blob: 4753dc083423f62f8702243189b8ffd938aa8bc0 [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>
Patrick McHardycd1df522013-04-17 06:47:05 +00006 * Patrick McHardy <kaber@trash.net>
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +090012 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
14 * added netlink_proto_exit
15 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
16 * use nlk_sk, as sk->protinfo is on a diet 8)
Harald Welte4fdb3bb2005-08-09 19:40:55 -070017 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
18 * - inc module use count of module that owns
19 * the kernel socket in case userspace opens
20 * socket of same protocol
21 * - remove all module support, since netlink is
22 * mandatory if CONFIG_NET=y these days
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 */
24
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/module.h>
26
Randy Dunlap4fc268d2006-01-11 12:17:47 -080027#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <linux/kernel.h>
29#include <linux/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/signal.h>
31#include <linux/sched.h>
32#include <linux/errno.h>
33#include <linux/string.h>
34#include <linux/stat.h>
35#include <linux/socket.h>
36#include <linux/un.h>
37#include <linux/fcntl.h>
38#include <linux/termios.h>
39#include <linux/sockios.h>
40#include <linux/net.h>
41#include <linux/fs.h>
42#include <linux/slab.h>
43#include <asm/uaccess.h>
44#include <linux/skbuff.h>
45#include <linux/netdevice.h>
46#include <linux/rtnetlink.h>
47#include <linux/proc_fs.h>
48#include <linux/seq_file.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#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>
Patrick McHardyaf65bdf2007-04-20 14:14:21 -070058#include <linux/mutex.h>
Patrick McHardyccdfcc32013-04-17 06:47:01 +000059#include <linux/vmalloc.h>
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +020060#include <linux/if_arp.h>
Thomas Grafe3416942014-08-02 11:47:45 +020061#include <linux/rhashtable.h>
Patrick McHardy9652e932013-04-17 06:47:02 +000062#include <asm/cacheflush.h>
Thomas Grafe3416942014-08-02 11:47:45 +020063#include <linux/hash.h>
Johannes Bergee1c24422015-01-16 11:37:14 +010064#include <linux/genetlink.h>
Andrew Morton54e0f522005-04-30 07:07:04 +010065
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020066#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <net/sock.h>
68#include <net/scm.h>
Thomas Graf82ace472005-11-10 02:25:53 +010069#include <net/netlink.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
Andrey Vagin0f29c762013-03-21 20:33:47 +040071#include "af_netlink.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Eric Dumazet5c398dc2010-10-24 04:27:10 +000073struct listeners {
74 struct rcu_head rcu;
75 unsigned long masks[0];
Johannes Berg6c04bb12009-07-10 09:51:32 +000076};
77
Patrick McHardycd967e02013-04-17 06:46:56 +000078/* state bits */
79#define NETLINK_CONGESTED 0x0
80
81/* flags */
Patrick McHardy77247bb2005-08-14 19:27:13 -070082#define NETLINK_KERNEL_SOCKET 0x1
Patrick McHardy9a4595b2005-08-15 12:32:15 -070083#define NETLINK_RECV_PKTINFO 0x2
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +000084#define NETLINK_BROADCAST_SEND_ERROR 0x4
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -070085#define NETLINK_RECV_NO_ENOBUFS 0x8
Patrick McHardy77247bb2005-08-14 19:27:13 -070086
David S. Miller035c4c12011-12-23 17:33:03 -050087static inline int netlink_is_kernel(struct sock *sk)
Denis V. Lunevaed81562007-10-10 21:14:32 -070088{
89 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
90}
91
Eric Dumazet91dd93f2015-05-12 17:24:50 -070092struct netlink_table *nl_table __read_mostly;
Andrey Vagin0f29c762013-03-21 20:33:47 +040093EXPORT_SYMBOL_GPL(nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
95static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
96
97static int netlink_dump(struct sock *sk);
Patrick McHardy9652e932013-04-17 06:47:02 +000098static void netlink_skb_destructor(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Thomas Graf78fd1d02014-10-21 22:05:38 +0200100/* nl_table locking explained:
Thomas Graf21e49022015-01-02 23:00:22 +0100101 * Lookup and traversal are protected with an RCU read-side lock. Insertion
Ying Xuec5adde92015-01-12 14:52:23 +0800102 * and removal are protected with per bucket lock while using RCU list
Thomas Graf21e49022015-01-02 23:00:22 +0100103 * modification primitives and may run in parallel to RCU protected lookups.
104 * Destruction of the Netlink socket may only occur *after* nl_table_lock has
105 * been acquired * either during or after the socket has been removed from
106 * the list and after an RCU grace period.
Thomas Graf78fd1d02014-10-21 22:05:38 +0200107 */
Andrey Vagin0f29c762013-03-21 20:33:47 +0400108DEFINE_RWLOCK(nl_table_lock);
109EXPORT_SYMBOL_GPL(nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110static atomic_t nl_table_users = ATOMIC_INIT(0);
111
Eric Dumazet6d772ac2012-10-18 03:21:55 +0000112#define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
113
Alan Sterne041c682006-03-27 01:16:30 -0800114static ATOMIC_NOTIFIER_HEAD(netlink_chain);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200116static DEFINE_SPINLOCK(netlink_tap_lock);
117static struct list_head netlink_tap_all __read_mostly;
118
Herbert Xuc428ecd2015-03-20 21:57:01 +1100119static const struct rhashtable_params netlink_rhashtable_params;
120
stephen hemmingerb57ef81f2011-12-22 08:52:02 +0000121static inline u32 netlink_group_mask(u32 group)
Patrick McHardyd629b832005-08-14 19:27:50 -0700122{
123 return group ? 1 << (group - 1) : 0;
124}
125
Daniel Borkmann65d48c62015-09-10 20:05:46 +0200126static struct sk_buff *netlink_to_full_skb(const struct sk_buff *skb,
127 gfp_t gfp_mask)
128{
129 unsigned int len = skb_end_offset(skb);
130 struct sk_buff *new;
131
132 new = alloc_skb(len, gfp_mask);
133 if (new == NULL)
134 return NULL;
135
136 NETLINK_CB(new).portid = NETLINK_CB(skb).portid;
137 NETLINK_CB(new).dst_group = NETLINK_CB(skb).dst_group;
138 NETLINK_CB(new).creds = NETLINK_CB(skb).creds;
139
140 memcpy(skb_put(new, len), skb->data, len);
141 return new;
142}
143
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200144int netlink_add_tap(struct netlink_tap *nt)
145{
146 if (unlikely(nt->dev->type != ARPHRD_NETLINK))
147 return -EINVAL;
148
149 spin_lock(&netlink_tap_lock);
150 list_add_rcu(&nt->list, &netlink_tap_all);
151 spin_unlock(&netlink_tap_lock);
152
Markus Elfringfcd4d352014-11-18 21:03:13 +0100153 __module_get(nt->module);
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200154
155 return 0;
156}
157EXPORT_SYMBOL_GPL(netlink_add_tap);
158
stephen hemminger2173f8d2013-12-30 10:49:22 -0800159static int __netlink_remove_tap(struct netlink_tap *nt)
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200160{
161 bool found = false;
162 struct netlink_tap *tmp;
163
164 spin_lock(&netlink_tap_lock);
165
166 list_for_each_entry(tmp, &netlink_tap_all, list) {
167 if (nt == tmp) {
168 list_del_rcu(&nt->list);
169 found = true;
170 goto out;
171 }
172 }
173
174 pr_warn("__netlink_remove_tap: %p not found\n", nt);
175out:
176 spin_unlock(&netlink_tap_lock);
177
178 if (found && nt->module)
179 module_put(nt->module);
180
181 return found ? 0 : -ENODEV;
182}
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200183
184int netlink_remove_tap(struct netlink_tap *nt)
185{
186 int ret;
187
188 ret = __netlink_remove_tap(nt);
189 synchronize_net();
190
191 return ret;
192}
193EXPORT_SYMBOL_GPL(netlink_remove_tap);
194
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200195static bool netlink_filter_tap(const struct sk_buff *skb)
196{
197 struct sock *sk = skb->sk;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200198
199 /* We take the more conservative approach and
200 * whitelist socket protocols that may pass.
201 */
202 switch (sk->sk_protocol) {
203 case NETLINK_ROUTE:
204 case NETLINK_USERSOCK:
205 case NETLINK_SOCK_DIAG:
206 case NETLINK_NFLOG:
207 case NETLINK_XFRM:
208 case NETLINK_FIB_LOOKUP:
209 case NETLINK_NETFILTER:
210 case NETLINK_GENERIC:
Varka Bhadram498044b2014-07-16 10:59:47 +0530211 return true;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200212 }
213
Varka Bhadram498044b2014-07-16 10:59:47 +0530214 return false;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200215}
216
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200217static int __netlink_deliver_tap_skb(struct sk_buff *skb,
218 struct net_device *dev)
219{
220 struct sk_buff *nskb;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200221 struct sock *sk = skb->sk;
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200222 int ret = -ENOMEM;
223
224 dev_hold(dev);
Daniel Borkmann65d48c62015-09-10 20:05:46 +0200225
226 if (netlink_skb_is_mmaped(skb) || is_vmalloc_addr(skb->head))
227 nskb = netlink_to_full_skb(skb, GFP_ATOMIC);
228 else
229 nskb = skb_clone(skb, GFP_ATOMIC);
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200230 if (nskb) {
231 nskb->dev = dev;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200232 nskb->protocol = htons((u16) sk->sk_protocol);
Daniel Borkmann604d13c2013-12-23 14:35:56 +0100233 nskb->pkt_type = netlink_is_kernel(sk) ?
234 PACKET_KERNEL : PACKET_USER;
Daniel Borkmann4e48ed82014-08-07 22:22:47 +0200235 skb_reset_network_header(nskb);
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200236 ret = dev_queue_xmit(nskb);
237 if (unlikely(ret > 0))
238 ret = net_xmit_errno(ret);
239 }
240
241 dev_put(dev);
242 return ret;
243}
244
245static void __netlink_deliver_tap(struct sk_buff *skb)
246{
247 int ret;
248 struct netlink_tap *tmp;
249
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200250 if (!netlink_filter_tap(skb))
251 return;
252
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200253 list_for_each_entry_rcu(tmp, &netlink_tap_all, list) {
254 ret = __netlink_deliver_tap_skb(skb, tmp->dev);
255 if (unlikely(ret))
256 break;
257 }
258}
259
260static void netlink_deliver_tap(struct sk_buff *skb)
261{
262 rcu_read_lock();
263
264 if (unlikely(!list_empty(&netlink_tap_all)))
265 __netlink_deliver_tap(skb);
266
267 rcu_read_unlock();
268}
269
Daniel Borkmann73bfd372013-12-23 14:35:55 +0100270static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src,
271 struct sk_buff *skb)
272{
273 if (!(netlink_is_kernel(dst) && netlink_is_kernel(src)))
274 netlink_deliver_tap(skb);
275}
276
Patrick McHardycd1df522013-04-17 06:47:05 +0000277static void netlink_overrun(struct sock *sk)
278{
279 struct netlink_sock *nlk = nlk_sk(sk);
280
281 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
282 if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) {
283 sk->sk_err = ENOBUFS;
284 sk->sk_error_report(sk);
285 }
286 }
287 atomic_inc(&sk->sk_drops);
288}
289
290static void netlink_rcv_wake(struct sock *sk)
291{
292 struct netlink_sock *nlk = nlk_sk(sk);
293
294 if (skb_queue_empty(&sk->sk_receive_queue))
295 clear_bit(NETLINK_CONGESTED, &nlk->state);
296 if (!test_bit(NETLINK_CONGESTED, &nlk->state))
297 wake_up_interruptible(&nlk->wait);
298}
299
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000300#ifdef CONFIG_NETLINK_MMAP
Patrick McHardyf9c22882013-04-17 06:47:04 +0000301static bool netlink_rx_is_mmaped(struct sock *sk)
302{
303 return nlk_sk(sk)->rx_ring.pg_vec != NULL;
304}
305
Patrick McHardy5fd96122013-04-17 06:47:03 +0000306static bool netlink_tx_is_mmaped(struct sock *sk)
307{
308 return nlk_sk(sk)->tx_ring.pg_vec != NULL;
309}
310
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000311static __pure struct page *pgvec_to_page(const void *addr)
312{
313 if (is_vmalloc_addr(addr))
314 return vmalloc_to_page(addr);
315 else
316 return virt_to_page(addr);
317}
318
319static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len)
320{
321 unsigned int i;
322
323 for (i = 0; i < len; i++) {
324 if (pg_vec[i] != NULL) {
325 if (is_vmalloc_addr(pg_vec[i]))
326 vfree(pg_vec[i]);
327 else
328 free_pages((unsigned long)pg_vec[i], order);
329 }
330 }
331 kfree(pg_vec);
332}
333
334static void *alloc_one_pg_vec_page(unsigned long order)
335{
336 void *buffer;
337 gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO |
338 __GFP_NOWARN | __GFP_NORETRY;
339
340 buffer = (void *)__get_free_pages(gfp_flags, order);
341 if (buffer != NULL)
342 return buffer;
343
344 buffer = vzalloc((1 << order) * PAGE_SIZE);
345 if (buffer != NULL)
346 return buffer;
347
348 gfp_flags &= ~__GFP_NORETRY;
349 return (void *)__get_free_pages(gfp_flags, order);
350}
351
352static void **alloc_pg_vec(struct netlink_sock *nlk,
353 struct nl_mmap_req *req, unsigned int order)
354{
355 unsigned int block_nr = req->nm_block_nr;
356 unsigned int i;
Daniel Borkmann8a849bb2013-08-02 17:32:39 +0200357 void **pg_vec;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000358
359 pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL);
360 if (pg_vec == NULL)
361 return NULL;
362
363 for (i = 0; i < block_nr; i++) {
Daniel Borkmann8a849bb2013-08-02 17:32:39 +0200364 pg_vec[i] = alloc_one_pg_vec_page(order);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000365 if (pg_vec[i] == NULL)
366 goto err1;
367 }
368
369 return pg_vec;
370err1:
371 free_pg_vec(pg_vec, order, block_nr);
372 return NULL;
373}
374
Florian Westphalb265c302015-07-21 16:33:50 +0200375
376static void
377__netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, bool tx_ring, void **pg_vec,
378 unsigned int order)
379{
380 struct netlink_sock *nlk = nlk_sk(sk);
381 struct sk_buff_head *queue;
382 struct netlink_ring *ring;
383
384 queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue;
385 ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring;
386
387 spin_lock_bh(&queue->lock);
388
389 ring->frame_max = req->nm_frame_nr - 1;
390 ring->head = 0;
391 ring->frame_size = req->nm_frame_size;
392 ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE;
393
394 swap(ring->pg_vec_len, req->nm_block_nr);
395 swap(ring->pg_vec_order, order);
396 swap(ring->pg_vec, pg_vec);
397
398 __skb_queue_purge(queue);
399 spin_unlock_bh(&queue->lock);
400
401 WARN_ON(atomic_read(&nlk->mapped));
402
403 if (pg_vec)
404 free_pg_vec(pg_vec, order, req->nm_block_nr);
405}
406
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000407static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req,
Florian Westphalb265c302015-07-21 16:33:50 +0200408 bool tx_ring)
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000409{
410 struct netlink_sock *nlk = nlk_sk(sk);
411 struct netlink_ring *ring;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000412 void **pg_vec = NULL;
413 unsigned int order = 0;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000414
415 ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000416
Florian Westphalb265c302015-07-21 16:33:50 +0200417 if (atomic_read(&nlk->mapped))
418 return -EBUSY;
419 if (atomic_read(&ring->pending))
420 return -EBUSY;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000421
422 if (req->nm_block_nr) {
