sk(sk);
986 int val;
987
988 if (p->exclude_sk == sk)
989 goto out;
990
Patrick McHardyf7fa9b12005-08-15 12:29:13 -0700991 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
992 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 goto out;
994
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900995 if (!net_eq(sock_net(sk), p->net))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +0200996 goto out;
997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 if (p->failure) {
999 netlink_overrun(sk);
1000 goto out;
1001 }
1002
1003 sock_hold(sk);
1004 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001005 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 p->skb2 = skb_clone(p->skb, p->allocation);
1007 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001008 p->skb2 = skb_get(p->skb);
1009 /*
1010 * skb ownership may have been set when
1011 * delivered to a previous socket.
1012 */
1013 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 }
1015 }
1016 if (p->skb2 == NULL) {
1017 netlink_overrun(sk);
1018 /* Clone failed. Notify ALL listeners. */
1019 p->failure = 1;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001020 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1021 p->delivery_failure = 1;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001022 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1023 kfree_skb(p->skb2);
1024 p->skb2 = NULL;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001025 } else if (sk_filter(sk, p->skb2)) {
1026 kfree_skb(p->skb2);
1027 p->skb2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1029 netlink_overrun(sk);
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001030 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1031 p->delivery_failure = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 } else {
1033 p->congested |= val;
1034 p->delivered = 1;
1035 p->skb2 = NULL;
1036 }
1037 sock_put(sk);
1038
1039out:
1040 return 0;
1041}
1042
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001043int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 pid,
1044 u32 group, gfp_t allocation,
1045 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1046 void *filter_data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001048 struct net *net = sock_net(ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 struct netlink_broadcast_data info;
1050 struct hlist_node *node;
1051 struct sock *sk;
1052
1053 skb = netlink_trim(skb, allocation);
1054
1055 info.exclude_sk = ssk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001056 info.net = net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 info.pid = pid;
1058 info.group = group;
1059 info.failure = 0;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001060 info.delivery_failure = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 info.congested = 0;
1062 info.delivered = 0;
1063 info.allocation = allocation;
1064 info.skb = skb;
1065 info.skb2 = NULL;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001066 info.tx_filter = filter;
1067 info.tx_data = filter_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
1069 /* While we sleep in clone, do not allow to change socket list */
1070
1071 netlink_lock_table();
1072
1073 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1074 do_one_broadcast(sk, &info);
1075
Neil Horman70d4bf62010-07-20 06:45:56 +00001076 consume_skb(skb);
Tommy S. Christensenaa1c6a62005-05-19 13:07:32 -07001077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 netlink_unlock_table();
1079
Neil Horman70d4bf62010-07-20 06:45:56 +00001080 if (info.delivery_failure) {
1081 kfree_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001082 return -ENOBUFS;
Neil Horman70d4bf62010-07-20 06:45:56 +00001083 } else
1084 consume_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001085
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 if (info.delivered) {
1087 if (info.congested && (allocation & __GFP_WAIT))
1088 yield();
1089 return 0;
1090 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 return -ESRCH;
1092}
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001093EXPORT_SYMBOL(netlink_broadcast_filtered);
1094
1095int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
1096 u32 group, gfp_t allocation)
1097{
1098 return netlink_broadcast_filtered(ssk, skb, pid, group, allocation,
1099 NULL, NULL);
1100}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001101EXPORT_SYMBOL(netlink_broadcast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
1103struct netlink_set_err_data {
1104 struct sock *exclude_sk;
1105 u32 pid;
1106 u32 group;
1107 int code;
1108};
1109
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001110static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111{
1112 struct netlink_sock *nlk = nlk_sk(sk);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001113 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114
1115 if (sk == p->exclude_sk)
1116 goto out;
1117
Octavian Purdila09ad9bc2009-11-25 15:14:13 -08001118 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001119 goto out;
1120
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001121 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
1122 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 goto out;
1124
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001125 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
1126 ret = 1;
1127 goto out;
1128 }
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 sk->sk_err = p->code;
1131 sk->sk_error_report(sk);
1132out:
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001133 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001136/**
1137 * netlink_set_err - report error to broadcast listeners
1138 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1139 * @pid: the PID of a process that we want to skip (if any)
1140 * @groups: the broadcast group that will notice the error
1141 * @code: error code, must be negative (as usual in kernelspace)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001142 *
1143 * This function returns the number of broadcast listeners that have set the
1144 * NETLINK_RECV_NO_ENOBUFS socket option.
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001145 */
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001146int netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147{
1148 struct netlink_set_err_data info;
1149 struct hlist_node *node;
1150 struct sock *sk;
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001151 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152
1153 info.exclude_sk = ssk;
1154 info.pid = pid;
1155 info.group = group;
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08001156 /* sk->sk_err wants a positive error value */
1157 info.code = -code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
1159 read_lock(&nl_table_lock);
1160
1161 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001162 ret += do_one_set_err(sk, &info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164 read_unlock(&nl_table_lock);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00001165 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166}
Pablo Neira Ayusodd5b6ce2009-03-23 13:21:06 +01001167EXPORT_SYMBOL(netlink_set_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
Johannes Berg84659eb2007-07-18 15:47:05 -07001169/* must be called with netlink table grabbed */
1170static void netlink_update_socket_mc(struct netlink_sock *nlk,
1171 unsigned int group,
1172 int is_new)
1173{
1174 int old, new = !!is_new, subscriptions;
1175
1176 old = test_bit(group - 1, nlk->groups);
1177 subscriptions = nlk->subscriptions - old + new;
1178 if (new)
1179 __set_bit(group - 1, nlk->groups);
1180 else
1181 __clear_bit(group - 1, nlk->groups);
1182 netlink_update_subscriptions(&nlk->sk, subscriptions);
1183 netlink_update_listeners(&nlk->sk);
1184}
1185
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001186static int netlink_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001187 char __user *optval, unsigned int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001188{
1189 struct sock *sk = sock->sk;
1190 struct netlink_sock *nlk = nlk_sk(sk);
Johannes Bergeb496532007-07-18 02:07:51 -07001191 unsigned int val = 0;
1192 int err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001193
1194 if (level != SOL_NETLINK)
1195 return -ENOPROTOOPT;
1196
1197 if (optlen >= sizeof(int) &&
Johannes Bergeb496532007-07-18 02:07:51 -07001198 get_user(val, (unsigned int __user *)optval))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001199 return -EFAULT;
1200
1201 switch (optname) {
1202 case NETLINK_PKTINFO:
1203 if (val)
1204 nlk->flags |= NETLINK_RECV_PKTINFO;
1205 else
1206 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1207 err = 0;
1208 break;
1209 case NETLINK_ADD_MEMBERSHIP:
1210 case NETLINK_DROP_MEMBERSHIP: {
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001211 if (!netlink_capable(sock, NL_NONROOT_RECV))
1212 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001213 err = netlink_realloc_groups(sk);
1214 if (err)
1215 return err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001216 if (!val || val - 1 >= nlk->ngroups)
1217 return -EINVAL;
1218 netlink_table_grab();
Johannes Berg84659eb2007-07-18 15:47:05 -07001219 netlink_update_socket_mc(nlk, val,
1220 optname == NETLINK_ADD_MEMBERSHIP);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001221 netlink_table_ungrab();
1222 err = 0;
1223 break;
1224 }
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001225 case NETLINK_BROADCAST_ERROR:
1226 if (val)
1227 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
1228 else
1229 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
1230 err = 0;
1231 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001232 case NETLINK_NO_ENOBUFS:
1233 if (val) {
1234 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
1235 clear_bit(0, &nlk->state);
1236 wake_up_interruptible(&nlk->wait);
1237 } else
1238 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
1239 err = 0;
1240 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001241 default:
1242 err = -ENOPROTOOPT;
1243 }
1244 return err;
1245}
1246
1247static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001248 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001249{
1250 struct sock *sk = sock->sk;
1251 struct netlink_sock *nlk = nlk_sk(sk);
1252 int len, val, err;
1253
1254 if (level != SOL_NETLINK)
1255 return -ENOPROTOOPT;
1256
1257 if (get_user(len, optlen))
1258 return -EFAULT;
1259 if (len < 0)
1260 return -EINVAL;
1261
1262 switch (optname) {
1263 case NETLINK_PKTINFO:
1264 if (len < sizeof(int))
1265 return -EINVAL;
1266 len = sizeof(int);
1267 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08001268 if (put_user(len, optlen) ||
1269 put_user(val, optval))
1270 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001271 err = 0;
1272 break;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001273 case NETLINK_BROADCAST_ERROR:
1274 if (len < sizeof(int))
1275 return -EINVAL;
1276 len = sizeof(int);
1277 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
1278 if (put_user(len, optlen) ||
1279 put_user(val, optval))
1280 return -EFAULT;
1281 err = 0;
1282 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001283 case NETLINK_NO_ENOBUFS:
1284 if (len < sizeof(int))
1285 return -EINVAL;
1286 len = sizeof(int);
1287 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
1288 if (put_user(len, optlen) ||
1289 put_user(val, optval))
1290 return -EFAULT;
1291 err = 0;
1292 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07001293 default:
1294 err = -ENOPROTOOPT;
1295 }
1296 return err;
1297}
1298
1299static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1300{
1301 struct nl_pktinfo info;
1302
1303 info.group = NETLINK_CB(skb).dst_group;
1304 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1305}
1306
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1308 struct msghdr *msg, size_t len)
1309{
1310 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1311 struct sock *sk = sock->sk;
1312 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001313 struct sockaddr_nl *addr = msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 u32 dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001315 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 struct sk_buff *skb;
1317 int err;
1318 struct scm_cookie scm;
1319
1320 if (msg->msg_flags&MSG_OOB)
1321 return -EOPNOTSUPP;
1322
Eric Dumazet16e57262011-09-19 05:52:27 +00001323 if (NULL == siocb->scm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 siocb->scm = &scm;