423 if (ring->pg_vec != NULL)
424 return -EBUSY;
425
426 if ((int)req->nm_block_size <= 0)
427 return -EINVAL;
Tobias Klauser74e83b22014-07-31 12:17:08 +0200428 if (!PAGE_ALIGNED(req->nm_block_size))
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000429 return -EINVAL;
430 if (req->nm_frame_size < NL_MMAP_HDRLEN)
431 return -EINVAL;
432 if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT))
433 return -EINVAL;
434
435 ring->frames_per_block = req->nm_block_size /
436 req->nm_frame_size;
437 if (ring->frames_per_block == 0)
438 return -EINVAL;
439 if (ring->frames_per_block * req->nm_block_nr !=
440 req->nm_frame_nr)
441 return -EINVAL;
442
443 order = get_order(req->nm_block_size);
444 pg_vec = alloc_pg_vec(nlk, req, order);
445 if (pg_vec == NULL)
446 return -ENOMEM;
447 } else {
448 if (req->nm_frame_nr)
449 return -EINVAL;
450 }
451
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000452 mutex_lock(&nlk->pg_vec_lock);
Florian Westphalb265c302015-07-21 16:33:50 +0200453 if (atomic_read(&nlk->mapped) == 0) {
454 __netlink_set_ring(sk, req, tx_ring, pg_vec, order);
455 mutex_unlock(&nlk->pg_vec_lock);
456 return 0;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000457 }
Florian Westphalb265c302015-07-21 16:33:50 +0200458
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000459 mutex_unlock(&nlk->pg_vec_lock);
460
461 if (pg_vec)
462 free_pg_vec(pg_vec, order, req->nm_block_nr);
Florian Westphalb265c302015-07-21 16:33:50 +0200463
464 return -EBUSY;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000465}
466
467static void netlink_mm_open(struct vm_area_struct *vma)
468{
469 struct file *file = vma->vm_file;
470 struct socket *sock = file->private_data;
471 struct sock *sk = sock->sk;
472
473 if (sk)
474 atomic_inc(&nlk_sk(sk)->mapped);
475}
476
477static void netlink_mm_close(struct vm_area_struct *vma)
478{
479 struct file *file = vma->vm_file;
480 struct socket *sock = file->private_data;
481 struct sock *sk = sock->sk;
482
483 if (sk)
484 atomic_dec(&nlk_sk(sk)->mapped);
485}
486
487static const struct vm_operations_struct netlink_mmap_ops = {
488 .open = netlink_mm_open,
489 .close = netlink_mm_close,
490};
491
492static int netlink_mmap(struct file *file, struct socket *sock,
493 struct vm_area_struct *vma)
494{
495 struct sock *sk = sock->sk;
496 struct netlink_sock *nlk = nlk_sk(sk);
497 struct netlink_ring *ring;
498 unsigned long start, size, expected;
499 unsigned int i;
500 int err = -EINVAL;
501
502 if (vma->vm_pgoff)
503 return -EINVAL;
504
505 mutex_lock(&nlk->pg_vec_lock);
506
507 expected = 0;
508 for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) {
509 if (ring->pg_vec == NULL)
510 continue;
511 expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE;
512 }
513
514 if (expected == 0)
515 goto out;
516
517 size = vma->vm_end - vma->vm_start;
518 if (size != expected)
519 goto out;
520
521 start = vma->vm_start;
522 for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) {
523 if (ring->pg_vec == NULL)
524 continue;
525
526 for (i = 0; i < ring->pg_vec_len; i++) {
527 struct page *page;
528 void *kaddr = ring->pg_vec[i];
529 unsigned int pg_num;
530
531 for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) {
532 page = pgvec_to_page(kaddr);
533 err = vm_insert_page(vma, start, page);
534 if (err < 0)
535 goto out;
536 start += PAGE_SIZE;
537 kaddr += PAGE_SIZE;
538 }
539 }
540 }
541
542 atomic_inc(&nlk->mapped);
543 vma->vm_ops = &netlink_mmap_ops;
544 err = 0;
545out:
546 mutex_unlock(&nlk->pg_vec_lock);
Patrick McHardy7cdbac72013-06-11 02:52:47 -0700547 return err;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000548}
Patrick McHardy9652e932013-04-17 06:47:02 +0000549
David Miller4682a032014-12-16 17:58:17 -0500550static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len)
Patrick McHardy9652e932013-04-17 06:47:02 +0000551{
552#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1
553 struct page *p_start, *p_end;
554
555 /* First page is flushed through netlink_{get,set}_status */
556 p_start = pgvec_to_page(hdr + PAGE_SIZE);
David Miller4682a032014-12-16 17:58:17 -0500557 p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1);
Patrick McHardy9652e932013-04-17 06:47:02 +0000558 while (p_start <= p_end) {
559 flush_dcache_page(p_start);
560 p_start++;
561 }
562#endif
563}
564
565static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr)
566{
567 smp_rmb();
568 flush_dcache_page(pgvec_to_page(hdr));
569 return hdr->nm_status;
570}
571
572static void netlink_set_status(struct nl_mmap_hdr *hdr,
573 enum nl_mmap_status status)
574{
Thomas Grafa18e6a12014-12-18 10:30:26 +0000575 smp_mb();
Patrick McHardy9652e932013-04-17 06:47:02 +0000576 hdr->nm_status = status;
577 flush_dcache_page(pgvec_to_page(hdr));
Patrick McHardy9652e932013-04-17 06:47:02 +0000578}
579
580static struct nl_mmap_hdr *
581__netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos)
582{
583 unsigned int pg_vec_pos, frame_off;
584
585 pg_vec_pos = pos / ring->frames_per_block;
586 frame_off = pos % ring->frames_per_block;
587
588 return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size);
589}
590
591static struct nl_mmap_hdr *
592netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos,
593 enum nl_mmap_status status)
594{
595 struct nl_mmap_hdr *hdr;
596
597 hdr = __netlink_lookup_frame(ring, pos);
598 if (netlink_get_status(hdr) != status)
599 return NULL;
600
601 return hdr;
602}
603
604static struct nl_mmap_hdr *
605netlink_current_frame(const struct netlink_ring *ring,
606 enum nl_mmap_status status)
607{
608 return netlink_lookup_frame(ring, ring->head, status);
609}
610
611static struct nl_mmap_hdr *
612netlink_previous_frame(const struct netlink_ring *ring,
613 enum nl_mmap_status status)
614{
615 unsigned int prev;
616
617 prev = ring->head ? ring->head - 1 : ring->frame_max;
618 return netlink_lookup_frame(ring, prev, status);
619}
620
621static void netlink_increment_head(struct netlink_ring *ring)
622{
623 ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0;
624}
625
626static void netlink_forward_ring(struct netlink_ring *ring)
627{
628 unsigned int head = ring->head, pos = head;
629 const struct nl_mmap_hdr *hdr;
630
631 do {
632 hdr = __netlink_lookup_frame(ring, pos);
633 if (hdr->nm_status == NL_MMAP_STATUS_UNUSED)
634 break;
635 if (hdr->nm_status != NL_MMAP_STATUS_SKIP)
636 break;
637 netlink_increment_head(ring);
638 } while (ring->head != head);
639}
640
Patrick McHardycd1df522013-04-17 06:47:05 +0000641static bool netlink_dump_space(struct netlink_sock *nlk)
642{
643 struct netlink_ring *ring = &nlk->rx_ring;
644 struct nl_mmap_hdr *hdr;
645 unsigned int n;
646
647 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
648 if (hdr == NULL)
649 return false;
650
651 n = ring->head + ring->frame_max / 2;
652 if (n > ring->frame_max)
653 n -= ring->frame_max;
654
655 hdr = __netlink_lookup_frame(ring, n);
656
657 return hdr->nm_status == NL_MMAP_STATUS_UNUSED;
658}
659
Patrick McHardy9652e932013-04-17 06:47:02 +0000660static unsigned int netlink_poll(struct file *file, struct socket *sock,
661 poll_table *wait)
662{
663 struct sock *sk = sock->sk;
664 struct netlink_sock *nlk = nlk_sk(sk);
665 unsigned int mask;
Patrick McHardycd1df522013-04-17 06:47:05 +0000666 int err;
Patrick McHardy9652e932013-04-17 06:47:02 +0000667
Patrick McHardycd1df522013-04-17 06:47:05 +0000668 if (nlk->rx_ring.pg_vec != NULL) {
669 /* Memory mapped sockets don't call recvmsg(), so flow control
670 * for dumps is performed here. A dump is allowed to continue
671 * if at least half the ring is unused.
672 */
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700673 while (nlk->cb_running && netlink_dump_space(nlk)) {
Patrick McHardycd1df522013-04-17 06:47:05 +0000674 err = netlink_dump(sk);
675 if (err < 0) {
Ben Pfaffac30ef82014-07-09 10:31:22 -0700676 sk->sk_err = -err;
Patrick McHardycd1df522013-04-17 06:47:05 +0000677 sk->sk_error_report(sk);
678 break;
679 }
680 }
681 netlink_rcv_wake(sk);
682 }
Patrick McHardy5fd96122013-04-17 06:47:03 +0000683
Patrick McHardy9652e932013-04-17 06:47:02 +0000684 mask = datagram_poll(file, sock, wait);
685
686 spin_lock_bh(&sk->sk_receive_queue.lock);
687 if (nlk->rx_ring.pg_vec) {
688 netlink_forward_ring(&nlk->rx_ring);
689 if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED))
690 mask |= POLLIN | POLLRDNORM;
691 }
692 spin_unlock_bh(&sk->sk_receive_queue.lock);
693
694 spin_lock_bh(&sk->sk_write_queue.lock);
695 if (nlk->tx_ring.pg_vec) {
696 if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED))
697 mask |= POLLOUT | POLLWRNORM;
698 }
699 spin_unlock_bh(&sk->sk_write_queue.lock);
700
701 return mask;
702}
703
704static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb)
705{
706 return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN);
707}
708
709static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk,
710 struct netlink_ring *ring,
711 struct nl_mmap_hdr *hdr)
712{
713 unsigned int size;
714 void *data;
715
716 size = ring->frame_size - NL_MMAP_HDRLEN;
717 data = (void *)hdr + NL_MMAP_HDRLEN;
718
719 skb->head = data;
720 skb->data = data;
721 skb_reset_tail_pointer(skb);
722 skb->end = skb->tail + size;
723 skb->len = 0;
724
725 skb->destructor = netlink_skb_destructor;
726 NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED;
727 NETLINK_CB(skb).sk = sk;
728}
Patrick McHardy5fd96122013-04-17 06:47:03 +0000729
730static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg,
731 u32 dst_portid, u32 dst_group,
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100732 struct scm_cookie *scm)
Patrick McHardy5fd96122013-04-17 06:47:03 +0000733{
734 struct netlink_sock *nlk = nlk_sk(sk);
735 struct netlink_ring *ring;
736 struct nl_mmap_hdr *hdr;
737 struct sk_buff *skb;
738 unsigned int maxlen;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000739 int err = 0, len = 0;
740
Patrick McHardy5fd96122013-04-17 06:47:03 +0000741 mutex_lock(&nlk->pg_vec_lock);
742
743 ring = &nlk->tx_ring;
744 maxlen = ring->frame_size - NL_MMAP_HDRLEN;
745
746 do {
David Miller4682a032014-12-16 17:58:17 -0500747 unsigned int nm_len;
748
Patrick McHardy5fd96122013-04-17 06:47:03 +0000749 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID);
750 if (hdr == NULL) {
751 if (!(msg->msg_flags & MSG_DONTWAIT) &&
752 atomic_read(&nlk->tx_ring.pending))
753 schedule();
754 continue;
755 }
David Miller4682a032014-12-16 17:58:17 -0500756
757 nm_len = ACCESS_ONCE(hdr->nm_len);
758 if (nm_len > maxlen) {
Patrick McHardy5fd96122013-04-17 06:47:03 +0000759 err = -EINVAL;
760 goto out;
761 }
762
David Miller4682a032014-12-16 17:58:17 -0500763 netlink_frame_flush_dcache(hdr, nm_len);
Patrick McHardy5fd96122013-04-17 06:47:03 +0000764
David Miller4682a032014-12-16 17:58:17 -0500765 skb = alloc_skb(nm_len, GFP_KERNEL);
766 if (skb == NULL) {
767 err = -ENOBUFS;
768 goto out;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000769 }
David Miller4682a032014-12-16 17:58:17 -0500770 __skb_put(skb, nm_len);
771 memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len);
772 netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED);
Patrick McHardy5fd96122013-04-17 06:47:03 +0000773
774 netlink_increment_head(ring);
775
776 NETLINK_CB(skb).portid = nlk->portid;
777 NETLINK_CB(skb).dst_group = dst_group;
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100778 NETLINK_CB(skb).creds = scm->creds;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000779
780 err = security_netlink_send(sk, skb);
781 if (err) {
782 kfree_skb(skb);
783 goto out;
784 }
785
786 if (unlikely(dst_group)) {
787 atomic_inc(&skb->users);
788 netlink_broadcast(sk, skb, dst_portid, dst_group,
789 GFP_KERNEL);
790 }
791 err = netlink_unicast(sk, skb, dst_portid,
792 msg->msg_flags & MSG_DONTWAIT);
793 if (err < 0)
794 goto out;
795 len += err;
796
797 } while (hdr != NULL ||
798 (!(msg->msg_flags & MSG_DONTWAIT) &&
799 atomic_read(&nlk->tx_ring.pending)));
800
801 if (len > 0)
802 err = len;
803out:
804 mutex_unlock(&nlk->pg_vec_lock);
805 return err;
806}
Patrick McHardyf9c22882013-04-17 06:47:04 +0000807
808static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb)
809{
810 struct nl_mmap_hdr *hdr;
811
812 hdr = netlink_mmap_hdr(skb);
813 hdr->nm_len = skb->len;
814 hdr->nm_group = NETLINK_CB(skb).dst_group;
815 hdr->nm_pid = NETLINK_CB(skb).creds.pid;
Nicolas Dichtel1bf93102013-04-24 10:36:23 +0200816 hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid);
817 hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid);
David Miller4682a032014-12-16 17:58:17 -0500818 netlink_frame_flush_dcache(hdr, hdr->nm_len);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000819 netlink_set_status(hdr, NL_MMAP_STATUS_VALID);
820
821 NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED;
822 kfree_skb(skb);
823}
824
825static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb)
826{
827 struct netlink_sock *nlk = nlk_sk(sk);
828 struct netlink_ring *ring = &nlk->rx_ring;
829 struct nl_mmap_hdr *hdr;
830
831 spin_lock_bh(&sk->sk_receive_queue.lock);
832 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
833 if (hdr == NULL) {
834 spin_unlock_bh(&sk->sk_receive_queue.lock);
835 kfree_skb(skb);
Patrick McHardycd1df522013-04-17 06:47:05 +0000836 netlink_overrun(sk);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000837 return;
838 }
839 netlink_increment_head(ring);
840 __skb_queue_tail(&sk->sk_receive_queue, skb);
841 spin_unlock_bh(&sk->sk_receive_queue.lock);
842
843 hdr->nm_len = skb->len;
844 hdr->nm_group = NETLINK_CB(skb).dst_group;
845 hdr->nm_pid = NETLINK_CB(skb).creds.pid;
Nicolas Dichtel1bf93102013-04-24 10:36:23 +0200846 hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid);
847 hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000848 netlink_set_status(hdr, NL_MMAP_STATUS_COPY);
849}
850
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000851#else /* CONFIG_NETLINK_MMAP */
Patrick McHardyf9c22882013-04-17 06:47:04 +0000852#define netlink_rx_is_mmaped(sk) false
Patrick McHardy5fd96122013-04-17 06:47:03 +0000853#define netlink_tx_is_mmaped(sk) false
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000854#define netlink_mmap sock_no_mmap
Patrick McHardy9652e932013-04-17 06:47:02 +0000855#define netlink_poll datagram_poll
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100856#define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000857#endif /* CONFIG_NETLINK_MMAP */
858
Patrick McHardycf0a0182013-04-17 06:47:00 +0000859static void netlink_skb_destructor(struct sk_buff *skb)
860{
Patrick McHardy9652e932013-04-17 06:47:02 +0000861#ifdef CONFIG_NETLINK_MMAP
862 struct nl_mmap_hdr *hdr;
863 struct netlink_ring *ring;
864 struct sock *sk;
865
866 /* If a packet from the kernel to userspace was freed because of an
867 * error without being delivered to userspace, the kernel must reset
868 * the status. In the direction userspace to kernel, the status is
869 * always reset here after the packet was processed and freed.