Eric Dumazet16e57262011-09-19 05:52:27 +00001325
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 err = scm_send(sock, msg, siocb->scm);
1327 if (err < 0)
1328 return err;
1329
1330 if (msg->msg_namelen) {
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001331 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 if (addr->nl_family != AF_NETLINK)
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001333 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 dst_pid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001335 dst_group = ffs(addr->nl_groups);
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001336 err = -EPERM;
Patrick McHardyd629b832005-08-14 19:27:50 -07001337 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001338 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 } else {
1340 dst_pid = nlk->dst_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001341 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 }
1343
1344 if (!nlk->pid) {
1345 err = netlink_autobind(sock);
1346 if (err)
1347 goto out;
1348 }
1349
1350 err = -EMSGSIZE;
1351 if (len > sk->sk_sndbuf - 32)
1352 goto out;
1353 err = -ENOBUFS;
Thomas Graf339bf982006-11-10 14:10:15 -08001354 skb = alloc_skb(len, GFP_KERNEL);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001355 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 goto out;
1357
1358 NETLINK_CB(skb).pid = nlk->pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001359 NETLINK_CB(skb).dst_group = dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 err = -EFAULT;
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001363 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 kfree_skb(skb);
1365 goto out;
1366 }
1367
1368 err = security_netlink_send(sk, skb);
1369 if (err) {
1370 kfree_skb(skb);
1371 goto out;
1372 }
1373
Patrick McHardyd629b832005-08-14 19:27:50 -07001374 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 atomic_inc(&skb->users);
Patrick McHardyd629b832005-08-14 19:27:50 -07001376 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 }
1378 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1379
1380out:
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00001381 scm_destroy(siocb->scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 return err;
1383}
1384
1385static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1386 struct msghdr *msg, size_t len,
1387 int flags)
1388{
1389 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1390 struct scm_cookie scm;
1391 struct sock *sk = sock->sk;
1392 struct netlink_sock *nlk = nlk_sk(sk);
1393 int noblock = flags&MSG_DONTWAIT;
1394 size_t copied;
Johannes Berg68d6ac62010-08-15 21:20:44 +00001395 struct sk_buff *skb, *data_skb;
Andrey Vaginb44d2112011-02-21 02:40:47 +00001396 int err, ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397
1398 if (flags&MSG_OOB)
1399 return -EOPNOTSUPP;
1400
1401 copied = 0;
1402
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001403 skb = skb_recv_datagram(sk, flags, noblock, &err);
1404 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 goto out;
1406
Johannes Berg68d6ac62010-08-15 21:20:44 +00001407 data_skb = skb;
1408
Johannes Berg1dacc762009-07-01 11:26:02 +00001409#ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1410 if (unlikely(skb_shinfo(skb)->frag_list)) {
Johannes Berg1dacc762009-07-01 11:26:02 +00001411 /*
Johannes Berg68d6ac62010-08-15 21:20:44 +00001412 * If this skb has a frag_list, then here that means that we
1413 * will have to use the frag_list skb's data for compat tasks
1414 * and the regular skb's data for normal (non-compat) tasks.
Johannes Berg1dacc762009-07-01 11:26:02 +00001415 *
Johannes Berg68d6ac62010-08-15 21:20:44 +00001416 * If we need to send the compat skb, assign it to the
1417 * 'data_skb' variable so that it will be used below for data
1418 * copying. We keep 'skb' for everything else, including
1419 * freeing both later.
Johannes Berg1dacc762009-07-01 11:26:02 +00001420 */
Johannes Berg68d6ac62010-08-15 21:20:44 +00001421 if (flags & MSG_CMSG_COMPAT)
1422 data_skb = skb_shinfo(skb)->frag_list;
Johannes Berg1dacc762009-07-01 11:26:02 +00001423 }
1424#endif
1425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 msg->msg_namelen = 0;
1427
Johannes Berg68d6ac62010-08-15 21:20:44 +00001428 copied = data_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 if (len < copied) {
1430 msg->msg_flags |= MSG_TRUNC;
1431 copied = len;
1432 }
1433
Johannes Berg68d6ac62010-08-15 21:20:44 +00001434 skb_reset_transport_header(data_skb);
1435 err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436
1437 if (msg->msg_name) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001438 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 addr->nl_family = AF_NETLINK;
1440 addr->nl_pad = 0;
1441 addr->nl_pid = NETLINK_CB(skb).pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001442 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 msg->msg_namelen = sizeof(*addr);
1444 }
1445
Patrick McHardycc9a06c2006-03-12 20:34:27 -08001446 if (nlk->flags & NETLINK_RECV_PKTINFO)
1447 netlink_cmsg_recv_pktinfo(msg, skb);
1448
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (NULL == siocb->scm) {
1450 memset(&scm, 0, sizeof(scm));
1451 siocb->scm = &scm;
1452 }
1453 siocb->scm->creds = *NETLINK_CREDS(skb);
Patrick McHardy188ccb52007-05-03 03:27:01 -07001454 if (flags & MSG_TRUNC)
Johannes Berg68d6ac62010-08-15 21:20:44 +00001455 copied = data_skb->len;
David S. Millerdaa37662010-08-15 23:21:50 -07001456
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 skb_free_datagram(sk, skb);
1458
Andrey Vaginb44d2112011-02-21 02:40:47 +00001459 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
1460 ret = netlink_dump(sk);
1461 if (ret) {
1462 sk->sk_err = ret;
1463 sk->sk_error_report(sk);
1464 }
1465 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466
1467 scm_recv(sock, msg, siocb->scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468out:
1469 netlink_rcv_wake(sk);
1470 return err ? : copied;
1471}
1472
1473static void netlink_data_ready(struct sock *sk, int len)
1474{
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001475 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476}
1477
1478/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001479 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 * complete set of kernel non-blocking support for message
1481 * queueing.
1482 */
1483
1484struct sock *
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001485netlink_kernel_create(struct net *net, int unit, unsigned int groups,
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001486 void (*input)(struct sk_buff *skb),
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001487 struct mutex *cb_mutex, struct module *module)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488{
1489 struct socket *sock;
1490 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001491 struct netlink_sock *nlk;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001492 struct listeners *listeners = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
Akinobu Mitafab2caf2006-08-29 02:15:24 -07001494 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001496 if (unit < 0 || unit >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 return NULL;
1498
1499 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1500 return NULL;
1501
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08001502 /*
1503 * We have to just have a reference on the net from sk, but don't
1504 * get_net it. Besides, we cannot get and then put the net here.
1505 * So we create one inside init_net and the move it to net.
1506 */
1507
1508 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
1509 goto out_sock_release_nosk;
1510
1511 sk = sock->sk;
Denis V. Lunevedf02082008-02-29 11:18:32 -08001512 sk_change_net(sk, net);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001513
Patrick McHardy4277a082006-03-20 18:52:01 -08001514 if (groups < 32)
1515 groups = 32;
1516
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001517 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
Patrick McHardy4277a082006-03-20 18:52:01 -08001518 if (!listeners)
1519 goto out_sock_release;
1520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 sk->sk_data_ready = netlink_data_ready;
1522 if (input)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001523 nlk_sk(sk)->netlink_rcv = input;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001525 if (netlink_insert(sk, net, 0))
Patrick McHardy77247bb2005-08-14 19:27:13 -07001526 goto out_sock_release;
1527
1528 nlk = nlk_sk(sk);
1529 nlk->flags |= NETLINK_KERNEL_SOCKET;
1530
1531 netlink_table_grab();
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001532 if (!nl_table[unit].registered) {
1533 nl_table[unit].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001534 rcu_assign_pointer(nl_table[unit].listeners, listeners);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001535 nl_table[unit].cb_mutex = cb_mutex;
1536 nl_table[unit].module = module;
1537 nl_table[unit].registered = 1;
Jesper Juhlf937f1f462007-10-15 01:39:12 -07001538 } else {
1539 kfree(listeners);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001540 nl_table[unit].registered++;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001541 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07001542 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001543 return sk;
1544
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001545out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08001546 kfree(listeners);
Denis V. Lunev9dfbec12008-02-29 11:17:56 -08001547 netlink_kernel_release(sk);
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08001548 return NULL;
1549
1550out_sock_release_nosk:
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001551 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001552 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001554EXPORT_SYMBOL(netlink_kernel_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001556
1557void
1558netlink_kernel_release(struct sock *sk)
1559{
Denis V. Lunevedf02082008-02-29 11:18:32 -08001560 sk_release_kernel(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08001561}
1562EXPORT_SYMBOL(netlink_kernel_release);
1563
Johannes Bergd136f1b2009-09-12 03:03:15 +00001564int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001565{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001566 struct listeners *new, *old;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001567 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001568
1569 if (groups < 32)
1570 groups = 32;
1571
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001572 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001573 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
1574 if (!new)
Johannes Bergd136f1b2009-09-12 03:03:15 +00001575 return -ENOMEM;
Eric Dumazet33d480c2011-08-11 19:30:52 +00001576 old = rcu_dereference_protected(tbl->listeners, 1);
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001577 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
1578 rcu_assign_pointer(tbl->listeners, new);
1579
Lai Jiangshan37b6b932011-03-15 18:01:42 +08001580 kfree_rcu(old, rcu);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001581 }
1582 tbl->groups = groups;
1583
Johannes Bergd136f1b2009-09-12 03:03:15 +00001584 return 0;
1585}
1586
1587/**
1588 * netlink_change_ngroups - change number of multicast groups
1589 *
1590 * This changes the number of multicast groups that are available
1591 * on a certain netlink family. Note that it is not possible to
1592 * change the number of groups to below 32. Also note that it does
1593 * not implicitly call netlink_clear_multicast_users() when the
1594 * number of groups is reduced.