870 */
871 if (netlink_skb_is_mmaped(skb)) {
872 hdr = netlink_mmap_hdr(skb);
873 sk = NETLINK_CB(skb).sk;
874
Patrick McHardy5fd96122013-04-17 06:47:03 +0000875 if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) {
876 netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED);
877 ring = &nlk_sk(sk)->tx_ring;
878 } else {
879 if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) {
880 hdr->nm_len = 0;
881 netlink_set_status(hdr, NL_MMAP_STATUS_VALID);
882 }
883 ring = &nlk_sk(sk)->rx_ring;
Patrick McHardy9652e932013-04-17 06:47:02 +0000884 }
Patrick McHardy9652e932013-04-17 06:47:02 +0000885
886 WARN_ON(atomic_read(&ring->pending) == 0);
887 atomic_dec(&ring->pending);
888 sock_put(sk);
889
Pablo Neira5e71d9d2013-06-03 09:28:43 +0000890 skb->head = NULL;
Patrick McHardy9652e932013-04-17 06:47:02 +0000891 }
892#endif
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +0000893 if (is_vmalloc_addr(skb->head)) {
Pablo Neira3a365152013-06-28 03:04:23 +0200894 if (!skb->cloned ||
895 !atomic_dec_return(&(skb_shinfo(skb)->dataref)))
896 vfree(skb->head);
897
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +0000898 skb->head = NULL;
899 }
Patrick McHardy9652e932013-04-17 06:47:02 +0000900 if (skb->sk != NULL)
901 sock_rfree(skb);
Patrick McHardycf0a0182013-04-17 06:47:00 +0000902}
903
904static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
905{
906 WARN_ON(skb->sk != NULL);
907 skb->sk = sk;
908 skb->destructor = netlink_skb_destructor;
909 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
910 sk_mem_charge(sk, skb->truesize);
911}
912
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913static void netlink_sock_destruct(struct sock *sk)
914{
Herbert Xu3f660d62007-05-03 03:17:14 -0700915 struct netlink_sock *nlk = nlk_sk(sk);
916
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700917 if (nlk->cb_running) {
918 if (nlk->cb.done)
919 nlk->cb.done(&nlk->cb);
Gao feng6dc878a2012-10-04 20:15:48 +0000920
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700921 module_put(nlk->cb.module);
922 kfree_skb(nlk->cb.skb);
Herbert Xu3f660d62007-05-03 03:17:14 -0700923 }
924
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 skb_queue_purge(&sk->sk_receive_queue);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000926#ifdef CONFIG_NETLINK_MMAP
927 if (1) {
928 struct nl_mmap_req req;
929
930 memset(&req, 0, sizeof(req));
931 if (nlk->rx_ring.pg_vec)
Florian Westphalb265c302015-07-21 16:33:50 +0200932 __netlink_set_ring(sk, &req, false, NULL, 0);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000933 memset(&req, 0, sizeof(req));
934 if (nlk->tx_ring.pg_vec)
Florian Westphalb265c302015-07-21 16:33:50 +0200935 __netlink_set_ring(sk, &req, true, NULL, 0);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000936 }
937#endif /* CONFIG_NETLINK_MMAP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939 if (!sock_flag(sk, SOCK_DEAD)) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800940 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 return;
942 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700943
944 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
945 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
946 WARN_ON(nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947}
948
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800949/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
950 * SMP. Look, when several writers sleep and reader wakes them up, all but one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
952 * this, _but_ remember, it adds useless work on UP machines.
953 */
954
Johannes Bergd136f1b2009-09-12 03:03:15 +0000955void netlink_table_grab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800956 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957{
Johannes Bergd136f1b2009-09-12 03:03:15 +0000958 might_sleep();
959
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700960 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
962 if (atomic_read(&nl_table_users)) {
963 DECLARE_WAITQUEUE(wait, current);
964
965 add_wait_queue_exclusive(&nl_table_wait, &wait);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800966 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 set_current_state(TASK_UNINTERRUPTIBLE);
968 if (atomic_read(&nl_table_users) == 0)
969 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700970 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700972 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 }
974
975 __set_current_state(TASK_RUNNING);
976 remove_wait_queue(&nl_table_wait, &wait);
977 }
978}
979
Johannes Bergd136f1b2009-09-12 03:03:15 +0000980void netlink_table_ungrab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800981 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700983 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 wake_up(&nl_table_wait);
985}
986
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800987static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988netlink_lock_table(void)
989{
990 /* read_lock() synchronizes us to netlink_table_grab */
991
992 read_lock(&nl_table_lock);
993 atomic_inc(&nl_table_users);
994 read_unlock(&nl_table_lock);
995}
996
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800997static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998netlink_unlock_table(void)
999{
1000 if (atomic_dec_and_test(&nl_table_users))
1001 wake_up(&nl_table_wait);
1002}
1003
Thomas Grafe3416942014-08-02 11:47:45 +02001004struct netlink_compare_arg
Gao fengda12c902013-06-06 14:49:11 +08001005{
Herbert Xuc428ecd2015-03-20 21:57:01 +11001006 possible_net_t pnet;
Thomas Grafe3416942014-08-02 11:47:45 +02001007 u32 portid;
1008};
1009
Herbert Xu8f2ddaa2015-03-21 14:14:03 +11001010/* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
1011#define netlink_compare_arg_len \
1012 (offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
Thomas Grafe3416942014-08-02 11:47:45 +02001013
Herbert Xuc428ecd2015-03-20 21:57:01 +11001014static inline int netlink_compare(struct rhashtable_compare_arg *arg,
1015 const void *ptr)
1016{
1017 const struct netlink_compare_arg *x = arg->key;
1018 const struct netlink_sock *nlk = ptr;
1019
Herbert Xu4e277622015-09-18 19:16:50 +08001020 return nlk->rhash_portid != x->portid ||
Herbert Xuc428ecd2015-03-20 21:57:01 +11001021 !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet));
1022}
1023
1024static void netlink_compare_arg_init(struct netlink_compare_arg *arg,
1025 struct net *net, u32 portid)
1026{
1027 memset(arg, 0, sizeof(*arg));
1028 write_pnet(&arg->pnet, net);
1029 arg->portid = portid;
Thomas Grafe3416942014-08-02 11:47:45 +02001030}
1031
1032static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
1033 struct net *net)
1034{
Herbert Xuc428ecd2015-03-20 21:57:01 +11001035 struct netlink_compare_arg arg;
Thomas Grafe3416942014-08-02 11:47:45 +02001036
Herbert Xuc428ecd2015-03-20 21:57:01 +11001037 netlink_compare_arg_init(&arg, net, portid);
1038 return rhashtable_lookup_fast(&table->hash, &arg,
1039 netlink_rhashtable_params);
Gao fengda12c902013-06-06 14:49:11 +08001040}
1041
Herbert Xuc428ecd2015-03-20 21:57:01 +11001042static int __netlink_insert(struct netlink_table *table, struct sock *sk)
Ying Xuec5adde92015-01-12 14:52:23 +08001043{
Herbert Xuc428ecd2015-03-20 21:57:01 +11001044 struct netlink_compare_arg arg;
Ying Xuec5adde92015-01-12 14:52:23 +08001045
Herbert Xu4e277622015-09-18 19:16:50 +08001046 netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->rhash_portid);
Herbert Xuc428ecd2015-03-20 21:57:01 +11001047 return rhashtable_lookup_insert_key(&table->hash, &arg,
1048 &nlk_sk(sk)->node,
1049 netlink_rhashtable_params);
Ying Xuec5adde92015-01-12 14:52:23 +08001050}
1051
Eric W. Biederman15e47302012-09-07 20:12:54 +00001052static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
Gao fengda12c902013-06-06 14:49:11 +08001054 struct netlink_table *table = &nl_table[protocol];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
Thomas Grafe3416942014-08-02 11:47:45 +02001057 rcu_read_lock();
1058 sk = __netlink_lookup(table, portid, net);
1059 if (sk)
1060 sock_hold(sk);
1061 rcu_read_unlock();
1062
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 return sk;
1064}
1065
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001066static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
Patrick McHardy4277a082006-03-20 18:52:01 -08001068static void
1069netlink_update_listeners(struct sock *sk)
1070{
1071 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Patrick McHardy4277a082006-03-20 18:52:01 -08001072 unsigned long mask;
1073 unsigned int i;
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001074 struct listeners *listeners;
1075
1076 listeners = nl_deref_protected(tbl->listeners);
1077 if (!listeners)
1078 return;
Patrick McHardy4277a082006-03-20 18:52:01 -08001079
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001080 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
Patrick McHardy4277a082006-03-20 18:52:01 -08001081 mask = 0;
Sasha Levinb67bfe02013-02-27 17:06:00 -08001082 sk_for_each_bound(sk, &tbl->mc_list) {
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001083 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
1084 mask |= nlk_sk(sk)->groups[i];
1085 }
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001086 listeners->masks[i] = mask;
Patrick McHardy4277a082006-03-20 18:52:01 -08001087 }
1088 /* this function is only called with the netlink table "grabbed", which
1089 * makes sure updates are visible before bind or setsockopt return. */
1090}
1091
Herbert Xu8ea65f42015-01-26 14:02:56 +11001092static int netlink_insert(struct sock *sk, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093{
Gao fengda12c902013-06-06 14:49:11 +08001094 struct netlink_table *table = &nl_table[sk->sk_protocol];
Herbert Xu919d9db2015-01-16 17:23:48 +11001095 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
Ying Xuec5adde92015-01-12 14:52:23 +08001097 lock_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
1099 err = -EBUSY;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001100 if (nlk_sk(sk)->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 goto err;
1102
1103 err = -ENOMEM;
Thomas Graf97defe12015-01-02 23:00:20 +01001104 if (BITS_PER_LONG > 32 &&
1105 unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 goto err;
1107
Herbert Xu4e277622015-09-18 19:16:50 +08001108 nlk_sk(sk)->rhash_portid = portid;
Thomas Grafe3416942014-08-02 11:47:45 +02001109 sock_hold(sk);
Herbert Xu919d9db2015-01-16 17:23:48 +11001110
Herbert Xuc428ecd2015-03-20 21:57:01 +11001111 err = __netlink_insert(table, sk);
1112 if (err) {
Daniel Borkmannd3976172015-08-07 00:26:41 +02001113 /* In case the hashtable backend returns with -EBUSY
1114 * from here, it must not escape to the caller.