1595 *
1596 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1597 * @groups: The new number of groups.
1598 */
1599int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1600{
1601 int err;
1602
1603 netlink_table_grab();
1604 err = __netlink_change_ngroups(sk, groups);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001605 netlink_table_ungrab();
Johannes Bergd136f1b2009-09-12 03:03:15 +00001606
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001607 return err;
1608}
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001609
Johannes Bergb8273572009-09-24 15:44:05 -07001610void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1611{
1612 struct sock *sk;
1613 struct hlist_node *node;
1614 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1615
1616 sk_for_each_bound(sk, node, &tbl->mc_list)
1617 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1618}
1619
Johannes Berg84659eb2007-07-18 15:47:05 -07001620/**
1621 * netlink_clear_multicast_users - kick off multicast listeners
1622 *
1623 * This function removes all listeners from the given group.
1624 * @ksk: The kernel netlink socket, as returned by
1625 * netlink_kernel_create().
1626 * @group: The multicast group to clear.
1627 */
1628void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1629{
Johannes Berg84659eb2007-07-18 15:47:05 -07001630 netlink_table_grab();
Johannes Bergb8273572009-09-24 15:44:05 -07001631 __netlink_clear_multicast_users(ksk, group);
Johannes Berg84659eb2007-07-18 15:47:05 -07001632 netlink_table_ungrab();
1633}
Johannes Berg84659eb2007-07-18 15:47:05 -07001634
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635void netlink_set_nonroot(int protocol, unsigned int flags)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001636{
1637 if ((unsigned int)protocol < MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 nl_table[protocol].nl_nonroot = flags;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001639}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001640EXPORT_SYMBOL(netlink_set_nonroot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
1642static void netlink_destroy_callback(struct netlink_callback *cb)
1643{
Wei Yongjun91744f62009-02-25 00:34:41 +00001644 kfree_skb(cb->skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 kfree(cb);
1646}
1647
1648/*
1649 * It looks a bit ugly.
1650 * It would be better to create kernel thread.
1651 */
1652
1653static int netlink_dump(struct sock *sk)
1654{
1655 struct netlink_sock *nlk = nlk_sk(sk);
1656 struct netlink_callback *cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00001657 struct sk_buff *skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001659 int len, err = -ENOBUFS;
Greg Rosec7ac8672011-06-10 01:27:09 +00001660 int alloc_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001662 mutex_lock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663
1664 cb = nlk->cb;
1665 if (cb == NULL) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001666 err = -EINVAL;
1667 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 }
1669
Greg Rosec7ac8672011-06-10 01:27:09 +00001670 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
1671
1672 skb = sock_rmalloc(sk, alloc_size, 0, GFP_KERNEL);
1673 if (!skb)
Dan Carpenterc63d6ea2011-06-15 03:11:42 +00001674 goto errout_skb;
Greg Rosec7ac8672011-06-10 01:27:09 +00001675
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 len = cb->dump(skb, cb);
1677
1678 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001679 mutex_unlock(nlk->cb_mutex);
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001680
1681 if (sk_filter(sk, skb))
1682 kfree_skb(skb);
1683 else {
1684 skb_queue_tail(&sk->sk_receive_queue, skb);
1685 sk->sk_data_ready(sk, skb->len);
1686 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 return 0;
1688 }
1689
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001690 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1691 if (!nlh)
1692 goto errout_skb;
1693
Johannes Berg670dc282011-06-20 13:40:46 +02001694 nl_dump_check_consistent(cb, nlh);
1695
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001696 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1697
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001698 if (sk_filter(sk, skb))
1699 kfree_skb(skb);
1700 else {
1701 skb_queue_tail(&sk->sk_receive_queue, skb);
1702 sk->sk_data_ready(sk, skb->len);
1703 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704
Thomas Grafa8f74b22005-11-10 02:25:52 +01001705 if (cb->done)
1706 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 nlk->cb = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001708 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
1710 netlink_destroy_callback(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07001712
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001713errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001714 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001715 kfree_skb(skb);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001716 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717}
1718
1719int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
Patrick McHardy3a6c2b42009-08-25 16:07:40 +02001720 const struct nlmsghdr *nlh,
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001721 int (*dump)(struct sk_buff *skb,
1722 struct netlink_callback *),
Greg Rosec7ac8672011-06-10 01:27:09 +00001723 int (*done)(struct netlink_callback *),
1724 u16 min_dump_alloc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
1726 struct netlink_callback *cb;
1727 struct sock *sk;
1728 struct netlink_sock *nlk;
Andrey Vaginb44d2112011-02-21 02:40:47 +00001729 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07001731 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 if (cb == NULL)
1733 return -ENOBUFS;
1734
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 cb->dump = dump;
1736 cb->done = done;
1737 cb->nlh = nlh;
Greg Rosec7ac8672011-06-10 01:27:09 +00001738 cb->min_dump_alloc = min_dump_alloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 atomic_inc(&skb->users);
1740 cb->skb = skb;
1741
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001742 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (sk == NULL) {
1744 netlink_destroy_callback(cb);
1745 return -ECONNREFUSED;
1746 }
1747 nlk = nlk_sk(sk);
Herbert Xu3f660d62007-05-03 03:17:14 -07001748 /* A dump is in progress... */
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001749 mutex_lock(nlk->cb_mutex);
Herbert Xu3f660d62007-05-03 03:17:14 -07001750 if (nlk->cb) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001751 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 netlink_destroy_callback(cb);
1753 sock_put(sk);
1754 return -EBUSY;
1755 }
1756 nlk->cb = cb;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001757 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
Andrey Vaginb44d2112011-02-21 02:40:47 +00001759 ret = netlink_dump(sk);
1760
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 sock_put(sk);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001762
Andrey Vaginb44d2112011-02-21 02:40:47 +00001763 if (ret)
1764 return ret;
1765
Denis V. Lunev5c582982007-10-23 20:29:25 -07001766 /* We successfully started a dump, by returning -EINTR we
1767 * signal not to send ACK even if it was requested.