1115 */
1116 if (unlikely(err == -EBUSY))
1117 err = -EOVERFLOW;
Herbert Xuc428ecd2015-03-20 21:57:01 +11001118 if (err == -EEXIST)
1119 err = -EADDRINUSE;
Ying Xuec5adde92015-01-12 14:52:23 +08001120 sock_put(sk);
Herbert Xu919d9db2015-01-16 17:23:48 +11001121 }
1122
Herbert Xu4e277622015-09-18 19:16:50 +08001123 nlk_sk(sk)->portid = portid;
1124
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125err:
Ying Xuec5adde92015-01-12 14:52:23 +08001126 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 return err;
1128}
1129
1130static void netlink_remove(struct sock *sk)
1131{
Thomas Grafe3416942014-08-02 11:47:45 +02001132 struct netlink_table *table;
1133
Thomas Grafe3416942014-08-02 11:47:45 +02001134 table = &nl_table[sk->sk_protocol];
Herbert Xuc428ecd2015-03-20 21:57:01 +11001135 if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node,
1136 netlink_rhashtable_params)) {
Thomas Grafe3416942014-08-02 11:47:45 +02001137 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
1138 __sock_put(sk);
1139 }
Thomas Grafe3416942014-08-02 11:47:45 +02001140
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 netlink_table_grab();
Johannes Bergb10dcb32014-12-22 18:56:37 +01001142 if (nlk_sk(sk)->subscriptions) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 __sk_del_bind_node(sk);
Johannes Bergb10dcb32014-12-22 18:56:37 +01001144 netlink_update_listeners(sk);
1145 }
Johannes Bergee1c24422015-01-16 11:37:14 +01001146 if (sk->sk_protocol == NETLINK_GENERIC)
1147 atomic_inc(&genl_sk_destructing_cnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 netlink_table_ungrab();
1149}
1150
1151static struct proto netlink_proto = {
1152 .name = "NETLINK",
1153 .owner = THIS_MODULE,
1154 .obj_size = sizeof(struct netlink_sock),
1155};
1156
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001157static int __netlink_create(struct net *net, struct socket *sock,
1158 struct mutex *cb_mutex, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159{
1160 struct sock *sk;
1161 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -07001162
1163 sock->ops = &netlink_ops;
1164
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001165 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
Patrick McHardyab33a172005-08-14 19:31:36 -07001166 if (!sk)
1167 return -ENOMEM;
1168
1169 sock_init_data(sock, sk);
1170
1171 nlk = nlk_sk(sk);
Eric Dumazet658cb352012-04-22 21:30:21 +00001172 if (cb_mutex) {
Patrick McHardyffa4d722007-04-25 14:01:17 -07001173 nlk->cb_mutex = cb_mutex;
Eric Dumazet658cb352012-04-22 21:30:21 +00001174 } else {
Patrick McHardyffa4d722007-04-25 14:01:17 -07001175 nlk->cb_mutex = &nlk->cb_def_mutex;
1176 mutex_init(nlk->cb_mutex);
1177 }
Patrick McHardyab33a172005-08-14 19:31:36 -07001178 init_waitqueue_head(&nlk->wait);
Patrick McHardyccdfcc32013-04-17 06:47:01 +00001179#ifdef CONFIG_NETLINK_MMAP
1180 mutex_init(&nlk->pg_vec_lock);
1181#endif
Patrick McHardyab33a172005-08-14 19:31:36 -07001182
1183 sk->sk_destruct = netlink_sock_destruct;
1184 sk->sk_protocol = protocol;
1185 return 0;
1186}
1187
Eric Paris3f378b62009-11-05 22:18:14 -08001188static int netlink_create(struct net *net, struct socket *sock, int protocol,
1189 int kern)
Patrick McHardyab33a172005-08-14 19:31:36 -07001190{
1191 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001192 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001193 struct netlink_sock *nlk;
Johannes Berg023e2cf2014-12-23 21:00:06 +01001194 int (*bind)(struct net *net, int group);
1195 void (*unbind)(struct net *net, int group);
Patrick McHardyab33a172005-08-14 19:31:36 -07001196 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
1198 sock->state = SS_UNCONNECTED;
1199
1200 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
1201 return -ESOCKTNOSUPPORT;
1202
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001203 if (protocol < 0 || protocol >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 return -EPROTONOSUPPORT;
1205
Patrick McHardy77247bb2005-08-14 19:27:13 -07001206 netlink_lock_table();
Johannes Berg95a5afc2008-10-16 15:24:51 -07001207#ifdef CONFIG_MODULES
Patrick McHardyab33a172005-08-14 19:31:36 -07001208 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -07001209 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001210 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001211 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001212 }
Patrick McHardyab33a172005-08-14 19:31:36 -07001213#endif
1214 if (nl_table[protocol].registered &&
1215 try_module_get(nl_table[protocol].module))
1216 module = nl_table[protocol].module;
Alexey Dobriyan974c37e2010-01-30 10:05:05 +00001217 else
1218 err = -EPROTONOSUPPORT;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001219 cb_mutex = nl_table[protocol].cb_mutex;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00001220 bind = nl_table[protocol].bind;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001221 unbind = nl_table[protocol].unbind;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001222 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001223
Alexey Dobriyan974c37e2010-01-30 10:05:05 +00001224 if (err < 0)
1225 goto out;
1226
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001227 err = __netlink_create(net, sock, cb_mutex, protocol);
1228 if (err < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -07001229 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230
David S. Miller6f756a82008-11-23 17:34:03 -08001231 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -08001232 sock_prot_inuse_add(net, &netlink_proto, 1);
David S. Miller6f756a82008-11-23 17:34:03 -08001233 local_bh_enable();
1234
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001235 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001236 nlk->module = module;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00001237 nlk->netlink_bind = bind;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001238 nlk->netlink_unbind = unbind;
Patrick McHardyab33a172005-08-14 19:31:36 -07001239out:
1240 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
Patrick McHardyab33a172005-08-14 19:31:36 -07001242out_module:
1243 module_put(module);
1244 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245}
1246
Thomas Graf21e49022015-01-02 23:00:22 +01001247static void deferred_put_nlk_sk(struct rcu_head *head)
1248{
1249 struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu);
1250
1251 sock_put(&nlk->sk);
1252}
1253
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254static int netlink_release(struct socket *sock)
1255{
1256 struct sock *sk = sock->sk;
1257 struct netlink_sock *nlk;
1258
1259 if (!sk)
1260 return 0;
1261
1262 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -07001263 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 nlk = nlk_sk(sk);
1265
Herbert Xu3f660d62007-05-03 03:17:14 -07001266 /*
1267 * OK. Socket is unlinked, any packets that arrive now
1268 * will be purged.
1269 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
Johannes Bergee1c24422015-01-16 11:37:14 +01001271 /* must not acquire netlink_table_lock in any way again before unbind
1272 * and notifying genetlink is done as otherwise it might deadlock
1273 */
1274 if (nlk->netlink_unbind) {
1275 int i;
1276
1277 for (i = 0; i < nlk->ngroups; i++)
1278 if (test_bit(i, nlk->groups))
1279 nlk->netlink_unbind(sock_net(sk), i + 1);
1280 }
1281 if (sk->sk_protocol == NETLINK_GENERIC &&
1282 atomic_dec_return(&genl_sk_destructing_cnt) == 0)
1283 wake_up(&genl_sk_destructing_waitq);
1284
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 sock->sk = NULL;
1286 wake_up_interruptible_all(&nlk->wait);
1287
1288 skb_queue_purge(&sk->sk_write_queue);
1289
Eric W. Biederman15e47302012-09-07 20:12:54 +00001290 if (nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 struct netlink_notify n = {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001292 .net = sock_net(sk),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 .protocol = sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00001294 .portid = nlk->portid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 };
Alan Sterne041c682006-03-27 01:16:30 -08001296 atomic_notifier_call_chain(&netlink_chain,
1297 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001298 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001299
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -08001300 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001301
Denis V. Lunevaed81562007-10-10 21:14:32 -07001302 if (netlink_is_kernel(sk)) {
Johannes Bergb10dcb32014-12-22 18:56:37 +01001303 netlink_table_grab();
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001304 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
1305 if (--nl_table[sk->sk_protocol].registered == 0) {
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001306 struct listeners *old;
1307
1308 old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
1309 RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
1310 kfree_rcu(old, rcu);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001311 nl_table[sk->sk_protocol].module = NULL;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001312 nl_table[sk->sk_protocol].bind = NULL;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001313 nl_table[sk->sk_protocol].unbind = NULL;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001314 nl_table[sk->sk_protocol].flags = 0;
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001315 nl_table[sk->sk_protocol].registered = 0;
1316 }
Johannes Bergb10dcb32014-12-22 18:56:37 +01001317 netlink_table_ungrab();
Eric Dumazet658cb352012-04-22 21:30:21 +00001318 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07001319
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001320 kfree(nlk->groups);
1321 nlk->groups = NULL;
1322
Eric Dumazet37558102008-11-24 14:05:22 -08001323 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -08001324 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
Eric Dumazet37558102008-11-24 14:05:22 -08001325 local_bh_enable();
Thomas Graf21e49022015-01-02 23:00:22 +01001326 call_rcu(&nlk->rcu, deferred_put_nlk_sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 return 0;
1328}
1329
1330static int netlink_autobind(struct socket *sock)
1331{
1332 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001333 struct net *net = sock_net(sk);
Gao fengda12c902013-06-06 14:49:11 +08001334 struct netlink_table *table = &nl_table[sk->sk_protocol];
Eric W. Biederman15e47302012-09-07 20:12:54 +00001335 s32 portid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 int err;
1337 static s32 rover = -4097;
1338
1339retry:
1340 cond_resched();
Thomas Grafe3416942014-08-02 11:47:45 +02001341 rcu_read_lock();
1342 if (__netlink_lookup(table, portid, net)) {
1343 /* Bind collision, search negative portid values. */
1344 portid = rover--;
1345 if (rover > -4097)
1346 rover = -4097;
1347 rcu_read_unlock();
1348 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 }
Thomas Grafe3416942014-08-02 11:47:45 +02001350 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
Herbert Xu8ea65f42015-01-26 14:02:56 +11001352 err = netlink_insert(sk, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 if (err == -EADDRINUSE)
1354 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -07001355
1356 /* If 2 threads race to autobind, that is fine. */
1357 if (err == -EBUSY)
1358 err = 0;
1359
1360 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361}
1362
Eric W. Biedermanaa4cf942014-04-23 14:28:03 -07001363/**
1364 * __netlink_ns_capable - General netlink message capability test
1365 * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
1366 * @user_ns: The user namespace of the capability to use
1367 * @cap: The capability to use
1368 *
1369 * Test to see if the opener of the socket we received the message
1370 * from had when the netlink socket was created and the sender of the
1371 * message has has the capability @cap in the user namespace @user_ns.
1372 */
1373bool __netlink_ns_capable(const struct netlink_skb_parms *nsp,
1374 struct user_namespace *user_ns, int cap)
1375{
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07001376 return ((nsp->flags & NETLINK_SKB_DST) ||
1377 file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) &&
1378 ns_capable(user_ns, cap);
Eric W. Biedermanaa4cf942014-04-23 14:28:03 -07001379}
1380EXPORT_SYMBOL(__netlink_ns_capable);
1381
1382/**
1383 * netlink_ns_capable - General netlink message capability test
1384 * @skb: socket buffer holding a netlink command from userspace
1385 * @user_ns: The user namespace of the capability to use
1386 * @cap: The capability to use
1387 *
1388 * Test to see if the opener of the socket we received the message
1389 * from had when the netlink socket was created and the sender of the
1390 * message has has the capability @cap in the user namespace @user_ns.
1391 */
1392bool netlink_ns_capable(const struct sk_buff *skb,
1393 struct user_namespace *user_ns, int cap)
1394{
1395 return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap);
1396}
1397EXPORT_SYMBOL(netlink_ns_capable);
1398
1399/**
1400 * netlink_capable - Netlink global message capability test
1401 * @skb: socket buffer holding a netlink command from userspace
1402 * @cap: The capability to use
1403 *
1404 * Test to see if the opener of the socket we received the message
1405 * from had when the netlink socket was created and the sender of the
1406 * message has has the capability @cap in all user namespaces.
1407 */
1408bool netlink_capable(const struct sk_buff *skb, int cap)
1409{
1410 return netlink_ns_capable(skb, &init_user_ns, cap);
1411}
1412EXPORT_SYMBOL(netlink_capable);
1413
1414/**
1415 * netlink_net_capable - Netlink network namespace message capability test
1416 * @skb: socket buffer holding a netlink command from userspace
1417 * @cap: The capability to use
1418 *
1419 * Test to see if the opener of the socket we received the message
1420 * from had when the netlink socket was created and the sender of the
1421 * message has has the capability @cap over the network namespace of
1422 * the socket we received the message from.