1768 */
1769 return -EINTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001771EXPORT_SYMBOL(netlink_dump_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
1773void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1774{
1775 struct sk_buff *skb;
1776 struct nlmsghdr *rep;
1777 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08001778 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Thomas Graf339bf982006-11-10 14:10:15 -08001780 /* error messages get the original request appened */
1781 if (err)
1782 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
Thomas Graf339bf982006-11-10 14:10:15 -08001784 skb = nlmsg_new(payload, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 if (!skb) {
1786 struct sock *sk;
1787
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001788 sk = netlink_lookup(sock_net(in_skb->sk),
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001789 in_skb->sk->sk_protocol,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 NETLINK_CB(in_skb).pid);
1791 if (sk) {
1792 sk->sk_err = ENOBUFS;
1793 sk->sk_error_report(sk);
1794 sock_put(sk);
1795 }
1796 return;
1797 }
1798
1799 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
John Fastabend5dba93a2009-09-25 13:11:44 +00001800 NLMSG_ERROR, payload, 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001801 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07001803 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1805}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001806EXPORT_SYMBOL(netlink_ack);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001808int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001809 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01001810{
Thomas Graf82ace472005-11-10 02:25:53 +01001811 struct nlmsghdr *nlh;
1812 int err;
1813
1814 while (skb->len >= nlmsg_total_size(0)) {
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001815 int msglen;
1816
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07001817 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07001818 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01001819
Martin Murrayad8e4b72006-01-10 13:02:29 -08001820 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01001821 return 0;
1822
Thomas Grafd35b6852007-03-22 23:28:46 -07001823 /* Only requests are handled by the kernel */
1824 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
Denis V. Lunev5c582982007-10-23 20:29:25 -07001825 goto ack;
Thomas Grafd35b6852007-03-22 23:28:46 -07001826
Thomas Graf45e7ae72007-03-22 23:29:10 -07001827 /* Skip control messages */
1828 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
Denis V. Lunev5c582982007-10-23 20:29:25 -07001829 goto ack;
Thomas Graf45e7ae72007-03-22 23:29:10 -07001830
Thomas Graf1d00a4e2007-03-22 23:30:12 -07001831 err = cb(skb, nlh);
Denis V. Lunev5c582982007-10-23 20:29:25 -07001832 if (err == -EINTR)
1833 goto skip;
1834
1835ack:
Thomas Grafd35b6852007-03-22 23:28:46 -07001836 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01001837 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01001838
Denis V. Lunev5c582982007-10-23 20:29:25 -07001839skip:
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001840 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001841 if (msglen > skb->len)
1842 msglen = skb->len;
1843 skb_pull(skb, msglen);
Thomas Graf82ace472005-11-10 02:25:53 +01001844 }
1845
1846 return 0;
1847}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001848EXPORT_SYMBOL(netlink_rcv_skb);
Thomas Graf82ace472005-11-10 02:25:53 +01001849
1850/**
Thomas Grafd387f6a2006-08-15 00:31:06 -07001851 * nlmsg_notify - send a notification netlink message
1852 * @sk: netlink socket to use
1853 * @skb: notification message
1854 * @pid: destination netlink pid for reports or 0
1855 * @group: destination multicast group or 0
1856 * @report: 1 to report back, 0 to disable
1857 * @flags: allocation flags
1858 */
1859int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid,
1860 unsigned int group, int report, gfp_t flags)
1861{
1862 int err = 0;
1863
1864 if (group) {
1865 int exclude_pid = 0;
1866
1867 if (report) {
1868 atomic_inc(&skb->users);
1869 exclude_pid = pid;
1870 }
1871
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08001872 /* errors reported via destination sk->sk_err, but propagate
1873 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
1874 err = nlmsg_multicast(sk, skb, exclude_pid, group, flags);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001875 }
1876
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08001877 if (report) {
1878 int err2;
1879
1880 err2 = nlmsg_unicast(sk, skb, pid);
1881 if (!err || err == -ESRCH)
1882 err = err2;
1883 }
Thomas Grafd387f6a2006-08-15 00:31:06 -07001884
1885 return err;
1886}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001887EXPORT_SYMBOL(nlmsg_notify);
Thomas Grafd387f6a2006-08-15 00:31:06 -07001888
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889#ifdef CONFIG_PROC_FS
1890struct nl_seq_iter {
Denis V. Luneve372c412007-11-19 22:31:54 -08001891 struct seq_net_private p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 int link;
1893 int hash_idx;
1894};
1895
1896static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1897{
1898 struct nl_seq_iter *iter = seq->private;
1899 int i, j;
1900 struct sock *s;
1901 struct hlist_node *node;
1902 loff_t off = 0;
1903
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001904 for (i = 0; i < MAX_LINKS; i++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 struct nl_pid_hash *hash = &nl_table[i].hash;
1906
1907 for (j = 0; j <= hash->mask; j++) {
1908 sk_for_each(s, node, &hash->table[j]) {
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09001909 if (sock_net(s) != seq_file_net(seq))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001910 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 if (off == pos) {
1912 iter->link = i;
1913 iter->hash_idx = j;
1914 return s;
1915 }
1916 ++off;
1917 }
1918 }
1919 }
1920 return NULL;
1921}
1922
1923static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001924 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925{
1926 read_lock(&nl_table_lock);
1927 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1928}
1929
1930static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1931{
1932 struct sock *s;
1933 struct nl_seq_iter *iter;
1934 int i, j;
1935
1936 ++*pos;
1937
1938 if (v == SEQ_START_TOKEN)
1939 return netlink_seq_socket_idx(seq, 0);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001940
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001941 iter = seq->private;
1942 s = v;
1943 do {
1944 s = sk_next(s);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09001945 } while (s && sock_net(s) != seq_file_net(seq));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 if (s)
1947 return s;
1948
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 i = iter->link;
1950 j = iter->hash_idx + 1;
1951
1952 do {
1953 struct nl_pid_hash *hash = &nl_table[i].hash;
1954
1955 for (; j <= hash->mask; j++) {
1956 s = sk_head(&hash->table[j]);
YOSHIFUJI Hideaki12188542008-03-26 02:36:06 +09001957 while (s && sock_net(s) != seq_file_net(seq))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001958 s = sk_next(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 if (s) {
1960 iter->link = i;
1961 iter->hash_idx = j;
1962 return s;
1963 }
1964 }
1965
1966 j = 0;
1967 } while (++i < MAX_LINKS);
1968
1969 return NULL;
1970}
1971
1972static void netlink_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08001973 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974{
1975 read_unlock(&nl_table_lock);
1976}
1977
1978
1979static int netlink_seq_show(struct seq_file *seq, void *v)
1980{
1981 if (v == SEQ_START_TOKEN)
1982 seq_puts(seq,
1983 "sk Eth Pid Groups "
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00001984 "Rmem Wmem Dump Locks Drops Inode\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 else {
1986 struct sock *s = v;
1987 struct netlink_sock *nlk = nlk_sk(s);
1988
Dan Rosenberg71338aa2011-05-23 12:17:35 +00001989 seq_printf(seq, "%pK %-3d %-6d %08x %-8d %-8d %pK %-8d %-8d %-8lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 s,
1991 s->sk_protocol,
1992 nlk->pid,
Patrick McHardy513c2502005-09-06 15:43:59 -07001993 nlk->groups ? (u32)nlk->groups[0] : 0,
Eric Dumazet31e6d362009-06-17 19:05:41 -07001994 sk_rmem_alloc_get(s),
1995 sk_wmem_alloc_get(s),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 nlk->cb,
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07001997 atomic_read(&s->sk_refcnt),
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00001998 atomic_read(&s->sk_drops),
1999 sock_i_ino(s)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 );
2001
2002 }
2003 return 0;
2004}
2005
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002006static const struct seq_operations netlink_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 .start = netlink_seq_start,
2008 .next = netlink_seq_next,
2009 .stop = netlink_seq_stop,
2010 .show = netlink_seq_show,
2011};
2012
2013
2014static int netlink_seq_open(struct inode *inode, struct file *file)
2015{
Denis V. Luneve372c412007-11-19 22:31:54 -08002016 return seq_open_net(inode, file, &netlink_seq_ops,
2017 sizeof(struct nl_seq_iter));
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002018}
2019
Arjan van de Venda7071d2007-02-12 00:55:36 -08002020static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 .owner = THIS_MODULE,
2022 .open = netlink_seq_open,
2023 .read = seq_read,
2024 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002025 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026};
2027
2028#endif
2029
2030int netlink_register_notifier(struct notifier_block *nb)
2031{
Alan Sterne041c682006-03-27 01:16:30 -08002032 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002034EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035
2036int netlink_unregister_notifier(struct notifier_block *nb)
2037{
Alan Sterne041c682006-03-27 01:16:30 -08002038 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002040EXPORT_SYMBOL(netlink_unregister_notifier);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002041
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08002042static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 .family = PF_NETLINK,