1423 */
1424bool netlink_net_capable(const struct sk_buff *skb, int cap)
1425{
1426 return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap);
1427}
1428EXPORT_SYMBOL(netlink_net_capable);
1429
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001430static inline int netlink_allowed(const struct socket *sock, unsigned int flag)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001431{
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001432 return (nl_table[sock->sk->sk_protocol].flags & flag) ||
Eric W. Biedermandf008c92012-11-16 03:03:07 +00001433 ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001434}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001436static void
1437netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
1438{
1439 struct netlink_sock *nlk = nlk_sk(sk);
1440
1441 if (nlk->subscriptions && !subscriptions)
1442 __sk_del_bind_node(sk);
1443 else if (!nlk->subscriptions && subscriptions)
1444 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
1445 nlk->subscriptions = subscriptions;
1446}
1447
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001448static int netlink_realloc_groups(struct sock *sk)
Patrick McHardy513c2502005-09-06 15:43:59 -07001449{
1450 struct netlink_sock *nlk = nlk_sk(sk);
1451 unsigned int groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001452 unsigned long *new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -07001453 int err = 0;
1454
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001455 netlink_table_grab();
1456
Patrick McHardy513c2502005-09-06 15:43:59 -07001457 groups = nl_table[sk->sk_protocol].groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001458 if (!nl_table[sk->sk_protocol].registered) {
Patrick McHardy513c2502005-09-06 15:43:59 -07001459 err = -ENOENT;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001460 goto out_unlock;
1461 }
Patrick McHardy513c2502005-09-06 15:43:59 -07001462
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001463 if (nlk->ngroups >= groups)
1464 goto out_unlock;
Patrick McHardy513c2502005-09-06 15:43:59 -07001465
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001466 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
1467 if (new_groups == NULL) {
1468 err = -ENOMEM;
1469 goto out_unlock;
1470 }
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001471 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001472 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
1473
1474 nlk->groups = new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -07001475 nlk->ngroups = groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001476 out_unlock:
1477 netlink_table_ungrab();
1478 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -07001479}
1480
Johannes Berg02c81ab2014-12-22 18:56:35 +01001481static void netlink_undo_bind(int group, long unsigned int groups,
Johannes Berg023e2cf2014-12-23 21:00:06 +01001482 struct sock *sk)
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001483{
Johannes Berg023e2cf2014-12-23 21:00:06 +01001484 struct netlink_sock *nlk = nlk_sk(sk);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001485 int undo;
1486
1487 if (!nlk->netlink_unbind)
1488 return;
1489
1490 for (undo = 0; undo < group; undo++)
Hiroaki SHIMODA6251edd2014-11-13 04:24:10 +09001491 if (test_bit(undo, &groups))
Pablo Neira8b7c36d2015-01-29 10:51:53 +01001492 nlk->netlink_unbind(sock_net(sk), undo + 1);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001493}
1494
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001495static int netlink_bind(struct socket *sock, struct sockaddr *addr,
1496 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497{
1498 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001499 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 struct netlink_sock *nlk = nlk_sk(sk);
1501 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
1502 int err;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001503 long unsigned int groups = nladdr->nl_groups;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001504
Hannes Frederic Sowa4e4b5372012-12-15 15:42:19 +00001505 if (addr_len < sizeof(struct sockaddr_nl))
1506 return -EINVAL;
1507
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 if (nladdr->nl_family != AF_NETLINK)
1509 return -EINVAL;
1510
1511 /* Only superuser is allowed to listen multicasts */
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001512 if (groups) {
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001513 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy513c2502005-09-06 15:43:59 -07001514 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001515 err = netlink_realloc_groups(sk);
1516 if (err)
1517 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -07001518 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001520 if (nlk->portid)
Eric W. Biederman15e47302012-09-07 20:12:54 +00001521 if (nladdr->nl_pid != nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 return -EINVAL;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001523
1524 if (nlk->netlink_bind && groups) {
1525 int group;
1526
1527 for (group = 0; group < nlk->ngroups; group++) {
1528 if (!test_bit(group, &groups))
1529 continue;
Pablo Neira8b7c36d2015-01-29 10:51:53 +01001530 err = nlk->netlink_bind(net, group + 1);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001531 if (!err)
1532 continue;
Johannes Berg023e2cf2014-12-23 21:00:06 +01001533 netlink_undo_bind(group, groups, sk);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001534 return err;
1535 }
1536 }
1537
1538 if (!nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 err = nladdr->nl_pid ?
Herbert Xu8ea65f42015-01-26 14:02:56 +11001540 netlink_insert(sk, nladdr->nl_pid) :
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 netlink_autobind(sock);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001542 if (err) {
Johannes Berg023e2cf2014-12-23 21:00:06 +01001543 netlink_undo_bind(nlk->ngroups, groups, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 return err;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001545 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 }
1547
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001548 if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 return 0;
1550
1551 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001552 netlink_update_subscriptions(sk, nlk->subscriptions +
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001553 hweight32(groups) -
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001554 hweight32(nlk->groups[0]));
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001555 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups;
Patrick McHardy4277a082006-03-20 18:52:01 -08001556 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 netlink_table_ungrab();
1558
1559 return 0;
1560}
1561
1562static int netlink_connect(struct socket *sock, struct sockaddr *addr,
1563 int alen, int flags)
1564{
1565 int err = 0;
1566 struct sock *sk = sock->sk;
1567 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001568 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
Changli Gao6503d962010-03-31 22:58:26 +00001570 if (alen < sizeof(addr->sa_family))
1571 return -EINVAL;
1572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 if (addr->sa_family == AF_UNSPEC) {
1574 sk->sk_state = NETLINK_UNCONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001575 nlk->dst_portid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -07001576 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 return 0;
1578 }
1579 if (addr->sa_family != AF_NETLINK)
1580 return -EINVAL;
1581
Mike Pecovnik46833a82014-02-24 21:11:16 +01001582 if ((nladdr->nl_groups || nladdr->nl_pid) &&
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001583 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 return -EPERM;
1585
Eric W. Biederman15e47302012-09-07 20:12:54 +00001586 if (!nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 err = netlink_autobind(sock);
1588
1589 if (err == 0) {
1590 sk->sk_state = NETLINK_CONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001591 nlk->dst_portid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001592 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 }
1594
1595 return err;
1596}
1597
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001598static int netlink_getname(struct socket *sock, struct sockaddr *addr,
1599 int *addr_len, int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600{
1601 struct sock *sk = sock->sk;
1602 struct netlink_sock *nlk = nlk_sk(sk);
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001603 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001604
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 nladdr->nl_family = AF_NETLINK;
1606 nladdr->nl_pad = 0;
1607 *addr_len = sizeof(*nladdr);
1608
1609 if (peer) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001610 nladdr->nl_pid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001611 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001613 nladdr->nl_pid = nlk->portid;
Patrick McHardy513c2502005-09-06 15:43:59 -07001614 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 }
1616 return 0;
1617}
1618
Eric W. Biederman15e47302012-09-07 20:12:54 +00001619static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 struct sock *sock;
1622 struct netlink_sock *nlk;
1623
Eric W. Biederman15e47302012-09-07 20:12:54 +00001624 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 if (!sock)
1626 return ERR_PTR(-ECONNREFUSED);
1627
1628 /* Don't bother queuing skb if kernel socket has no input function */
1629 nlk = nlk_sk(sock);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001630 if (sock->sk_state == NETLINK_CONNECTED &&
Eric W. Biederman15e47302012-09-07 20:12:54 +00001631 nlk->dst_portid != nlk_sk(ssk)->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 sock_put(sock);
1633 return ERR_PTR(-ECONNREFUSED);
1634 }
1635 return sock;
1636}
1637
1638struct sock *netlink_getsockbyfilp(struct file *filp)
1639{
Al Viro496ad9a2013-01-23 17:07:38 -05001640 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 struct sock *sock;
1642
1643 if (!S_ISSOCK(inode->i_mode))
1644 return ERR_PTR(-ENOTSOCK);
1645
1646 sock = SOCKET_I(inode)->sk;
1647 if (sock->sk_family != AF_NETLINK)
1648 return ERR_PTR(-EINVAL);
1649
1650 sock_hold(sock);
1651 return sock;
1652}
1653
Pablo Neira3a365152013-06-28 03:04:23 +02001654static struct sk_buff *netlink_alloc_large_skb(unsigned int size,
1655 int broadcast)
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001656{
1657 struct sk_buff *skb;
1658 void *data;
1659
Pablo Neira3a365152013-06-28 03:04:23 +02001660 if (size <= NLMSG_GOODSIZE || broadcast)
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001661 return alloc_skb(size, GFP_KERNEL);
1662
Pablo Neira3a365152013-06-28 03:04:23 +02001663 size = SKB_DATA_ALIGN(size) +
1664 SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001665
1666 data = vmalloc(size);
1667 if (data == NULL)
Pablo Neira3a365152013-06-28 03:04:23 +02001668 return NULL;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001669
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001670 skb = __build_skb(data, size);
Pablo Neira3a365152013-06-28 03:04:23 +02001671 if (skb == NULL)
1672 vfree(data);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001673 else
Pablo Neira3a365152013-06-28 03:04:23 +02001674 skb->destructor = netlink_skb_destructor;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001675
1676 return skb;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001677}
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679/*
1680 * Attach a skb to a netlink socket.
1681 * The caller must hold a reference to the destination socket. On error, the
1682 * reference is dropped. The skb is not send to the destination, just all
1683 * all error checks are performed and memory in the queue is reserved.
1684 * Return values:
1685 * < 0: error. skb freed, reference to sock dropped.
1686 * 0: continue
1687 * 1: repeat lookup - reference dropped while waiting for socket memory.
1688 */
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001689int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001690 long *timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691{
1692 struct netlink_sock *nlk;
1693
1694 nlk = nlk_sk(sk);
1695
Patrick McHardy5fd96122013-04-17 06:47:03 +00001696 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1697 test_bit(NETLINK_CONGESTED, &nlk->state)) &&
1698 !netlink_skb_is_mmaped(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 DECLARE_WAITQUEUE(wait, current);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001700 if (!*timeo) {
Denis V. Lunevaed81562007-10-10 21:14:32 -07001701 if (!ssk || netlink_is_kernel(ssk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 netlink_overrun(sk);
1703 sock_put(sk);
1704 kfree_skb(skb);
1705 return -EAGAIN;
1706 }
1707
1708 __set_current_state(TASK_INTERRUPTIBLE);
1709 add_wait_queue(&nlk->wait, &wait);
1710
1711 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
Patrick McHardycd967e02013-04-17 06:46:56 +00001712 test_bit(NETLINK_CONGESTED, &nlk->state)) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 !sock_flag(sk, SOCK_DEAD))
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001714 *timeo = schedule_timeout(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
1716 __set_current_state(TASK_RUNNING);
1717 remove_wait_queue(&nlk->wait, &wait);
1718 sock_put(sk);
1719
1720 if (signal_pending(current)) {
1721 kfree_skb(skb);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001722 return sock_intr_errno(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 }
1724 return 1;
1725 }
Patrick McHardycf0a0182013-04-17 06:47:00 +00001726 netlink_skb_set_owner_r(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 return 0;
1728}
1729
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001730static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 int len = skb->len;
1733
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +02001734 netlink_deliver_tap(skb);
1735
Patrick McHardyf9c22882013-04-17 06:47:04 +00001736#ifdef CONFIG_NETLINK_MMAP
1737 if (netlink_skb_is_mmaped(skb))
1738 netlink_queue_mmaped_skb(sk, skb);
1739 else if (netlink_rx_is_mmaped(sk))
1740 netlink_ring_set_copied(sk, skb);
1741 else
1742#endif /* CONFIG_NETLINK_MMAP */
1743 skb_queue_tail(&sk->sk_receive_queue, skb);
David S. Miller676d2362014-04-11 16:15:36 -04001744 sk->sk_data_ready(sk);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001745 return len;
1746}
1747
1748int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1749{
1750 int len = __netlink_sendskb(sk, skb);
1751
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 sock_put(sk);
1753 return len;
1754}
1755
1756void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
1757{
1758 kfree_skb(skb);
1759 sock_put(sk);
1760}
1761
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001762static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763{
1764 int delta;
1765
Patrick McHardy1298ca42013-04-17 06:46:59 +00001766 WARN_ON(skb->sk != NULL);
Patrick McHardy5fd96122013-04-17 06:47:03 +00001767 if (netlink_skb_is_mmaped(skb))
1768 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001770 delta = skb->end - skb->tail;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001771 if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 return skb;
1773
1774 if (skb_shared(skb)) {
1775 struct sk_buff *nskb = skb_clone(skb, allocation);
1776 if (!nskb)
1777 return skb;
Eric Dumazet8460c002012-04-19 02:24:28 +00001778 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 skb = nskb;
1780 }
1781
1782 if (!pskb_expand_head(skb, 0, -delta, allocation))
1783 skb->truesize -= delta;
1784
1785 return skb;
1786}
1787
Eric W. Biederman3fbc2902012-05-24 17:21:27 -06001788static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
1789 struct sock *ssk)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001790{
1791 int ret;
1792 struct netlink_sock *nlk = nlk_sk(sk);
1793
1794 ret = -ECONNREFUSED;
1795 if (nlk->netlink_rcv != NULL) {
1796 ret = skb->len;
Patrick McHardycf0a0182013-04-17 06:47:00 +00001797 netlink_skb_set_owner_r(skb, sk);
Patrick McHardye32123e2013-04-17 06:46:57 +00001798 NETLINK_CB(skb).sk = ssk;
Daniel Borkmann73bfd372013-12-23 14:35:55 +01001799 netlink_deliver_tap_kernel(sk, ssk, skb);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001800 nlk->netlink_rcv(skb);
Eric Dumazetbfb253c2012-04-22 21:30:29 +00001801 consume_skb(skb);
1802 } else {
1803 kfree_skb(skb);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001804 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001805 sock_put(sk);
1806 return ret;
1807}
1808
1809int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
Eric W. Biederman15e47302012-09-07 20:12:54 +00001810 u32 portid, int nonblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811{
1812 struct sock *sk;
1813 int err;
1814 long timeo;
1815
1816 skb = netlink_trim(skb, gfp_any());
1817
1818 timeo = sock_sndtimeo(ssk, nonblock);
1819retry:
Eric W. Biederman15e47302012-09-07 20:12:54 +00001820 sk = netlink_getsockbyportid(ssk, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 if (IS_ERR(sk)) {
1822 kfree_skb(skb);
1823 return PTR_ERR(sk);
1824 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001825 if (netlink_is_kernel(sk))
Eric W. Biederman3fbc2902012-05-24 17:21:27 -06001826 return netlink_unicast_kernel(sk, skb, ssk);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001827
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001828 if (sk_filter(sk, skb)) {
Wang Chen84874602008-07-01 19:55:09 -07001829 err = skb->len;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001830 kfree_skb(skb);
1831 sock_put(sk);
1832 return err;
1833 }
1834
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001835 err = netlink_attachskb(sk, skb, &timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 if (err == 1)
1837 goto retry;
1838 if (err)
1839 return err;
1840
Denis V. Lunev7ee015e2007-10-10 21:14:03 -07001841 return netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001843EXPORT_SYMBOL(netlink_unicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844
Patrick McHardyf9c22882013-04-17 06:47:04 +00001845struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size,
1846 u32 dst_portid, gfp_t gfp_mask)
1847{
1848#ifdef CONFIG_NETLINK_MMAP
1849 struct sock *sk = NULL;
1850 struct sk_buff *skb;
1851 struct netlink_ring *ring;
1852 struct nl_mmap_hdr *hdr;
1853 unsigned int maxlen;
1854
1855 sk = netlink_getsockbyportid(ssk, dst_portid);
1856 if (IS_ERR(sk))
1857 goto out;
1858
1859 ring = &nlk_sk(sk)->rx_ring;
1860 /* fast-path without atomic ops for common case: non-mmaped receiver */
1861 if (ring->pg_vec == NULL)
1862 goto out_put;
1863
Thomas Grafaae9f0e2013-11-30 13:21:31 +01001864 if (ring->frame_size - NL_MMAP_HDRLEN < size)
1865 goto out_put;
1866
Patrick McHardyf9c22882013-04-17 06:47:04 +00001867 skb = alloc_skb_head(gfp_mask);
1868 if (skb == NULL)
1869 goto err1;
1870
1871 spin_lock_bh(&sk->sk_receive_queue.lock);
1872 /* check again under lock */
1873 if (ring->pg_vec == NULL)
1874 goto out_free;
1875
Thomas Grafaae9f0e2013-11-30 13:21:31 +01001876 /* check again under lock */
Patrick McHardyf9c22882013-04-17 06:47:04 +00001877 maxlen = ring->frame_size - NL_MMAP_HDRLEN;
1878 if (maxlen < size)
1879 goto out_free;
1880
1881 netlink_forward_ring(ring);
1882 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
1883 if (hdr == NULL)
1884 goto err2;
1885 netlink_ring_setup_skb(skb, sk, ring, hdr);
1886 netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED);
1887 atomic_inc(&ring->pending);
1888 netlink_increment_head(ring);
1889
1890 spin_unlock_bh(&sk->sk_receive_queue.lock);
1891 return skb;
1892
1893err2:
1894 kfree_skb(skb);
1895 spin_unlock_bh(&sk->sk_receive_queue.lock);
Patrick McHardycd1df522013-04-17 06:47:05 +00001896 netlink_overrun(sk);
Patrick McHardyf9c22882013-04-17 06:47:04 +00001897err1:
1898 sock_put(sk);
1899 return NULL;
1900
1901out_free:
1902 kfree_skb(skb);
1903 spin_unlock_bh(&sk->sk_receive_queue.lock);
1904out_put:
1905 sock_put(sk);
1906out:
1907#endif
1908 return alloc_skb(size, gfp_mask);
1909}
1910EXPORT_SYMBOL_GPL(netlink_alloc_skb);
1911
Patrick McHardy4277a082006-03-20 18:52:01 -08001912int netlink_has_listeners(struct sock *sk, unsigned int group)
1913{
1914 int res = 0;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001915 struct listeners *listeners;
Patrick McHardy4277a082006-03-20 18:52:01 -08001916
Denis V. Lunevaed81562007-10-10 21:14:32 -07001917 BUG_ON(!netlink_is_kernel(sk));
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001918
1919 rcu_read_lock();
1920 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
1921
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001922 if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001923 res = test_bit(group - 1, listeners->masks);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001924
1925 rcu_read_unlock();
1926
Patrick McHardy4277a082006-03-20 18:52:01 -08001927 return res;
1928}
1929EXPORT_SYMBOL_GPL(netlink_has_listeners);
1930
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001931static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932{
1933 struct netlink_sock *nlk = nlk_sk(sk);
1934
1935 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
Patrick McHardycd967e02013-04-17 06:46:56 +00001936 !test_bit(NETLINK_CONGESTED, &nlk->state)) {
Patrick McHardycf0a0182013-04-17 06:47:00 +00001937 netlink_skb_set_owner_r(skb, sk);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001938 __netlink_sendskb(sk, skb);
stephen hemminger2c6458002011-12-22 08:52:03 +00001939 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 }
1941 return -1;
1942}
1943
1944struct netlink_broadcast_data {
1945 struct sock *exclude_sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001946 struct net *net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001947 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 u32 group;
1949 int failure;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001950 int delivery_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 int congested;
1952 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -04001953 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 struct sk_buff *skb, *skb2;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001955 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
1956 void *tx_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957};
1958
Rami Rosen46c95212014-07-01 21:17:35 +03001959static void do_one_broadcast(struct sock *sk,
1960 struct netlink_broadcast_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961{
1962 struct netlink_sock *nlk = nlk_sk(sk);
1963 int val;
1964
1965 if (p->exclude_sk == sk)
Rami Rosen46c95212014-07-01 21:17:35 +03001966 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967
Eric W. Biederman15e47302012-09-07 20:12:54 +00001968 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001969 !test_bit(p->group - 1, nlk->groups))
Rami Rosen46c95212014-07-01 21:17:35 +03001970 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001972 if (!net_eq(sock_net(sk), p->net))
Rami Rosen46c95212014-07-01 21:17:35 +03001973 return;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001974
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975 if (p->failure) {
1976 netlink_overrun(sk);
Rami Rosen46c95212014-07-01 21:17:35 +03001977 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 }
1979
1980 sock_hold(sk);
1981 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001982 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 p->skb2 = skb_clone(p->skb, p->allocation);
1984 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001985 p->skb2 = skb_get(p->skb);
1986 /*
1987 * skb ownership may have been set when
1988 * delivered to a previous socket.
1989 */
1990 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991 }
1992 }
1993 if (p->skb2 == NULL) {
1994 netlink_overrun(sk);
1995 /* Clone failed. Notify ALL listeners. */
1996 p->failure = 1;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001997 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1998 p->delivery_failure = 1;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001999 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
2000 kfree_skb(p->skb2);
2001 p->skb2 = NULL;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07002002 } else if (sk_filter(sk, p->skb2)) {
2003 kfree_skb(p->skb2);
2004 p->skb2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
2006 netlink_overrun(sk);
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00002007 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
2008 p->delivery_failure = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 } else {
2010 p->congested |= val;
2011 p->delivered = 1;
2012 p->skb2 = NULL;
2013 }
2014 sock_put(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015}
2016
Eric W. Biederman15e47302012-09-07 20:12:54 +00002017int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002018 u32 group, gfp_t allocation,
2019 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
2020 void *filter_data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002022 struct net *net = sock_net(ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 struct netlink_broadcast_data info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 struct sock *sk;
2025
2026 skb = netlink_trim(skb, allocation);
2027
2028 info.exclude_sk = ssk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002029 info.net = net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002030 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 info.group = group;
2032 info.failure = 0;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08002033 info.delivery_failure = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 info.congested = 0;
2035 info.delivered = 0;
2036 info.allocation = allocation;
2037 info.skb = skb;
2038 info.skb2 = NULL;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002039 info.tx_filter = filter;
2040 info.tx_data = filter_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
2042 /* While we sleep in clone, do not allow to change socket list */
2043
2044 netlink_lock_table();
2045
Sasha Levinb67bfe02013-02-27 17:06:00 -08002046 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 do_one_broadcast(sk, &info);
2048
Neil Horman70d4bf62010-07-20 06:45:56 +00002049 consume_skb(skb);
Tommy S. Christensenaa1c6a6f2005-05-19 13:07:32 -07002050
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 netlink_unlock_table();
2052
Neil Horman70d4bf62010-07-20 06:45:56 +00002053 if (info.delivery_failure) {
2054 kfree_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08002055 return -ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00002056 }
2057 consume_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08002058
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 if (info.delivered) {
2060 if (info.congested && (allocation & __GFP_WAIT))
2061 yield();
2062 return 0;
2063 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 return -ESRCH;
2065}
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002066EXPORT_SYMBOL(netlink_broadcast_filtered);
2067
Eric W. Biederman15e47302012-09-07 20:12:54 +00002068int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002069 u32 group, gfp_t allocation)
2070{
Eric W. Biederman15e47302012-09-07 20:12:54 +00002071 return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002072 NULL, NULL);
2073}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002074EXPORT_SYMBOL(netlink_broadcast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
2076struct netlink_set_err_data {
2077 struct sock *exclude_sk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002078 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 u32 group;
2080 int code;
2081};
2082
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00002083static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084{
2085 struct netlink_sock *nlk = nlk_sk(sk);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002086 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
2088 if (sk == p->exclude_sk)
2089 goto out;
2090
Octavian Purdila09ad9bc2009-11-25 15:14:13 -08002091 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002092 goto out;
2093
Eric W. Biederman15e47302012-09-07 20:12:54 +00002094 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07002095 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 goto out;
2097
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002098 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
2099 ret = 1;
2100 goto out;
2101 }
2102
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 sk->sk_err = p->code;
2104 sk->sk_error_report(sk);
2105out:
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002106 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107}
2108
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002109/**
2110 * netlink_set_err - report error to broadcast listeners
2111 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
Eric W. Biederman15e47302012-09-07 20:12:54 +00002112 * @portid: the PORTID of a process that we want to skip (if any)
Johannes Berg840e93f22013-11-19 10:35:40 +01002113 * @group: the broadcast group that will notice the error
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002114 * @code: error code, must be negative (as usual in kernelspace)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002115 *
2116 * This function returns the number of broadcast listeners that have set the
2117 * NETLINK_RECV_NO_ENOBUFS socket option.
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002118 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002119int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120{
2121 struct netlink_set_err_data info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 struct sock *sk;
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002123 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124
2125 info.exclude_sk = ssk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002126 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 info.group = group;
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002128 /* sk->sk_err wants a positive error value */
2129 info.code = -code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130
2131 read_lock(&nl_table_lock);
2132
Sasha Levinb67bfe02013-02-27 17:06:00 -08002133 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002134 ret += do_one_set_err(sk, &info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135
2136 read_unlock(&nl_table_lock);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002137 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138}
Pablo Neira Ayusodd5b6ce2009-03-23 13:21:06 +01002139EXPORT_SYMBOL(netlink_set_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Johannes Berg84659eb2007-07-18 15:47:05 -07002141/* must be called with netlink table grabbed */
2142static void netlink_update_socket_mc(struct netlink_sock *nlk,
2143 unsigned int group,
2144 int is_new)
2145{
2146 int old, new = !!is_new, subscriptions;
2147
2148 old = test_bit(group - 1, nlk->groups);
2149 subscriptions = nlk->subscriptions - old + new;
2150 if (new)
2151 __set_bit(group - 1, nlk->groups);
2152 else
2153 __clear_bit(group - 1, nlk->groups);
2154 netlink_update_subscriptions(&nlk->sk, subscriptions);
2155 netlink_update_listeners(&nlk->sk);
2156}
2157
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002158static int netlink_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002159 char __user *optval, unsigned int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002160{
2161 struct sock *sk = sock->sk;
2162 struct netlink_sock *nlk = nlk_sk(sk);
Johannes Bergeb496532007-07-18 02:07:51 -07002163 unsigned int val = 0;
2164 int err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002165
2166 if (level != SOL_NETLINK)
2167 return -ENOPROTOOPT;
2168
Patrick McHardyccdfcc32013-04-17 06:47:01 +00002169 if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING &&
2170 optlen >= sizeof(int) &&
Johannes Bergeb496532007-07-18 02:07:51 -07002171 get_user(val, (unsigned int __user *)optval))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002172 return -EFAULT;
2173
2174 switch (optname) {
2175 case NETLINK_PKTINFO:
2176 if (val)
2177 nlk->flags |= NETLINK_RECV_PKTINFO;
2178 else
2179 nlk->flags &= ~NETLINK_RECV_PKTINFO;
2180 err = 0;
2181 break;
2182 case NETLINK_ADD_MEMBERSHIP:
2183 case NETLINK_DROP_MEMBERSHIP: {
Eric W. Biederman5187cd02014-04-23 14:25:48 -07002184 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002185 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002186 err = netlink_realloc_groups(sk);
2187 if (err)
2188 return err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002189 if (!val || val - 1 >= nlk->ngroups)
2190 return -EINVAL;
Richard Guy Briggs7774d5e2014-04-22 21:31:55 -04002191 if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) {
Johannes Berg023e2cf2014-12-23 21:00:06 +01002192 err = nlk->netlink_bind(sock_net(sk), val);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04002193 if (err)
2194 return err;
2195 }
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002196 netlink_table_grab();
Johannes Berg84659eb2007-07-18 15:47:05 -07002197 netlink_update_socket_mc(nlk, val,
2198 optname == NETLINK_ADD_MEMBERSHIP);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002199 netlink_table_ungrab();
Richard Guy Briggs7774d5e2014-04-22 21:31:55 -04002200 if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind)
Johannes Berg023e2cf2014-12-23 21:00:06 +01002201 nlk->netlink_unbind(sock_net(sk), val);
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00002202
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002203 err = 0;
2204 break;
2205 }
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00002206 case NETLINK_BROADCAST_ERROR:
2207 if (val)
2208 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
2209 else
2210 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
2211 err = 0;
2212 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002213 case NETLINK_NO_ENOBUFS:
2214 if (val) {
2215 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
Patrick McHardycd967e02013-04-17 06:46:56 +00002216 clear_bit(NETLINK_CONGESTED, &nlk->state);
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002217 wake_up_interruptible(&nlk->wait);
Eric Dumazet658cb352012-04-22 21:30:21 +00002218 } else {
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002219 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00002220 }
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002221 err = 0;
2222 break;
Patrick McHardyccdfcc32013-04-17 06:47:01 +00002223#ifdef CONFIG_NETLINK_MMAP
2224 case NETLINK_RX_RING:
2225 case NETLINK_TX_RING: {
2226 struct nl_mmap_req req;
2227
2228 /* Rings might consume more memory than queue limits, require
2229 * CAP_NET_ADMIN.