2044 .owner = THIS_MODULE,
2045 .release = netlink_release,
2046 .bind = netlink_bind,
2047 .connect = netlink_connect,
2048 .socketpair = sock_no_socketpair,
2049 .accept = sock_no_accept,
2050 .getname = netlink_getname,
2051 .poll = datagram_poll,
2052 .ioctl = sock_no_ioctl,
2053 .listen = sock_no_listen,
2054 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002055 .setsockopt = netlink_setsockopt,
2056 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 .sendmsg = netlink_sendmsg,
2058 .recvmsg = netlink_recvmsg,
2059 .mmap = sock_no_mmap,
2060 .sendpage = sock_no_sendpage,
2061};
2062
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002063static const struct net_proto_family netlink_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 .family = PF_NETLINK,
2065 .create = netlink_create,
2066 .owner = THIS_MODULE, /* for consistency 8) */
2067};
2068
Pavel Emelyanov46650792007-10-08 20:38:39 -07002069static int __net_init netlink_net_init(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002070{
2071#ifdef CONFIG_PROC_FS
2072 if (!proc_net_fops_create(net, "netlink", 0, &netlink_seq_fops))
2073 return -ENOMEM;
2074#endif
2075 return 0;
2076}
2077
Pavel Emelyanov46650792007-10-08 20:38:39 -07002078static void __net_exit netlink_net_exit(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002079{
2080#ifdef CONFIG_PROC_FS
2081 proc_net_remove(net, "netlink");
2082#endif
2083}
2084
David S. Millerb963ea82010-08-30 19:08:01 -07002085static void __init netlink_add_usersock_entry(void)
2086{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002087 struct listeners *listeners;
David S. Millerb963ea82010-08-30 19:08:01 -07002088 int groups = 32;
2089
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002090 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
David S. Millerb963ea82010-08-30 19:08:01 -07002091 if (!listeners)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002092 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
David S. Millerb963ea82010-08-30 19:08:01 -07002093
2094 netlink_table_grab();
2095
2096 nl_table[NETLINK_USERSOCK].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002097 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
David S. Millerb963ea82010-08-30 19:08:01 -07002098 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
2099 nl_table[NETLINK_USERSOCK].registered = 1;
2100
2101 netlink_table_ungrab();
2102}
2103
Denis V. Lunev022cbae2007-11-13 03:23:50 -08002104static struct pernet_operations __net_initdata netlink_net_ops = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002105 .init = netlink_net_init,
2106 .exit = netlink_net_exit,
2107};
2108
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109static int __init netlink_proto_init(void)
2110{
2111 struct sk_buff *dummy_skb;
2112 int i;
Denis Cheng26ff5dd2007-09-16 16:36:02 -07002113 unsigned long limit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 unsigned int order;
2115 int err = proto_register(&netlink_proto, 0);
2116
2117 if (err != 0)
2118 goto out;
2119
YOSHIFUJI Hideakief047f52006-09-01 00:29:06 -07002120 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07002122 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002123 if (!nl_table)
2124 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125
Jan Beulich44813742009-09-21 17:03:05 -07002126 if (totalram_pages >= (128 * 1024))
2127 limit = totalram_pages >> (21 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 else
Jan Beulich44813742009-09-21 17:03:05 -07002129 limit = totalram_pages >> (23 - PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130
Denis Cheng26ff5dd2007-09-16 16:36:02 -07002131 order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
2132 limit = (1UL << order) / sizeof(struct hlist_head);
2133 order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134
2135 for (i = 0; i < MAX_LINKS; i++) {
2136 struct nl_pid_hash *hash = &nl_table[i].hash;
2137
Eric Dumazetea729122007-12-11 02:09:47 -08002138 hash->table = nl_pid_hash_zalloc(1 * sizeof(*hash->table));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 if (!hash->table) {
2140 while (i-- > 0)
2141 nl_pid_hash_free(nl_table[i].hash.table,
2142 1 * sizeof(*hash->table));
2143 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002144 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 hash->max_shift = order;
2147 hash->shift = 0;
2148 hash->mask = 0;
2149 hash->rehash_time = jiffies;
2150 }
2151
David S. Millerb963ea82010-08-30 19:08:01 -07002152 netlink_add_usersock_entry();
2153
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 sock_register(&netlink_family_ops);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002155 register_pernet_subsys(&netlink_net_ops);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002156 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 rtnetlink_init();
2158out:
2159 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002160panic:
2161 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162}
2163
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164core_initcall(netlink_proto_init);