2230 */
2231 if (!capable(CAP_NET_ADMIN))
2232 return -EPERM;
2233 if (optlen < sizeof(req))
2234 return -EINVAL;
2235 if (copy_from_user(&req, optval, sizeof(req)))
2236 return -EFAULT;
Florian Westphalb265c302015-07-21 16:33:50 +02002237 err = netlink_set_ring(sk, &req,
Patrick McHardyccdfcc32013-04-17 06:47:01 +00002238 optname == NETLINK_TX_RING);
2239 break;
2240 }
2241#endif /* CONFIG_NETLINK_MMAP */
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002242 default:
2243 err = -ENOPROTOOPT;
2244 }
2245 return err;
2246}
2247
2248static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002249 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002250{
2251 struct sock *sk = sock->sk;
2252 struct netlink_sock *nlk = nlk_sk(sk);
2253 int len, val, err;
2254
2255 if (level != SOL_NETLINK)
2256 return -ENOPROTOOPT;
2257
2258 if (get_user(len, optlen))
2259 return -EFAULT;
2260 if (len < 0)
2261 return -EINVAL;
2262
2263 switch (optname) {
2264 case NETLINK_PKTINFO:
2265 if (len < sizeof(int))
2266 return -EINVAL;
2267 len = sizeof(int);
2268 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08002269 if (put_user(len, optlen) ||
2270 put_user(val, optval))
2271 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002272 err = 0;
2273 break;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00002274 case NETLINK_BROADCAST_ERROR:
2275 if (len < sizeof(int))
2276 return -EINVAL;
2277 len = sizeof(int);
2278 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
2279 if (put_user(len, optlen) ||
2280 put_user(val, optval))
2281 return -EFAULT;
2282 err = 0;
2283 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002284 case NETLINK_NO_ENOBUFS:
2285 if (len < sizeof(int))
2286 return -EINVAL;
2287 len = sizeof(int);
2288 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
2289 if (put_user(len, optlen) ||
2290 put_user(val, optval))
2291 return -EFAULT;
2292 err = 0;
2293 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002294 default:
2295 err = -ENOPROTOOPT;
2296 }
2297 return err;
2298}
2299
2300static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
2301{
2302 struct nl_pktinfo info;
2303
2304 info.group = NETLINK_CB(skb).dst_group;
2305 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
2306}
2307
Ying Xue1b784142015-03-02 15:37:48 +08002308static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310 struct sock *sk = sock->sk;
2311 struct netlink_sock *nlk = nlk_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01002312 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002313 u32 dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002314 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 struct sk_buff *skb;
2316 int err;
2317 struct scm_cookie scm;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002318 u32 netlink_skb_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319
2320 if (msg->msg_flags&MSG_OOB)
2321 return -EOPNOTSUPP;
2322
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002323 err = scm_send(sock, msg, &scm, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 if (err < 0)
2325 return err;
2326
2327 if (msg->msg_namelen) {
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002328 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 if (addr->nl_family != AF_NETLINK)
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002330 goto out;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002331 dst_portid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002332 dst_group = ffs(addr->nl_groups);
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002333 err = -EPERM;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002334 if ((dst_group || dst_portid) &&
Eric W. Biederman5187cd02014-04-23 14:25:48 -07002335 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002336 goto out;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002337 netlink_skb_flags |= NETLINK_SKB_DST;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002339 dst_portid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002340 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341 }
2342
Eric W. Biederman15e47302012-09-07 20:12:54 +00002343 if (!nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 err = netlink_autobind(sock);
2345 if (err)
2346 goto out;
2347 }
2348
Al Viroa8866ff2014-12-12 23:02:36 -05002349 /* It's a really convoluted way for userland to ask for mmaped
2350 * sendmsg(), but that's what we've got...
2351 */
Patrick McHardy5fd96122013-04-17 06:47:03 +00002352 if (netlink_tx_is_mmaped(sk) &&
Al Viroa8866ff2014-12-12 23:02:36 -05002353 msg->msg_iter.type == ITER_IOVEC &&
2354 msg->msg_iter.nr_segs == 1 &&
Al Viroc0371da2014-11-24 10:42:55 -05002355 msg->msg_iter.iov->iov_base == NULL) {
Patrick McHardy5fd96122013-04-17 06:47:03 +00002356 err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group,
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002357 &scm);
Patrick McHardy5fd96122013-04-17 06:47:03 +00002358 goto out;
2359 }
2360
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361 err = -EMSGSIZE;
2362 if (len > sk->sk_sndbuf - 32)
2363 goto out;
2364 err = -ENOBUFS;
Pablo Neira3a365152013-06-28 03:04:23 +02002365 skb = netlink_alloc_large_skb(len, dst_group);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002366 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367 goto out;
2368
Eric W. Biederman15e47302012-09-07 20:12:54 +00002369 NETLINK_CB(skb).portid = nlk->portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002370 NETLINK_CB(skb).dst_group = dst_group;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002371 NETLINK_CB(skb).creds = scm.creds;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002372 NETLINK_CB(skb).flags = netlink_skb_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 err = -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002375 if (memcpy_from_msg(skb_put(skb, len), msg, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 kfree_skb(skb);
2377 goto out;
2378 }
2379
2380 err = security_netlink_send(sk, skb);
2381 if (err) {
2382 kfree_skb(skb);
2383 goto out;
2384 }
2385
Patrick McHardyd629b832005-08-14 19:27:50 -07002386 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002388 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 }
Eric W. Biederman15e47302012-09-07 20:12:54 +00002390 err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391
2392out:
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002393 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 return err;
2395}
2396
Ying Xue1b784142015-03-02 15:37:48 +08002397static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398 int flags)
2399{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 struct scm_cookie scm;
2401 struct sock *sk = sock->sk;
2402 struct netlink_sock *nlk = nlk_sk(sk);
2403 int noblock = flags&MSG_DONTWAIT;
2404 size_t copied;
Johannes Berg68d6ac62010-08-15 21:20:44 +00002405 struct sk_buff *skb, *data_skb;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002406 int err, ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407
2408 if (flags&MSG_OOB)
2409 return -EOPNOTSUPP;
2410
2411 copied = 0;
2412
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002413 skb = skb_recv_datagram(sk, flags, noblock, &err);
2414 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415 goto out;
2416
Johannes Berg68d6ac62010-08-15 21:20:44 +00002417 data_skb = skb;
2418
Johannes Berg1dacc762009-07-01 11:26:02 +00002419#ifdef CONFIG_COMPAT_NETLINK_MESSAGES
2420 if (unlikely(skb_shinfo(skb)->frag_list)) {
Johannes Berg1dacc762009-07-01 11:26:02 +00002421 /*
Johannes Berg68d6ac62010-08-15 21:20:44 +00002422 * If this skb has a frag_list, then here that means that we
2423 * will have to use the frag_list skb's data for compat tasks
2424 * and the regular skb's data for normal (non-compat) tasks.
Johannes Berg1dacc762009-07-01 11:26:02 +00002425 *
Johannes Berg68d6ac62010-08-15 21:20:44 +00002426 * If we need to send the compat skb, assign it to the
2427 * 'data_skb' variable so that it will be used below for data
2428 * copying. We keep 'skb' for everything else, including
2429 * freeing both later.
Johannes Berg1dacc762009-07-01 11:26:02 +00002430 */
Johannes Berg68d6ac62010-08-15 21:20:44 +00002431 if (flags & MSG_CMSG_COMPAT)
2432 data_skb = skb_shinfo(skb)->frag_list;
Johannes Berg1dacc762009-07-01 11:26:02 +00002433 }
2434#endif
2435
Eric Dumazet9063e212014-03-07 12:02:33 -08002436 /* Record the max length of recvmsg() calls for future allocations */
2437 nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len);
2438 nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len,
2439 16384);
2440
Johannes Berg68d6ac62010-08-15 21:20:44 +00002441 copied = data_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 if (len < copied) {
2443 msg->msg_flags |= MSG_TRUNC;
2444 copied = len;
2445 }
2446
Johannes Berg68d6ac62010-08-15 21:20:44 +00002447 skb_reset_transport_header(data_skb);
David S. Miller51f3d022014-11-05 16:46:40 -05002448 err = skb_copy_datagram_msg(data_skb, 0, msg, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449
2450 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +01002451 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 addr->nl_family = AF_NETLINK;
2453 addr->nl_pad = 0;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002454 addr->nl_pid = NETLINK_CB(skb).portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002455 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456 msg->msg_namelen = sizeof(*addr);
2457 }
2458
Patrick McHardycc9a06c2006-03-12 20:34:27 -08002459 if (nlk->flags & NETLINK_RECV_PKTINFO)
2460 netlink_cmsg_recv_pktinfo(msg, skb);
2461
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002462 memset(&scm, 0, sizeof(scm));
2463 scm.creds = *NETLINK_CREDS(skb);
Patrick McHardy188ccb52007-05-03 03:27:01 -07002464 if (flags & MSG_TRUNC)
Johannes Berg68d6ac62010-08-15 21:20:44 +00002465 copied = data_skb->len;
David S. Millerdaa37662010-08-15 23:21:50 -07002466
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467 skb_free_datagram(sk, skb);
2468
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002469 if (nlk->cb_running &&
2470 atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
Andrey Vaginb44d2112011-02-21 02:40:47 +00002471 ret = netlink_dump(sk);
2472 if (ret) {
Ben Pfaffac30ef82014-07-09 10:31:22 -07002473 sk->sk_err = -ret;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002474 sk->sk_error_report(sk);
2475 }
2476 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002478 scm_recv(sock, msg, &scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479out:
2480 netlink_rcv_wake(sk);
2481 return err ? : copied;
2482}
2483
David S. Miller676d2362014-04-11 16:15:36 -04002484static void netlink_data_ready(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485{
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002486 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487}
2488
2489/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002490 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491 * complete set of kernel non-blocking support for message
2492 * queueing.
2493 */
2494
2495struct sock *
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00002496__netlink_kernel_create(struct net *net, int unit, struct module *module,
2497 struct netlink_kernel_cfg *cfg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002498{
2499 struct socket *sock;
2500 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07002501 struct netlink_sock *nlk;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002502 struct listeners *listeners = NULL;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002503 struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
2504 unsigned int groups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002505
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002506 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002508 if (unit < 0 || unit >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509 return NULL;
2510
2511 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
2512 return NULL;
2513
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08002514 /*
2515 * We have to just have a reference on the net from sk, but don't
2516 * get_net it. Besides, we cannot get and then put the net here.
2517 * So we create one inside init_net and the move it to net.
2518 */
2519
2520 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
2521 goto out_sock_release_nosk;
2522
2523 sk = sock->sk;
Denis V. Lunevedf02082008-02-29 11:18:32 -08002524 sk_change_net(sk, net);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002525
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002526 if (!cfg || cfg->groups < 32)
Patrick McHardy4277a082006-03-20 18:52:01 -08002527 groups = 32;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002528 else
2529 groups = cfg->groups;
Patrick McHardy4277a082006-03-20 18:52:01 -08002530
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002531 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
Patrick McHardy4277a082006-03-20 18:52:01 -08002532 if (!listeners)
2533 goto out_sock_release;
2534
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535 sk->sk_data_ready = netlink_data_ready;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002536 if (cfg && cfg->input)
2537 nlk_sk(sk)->netlink_rcv = cfg->input;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
Herbert Xu8ea65f42015-01-26 14:02:56 +11002539 if (netlink_insert(sk, 0))
Patrick McHardy77247bb2005-08-14 19:27:13 -07002540 goto out_sock_release;
2541
2542 nlk = nlk_sk(sk);
2543 nlk->flags |= NETLINK_KERNEL_SOCKET;
2544
2545 netlink_table_grab();
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002546 if (!nl_table[unit].registered) {
2547 nl_table[unit].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002548 rcu_assign_pointer(nl_table[unit].listeners, listeners);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002549 nl_table[unit].cb_mutex = cb_mutex;
2550 nl_table[unit].module = module;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002551 if (cfg) {
2552 nl_table[unit].bind = cfg->bind;
Hiroaki SHIMODA6251edd2014-11-13 04:24:10 +09002553 nl_table[unit].unbind = cfg->unbind;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002554 nl_table[unit].flags = cfg->flags;
Gao fengda12c902013-06-06 14:49:11 +08002555 if (cfg->compare)
2556 nl_table[unit].compare = cfg->compare;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002557 }
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002558 nl_table[unit].registered = 1;
Jesper Juhlf937f1f462007-10-15 01:39:12 -07002559 } else {
2560 kfree(listeners);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08002561 nl_table[unit].registered++;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002562 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07002563 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002564 return sk;
2565
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002566out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08002567 kfree(listeners);
Denis V. Lunev9dfbec12008-02-29 11:17:56 -08002568 netlink_kernel_release(sk);
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08002569 return NULL;
2570
2571out_sock_release_nosk:
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002572 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07002573 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574}
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00002575EXPORT_SYMBOL(__netlink_kernel_create);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08002576
2577void
2578netlink_kernel_release(struct sock *sk)
2579{
Denis V. Lunevedf02082008-02-29 11:18:32 -08002580 sk_release_kernel(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08002581}
2582EXPORT_SYMBOL(netlink_kernel_release);
2583
Johannes Bergd136f1b2009-09-12 03:03:15 +00002584int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002585{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002586 struct listeners *new, *old;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002587 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002588
2589 if (groups < 32)
2590 groups = 32;
2591
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002592 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002593 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
2594 if (!new)
Johannes Bergd136f1b2009-09-12 03:03:15 +00002595 return -ENOMEM;
Eric Dumazet6d772ac2012-10-18 03:21:55 +00002596 old = nl_deref_protected(tbl->listeners);
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002597 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
2598 rcu_assign_pointer(tbl->listeners, new);
2599
Lai Jiangshan37b6b932011-03-15 18:01:42 +08002600 kfree_rcu(old, rcu);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002601 }
2602 tbl->groups = groups;
2603
Johannes Bergd136f1b2009-09-12 03:03:15 +00002604 return 0;
2605}
2606
2607/**
2608 * netlink_change_ngroups - change number of multicast groups
2609 *
2610 * This changes the number of multicast groups that are available
2611 * on a certain netlink family. Note that it is not possible to
2612 * change the number of groups to below 32. Also note that it does
2613 * not implicitly call netlink_clear_multicast_users() when the
2614 * number of groups is reduced.
2615 *
2616 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
2617 * @groups: The new number of groups.
2618 */
2619int netlink_change_ngroups(struct sock *sk, unsigned int groups)
2620{
2621 int err;
2622
2623 netlink_table_grab();
2624 err = __netlink_change_ngroups(sk, groups);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002625 netlink_table_ungrab();
Johannes Bergd136f1b2009-09-12 03:03:15 +00002626
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002627 return err;
2628}
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002629
Johannes Bergb8273572009-09-24 15:44:05 -07002630void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
2631{
2632 struct sock *sk;
Johannes Bergb8273572009-09-24 15:44:05 -07002633 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
2634
Sasha Levinb67bfe02013-02-27 17:06:00 -08002635 sk_for_each_bound(sk, &tbl->mc_list)
Johannes Bergb8273572009-09-24 15:44:05 -07002636 netlink_update_socket_mc(nlk_sk(sk), group, 0);
2637}
2638
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002639struct nlmsghdr *
Eric W. Biederman15e47302012-09-07 20:12:54 +00002640__nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002641{
2642 struct nlmsghdr *nlh;
Hong zhi guo573ce262013-03-27 06:47:04 +00002643 int size = nlmsg_msg_size(len);
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002644
Wang Yufen23b45672014-02-17 16:53:32 +08002645 nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size));
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002646 nlh->nlmsg_type = type;
2647 nlh->nlmsg_len = size;
2648 nlh->nlmsg_flags = flags;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002649 nlh->nlmsg_pid = portid;
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002650 nlh->nlmsg_seq = seq;
2651 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
Hong zhi guo573ce262013-03-27 06:47:04 +00002652 memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002653 return nlh;
2654}
2655EXPORT_SYMBOL(__nlmsg_put);
2656
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657/*
2658 * It looks a bit ugly.
2659 * It would be better to create kernel thread.
2660 */
2661
2662static int netlink_dump(struct sock *sk)
2663{
2664 struct netlink_sock *nlk = nlk_sk(sk);
2665 struct netlink_callback *cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00002666 struct sk_buff *skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002667 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002668 int len, err = -ENOBUFS;
Greg Rosec7ac8672011-06-10 01:27:09 +00002669 int alloc_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002671 mutex_lock(nlk->cb_mutex);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002672 if (!nlk->cb_running) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002673 err = -EINVAL;
2674 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002675 }
2676
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002677 cb = &nlk->cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00002678 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
2679
Patrick McHardyf9c22882013-04-17 06:47:04 +00002680 if (!netlink_rx_is_mmaped(sk) &&
2681 atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
2682 goto errout_skb;
Eric Dumazet9063e212014-03-07 12:02:33 -08002683
2684 /* NLMSG_GOODSIZE is small to avoid high order allocations being
2685 * required, but it makes sense to _attempt_ a 16K bytes allocation
2686 * to reduce number of system calls on dump operations, if user
2687 * ever provided a big enough buffer.
2688 */
2689 if (alloc_size < nlk->max_recvmsg_len) {
2690 skb = netlink_alloc_skb(sk,
2691 nlk->max_recvmsg_len,
2692 nlk->portid,
2693 GFP_KERNEL |
2694 __GFP_NOWARN |
2695 __GFP_NORETRY);
2696 /* available room should be exact amount to avoid MSG_TRUNC */
2697 if (skb)
2698 skb_reserve(skb, skb_tailroom(skb) -
2699 nlk->max_recvmsg_len);
2700 }
2701 if (!skb)
2702 skb = netlink_alloc_skb(sk, alloc_size, nlk->portid,
2703 GFP_KERNEL);
Greg Rosec7ac8672011-06-10 01:27:09 +00002704 if (!skb)
Dan Carpenterc63d6ea2011-06-15 03:11:42 +00002705 goto errout_skb;
Patrick McHardyf9c22882013-04-17 06:47:04 +00002706 netlink_skb_set_owner_r(skb, sk);
Greg Rosec7ac8672011-06-10 01:27:09 +00002707
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708 len = cb->dump(skb, cb);
2709
2710 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002711 mutex_unlock(nlk->cb_mutex);
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07002712
2713 if (sk_filter(sk, skb))
2714 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00002715 else
2716 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002717 return 0;
2718 }
2719
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002720 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
2721 if (!nlh)
2722 goto errout_skb;
2723
Johannes Berg670dc282011-06-20 13:40:46 +02002724 nl_dump_check_consistent(cb, nlh);
2725
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002726 memcpy(nlmsg_data(nlh), &len, sizeof(len));
2727
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07002728 if (sk_filter(sk, skb))
2729 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00002730 else
2731 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732
Thomas Grafa8f74b22005-11-10 02:25:52 +01002733 if (cb->done)
2734 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002735
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002736 nlk->cb_running = false;
2737 mutex_unlock(nlk->cb_mutex);
Gao feng6dc878a2012-10-04 20:15:48 +00002738 module_put(cb->module);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002739 consume_skb(cb->skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002740 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07002741
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002742errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002743 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002744 kfree_skb(skb);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002745 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002746}
2747
Gao feng6dc878a2012-10-04 20:15:48 +00002748int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
2749 const struct nlmsghdr *nlh,
2750 struct netlink_dump_control *control)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002751{
2752 struct netlink_callback *cb;
2753 struct sock *sk;
2754 struct netlink_sock *nlk;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002755 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756
Patrick McHardyf9c22882013-04-17 06:47:04 +00002757 /* Memory mapped dump requests need to be copied to avoid looping
2758 * on the pending state in netlink_mmap_sendmsg() while the CB hold
2759 * a reference to the skb.
2760 */
2761 if (netlink_skb_is_mmaped(skb)) {
2762 skb = skb_copy(skb, GFP_KERNEL);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002763 if (skb == NULL)
Patrick McHardyf9c22882013-04-17 06:47:04 +00002764 return -ENOBUFS;
Patrick McHardyf9c22882013-04-17 06:47:04 +00002765 } else
2766 atomic_inc(&skb->users);
2767
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002768 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
2769 if (sk == NULL) {
2770 ret = -ECONNREFUSED;
2771 goto error_free;
2772 }
2773
2774 nlk = nlk_sk(sk);
2775 mutex_lock(nlk->cb_mutex);
2776 /* A dump is in progress... */
2777 if (nlk->cb_running) {
2778 ret = -EBUSY;
2779 goto error_unlock;
2780 }
2781 /* add reference of module which cb->dump belongs to */
2782 if (!try_module_get(control->module)) {
2783 ret = -EPROTONOSUPPORT;
2784 goto error_unlock;
2785 }
2786
2787 cb = &nlk->cb;
2788 memset(cb, 0, sizeof(*cb));
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00002789 cb->dump = control->dump;
2790 cb->done = control->done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002791 cb->nlh = nlh;
Pablo Neira Ayuso7175c882012-02-24 14:30:16 +00002792 cb->data = control->data;
Gao feng6dc878a2012-10-04 20:15:48 +00002793 cb->module = control->module;
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00002794 cb->min_dump_alloc = control->min_dump_alloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795 cb->skb = skb;
2796
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002797 nlk->cb_running = true;
Gao feng6dc878a2012-10-04 20:15:48 +00002798
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002799 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002800
Andrey Vaginb44d2112011-02-21 02:40:47 +00002801 ret = netlink_dump(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802 sock_put(sk);
Denis V. Lunev5c582982007-10-23 20:29:25 -07002803
Andrey Vaginb44d2112011-02-21 02:40:47 +00002804 if (ret)
2805 return ret;
2806
Denis V. Lunev5c582982007-10-23 20:29:25 -07002807 /* We successfully started a dump, by returning -EINTR we
2808 * signal not to send ACK even if it was requested.
2809 */
2810 return -EINTR;
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002811
2812error_unlock:
2813 sock_put(sk);
2814 mutex_unlock(nlk->cb_mutex);
2815error_free:
2816 kfree_skb(skb);
2817 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818}
Gao feng6dc878a2012-10-04 20:15:48 +00002819EXPORT_SYMBOL(__netlink_dump_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002820
2821void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
2822{
2823 struct sk_buff *skb;
2824 struct nlmsghdr *rep;
2825 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08002826 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002827
Thomas Graf339bf982006-11-10 14:10:15 -08002828 /* error messages get the original request appened */
2829 if (err)
2830 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831
Patrick McHardyf9c22882013-04-17 06:47:04 +00002832 skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload),
2833 NETLINK_CB(in_skb).portid, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002834 if (!skb) {
2835 struct sock *sk;
2836
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002837 sk = netlink_lookup(sock_net(in_skb->sk),
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002838 in_skb->sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00002839 NETLINK_CB(in_skb).portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840 if (sk) {
2841 sk->sk_err = ENOBUFS;
2842 sk->sk_error_report(sk);
2843 sock_put(sk);
2844 }
2845 return;
2846 }
2847
Eric W. Biederman15e47302012-09-07 20:12:54 +00002848 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
John Fastabend5dba93a2009-09-25 13:11:44 +00002849 NLMSG_ERROR, payload, 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002850 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002851 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002852 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Eric W. Biederman15e47302012-09-07 20:12:54 +00002853 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002854}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002855EXPORT_SYMBOL(netlink_ack);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002857int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07002858 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01002859{
Thomas Graf82ace472005-11-10 02:25:53 +01002860 struct nlmsghdr *nlh;
2861 int err;
2862
2863 while (skb->len >= nlmsg_total_size(0)) {
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002864 int msglen;
2865
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07002866 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07002867 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01002868
Martin Murrayad8e4b72006-01-10 13:02:29 -08002869 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01002870 return 0;
2871
Thomas Grafd35b6852007-03-22 23:28:46 -07002872 /* Only requests are handled by the kernel */
2873 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
Denis V. Lunev5c582982007-10-23 20:29:25 -07002874 goto ack;
Thomas Grafd35b6852007-03-22 23:28:46 -07002875
Thomas Graf45e7ae72007-03-22 23:29:10 -07002876 /* Skip control messages */
2877 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
Denis V. Lunev5c582982007-10-23 20:29:25 -07002878 goto ack;
Thomas Graf45e7ae72007-03-22 23:29:10 -07002879
Thomas Graf1d00a4e2007-03-22 23:30:12 -07002880 err = cb(skb, nlh);
Denis V. Lunev5c582982007-10-23 20:29:25 -07002881 if (err == -EINTR)
2882 goto skip;
2883
2884ack:
Thomas Grafd35b6852007-03-22 23:28:46 -07002885 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01002886 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01002887
Denis V. Lunev5c582982007-10-23 20:29:25 -07002888skip:
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002889 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002890 if (msglen > skb->len)
2891 msglen = skb->len;
2892 skb_pull(skb, msglen);
Thomas Graf82ace472005-11-10 02:25:53 +01002893 }
2894
2895 return 0;
2896}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002897EXPORT_SYMBOL(netlink_rcv_skb);
Thomas Graf82ace472005-11-10 02:25:53 +01002898
2899/**
Thomas Grafd387f6a2006-08-15 00:31:06 -07002900 * nlmsg_notify - send a notification netlink message
2901 * @sk: netlink socket to use
2902 * @skb: notification message
Eric W. Biederman15e47302012-09-07 20:12:54 +00002903 * @portid: destination netlink portid for reports or 0
Thomas Grafd387f6a2006-08-15 00:31:06 -07002904 * @group: destination multicast group or 0
2905 * @report: 1 to report back, 0 to disable
2906 * @flags: allocation flags
2907 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002908int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
Thomas Grafd387f6a2006-08-15 00:31:06 -07002909 unsigned int group, int report, gfp_t flags)
2910{
2911 int err = 0;
2912
2913 if (group) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002914 int exclude_portid = 0;
Thomas Grafd387f6a2006-08-15 00:31:06 -07002915
2916 if (report) {
2917 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002918 exclude_portid = portid;
Thomas Grafd387f6a2006-08-15 00:31:06 -07002919 }
2920
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002921 /* errors reported via destination sk->sk_err, but propagate
2922 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002923 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
Thomas Grafd387f6a2006-08-15 00:31:06 -07002924 }
2925
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002926 if (report) {
2927 int err2;
2928
Eric W. Biederman15e47302012-09-07 20:12:54 +00002929 err2 = nlmsg_unicast(sk, skb, portid);
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002930 if (!err || err == -ESRCH)
2931 err = err2;
2932 }
Thomas Grafd387f6a2006-08-15 00:31:06 -07002933
2934 return err;
2935}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002936EXPORT_SYMBOL(nlmsg_notify);
Thomas Grafd387f6a2006-08-15 00:31:06 -07002937
Linus Torvalds1da177e2005-04-16 15:20:36 -07002938#ifdef CONFIG_PROC_FS
2939struct nl_seq_iter {
Denis V. Luneve372c412007-11-19 22:31:54 -08002940 struct seq_net_private p;
Herbert Xu56d28b12015-02-04 07:33:24 +11002941 struct rhashtable_iter hti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002942 int link;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002943};
2944
Herbert Xu56d28b12015-02-04 07:33:24 +11002945static int netlink_walk_start(struct nl_seq_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002946{
Herbert Xu56d28b12015-02-04 07:33:24 +11002947 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002948
Herbert Xu56d28b12015-02-04 07:33:24 +11002949 err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti);
2950 if (err) {
2951 iter->link = MAX_LINKS;
2952 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002953 }
Herbert Xu56d28b12015-02-04 07:33:24 +11002954
2955 err = rhashtable_walk_start(&iter->hti);
2956 return err == -EAGAIN ? 0 : err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002957}
2958
Herbert Xu56d28b12015-02-04 07:33:24 +11002959static void netlink_walk_stop(struct nl_seq_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002960{
Herbert Xu56d28b12015-02-04 07:33:24 +11002961 rhashtable_walk_stop(&iter->hti);
2962 rhashtable_walk_exit(&iter->hti);
2963}
2964
2965static void *__netlink_seq_next(struct seq_file *seq)
2966{
2967 struct nl_seq_iter *iter = seq->private;
2968 struct netlink_sock *nlk;
2969
2970 do {
2971 for (;;) {
2972 int err;
2973
2974 nlk = rhashtable_walk_next(&iter->hti);
2975
2976 if (IS_ERR(nlk)) {
2977 if (PTR_ERR(nlk) == -EAGAIN)
2978 continue;
2979
2980 return nlk;
2981 }
2982
2983 if (nlk)
2984 break;
2985
2986 netlink_walk_stop(iter);
2987 if (++iter->link >= MAX_LINKS)
2988 return NULL;
2989
2990 err = netlink_walk_start(iter);
2991 if (err)
2992 return ERR_PTR(err);
2993 }
2994 } while (sock_net(&nlk->sk) != seq_file_net(seq));
2995
2996 return nlk;
2997}
2998
2999static void *netlink_seq_start(struct seq_file *seq, loff_t *posp)
3000{
3001 struct nl_seq_iter *iter = seq->private;
3002 void *obj = SEQ_START_TOKEN;
3003 loff_t pos;
3004 int err;
3005
3006 iter->link = 0;
3007
3008 err = netlink_walk_start(iter);
3009 if (err)
3010 return ERR_PTR(err);
3011
3012 for (pos = *posp; pos && obj && !IS_ERR(obj); pos--)
3013 obj = __netlink_seq_next(seq);
3014
3015 return obj;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003016}
3017
3018static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3019{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003020 ++*pos;
Herbert Xu56d28b12015-02-04 07:33:24 +11003021 return __netlink_seq_next(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003022}
3023
3024static void netlink_seq_stop(struct seq_file *seq, void *v)
3025{
Herbert Xu56d28b12015-02-04 07:33:24 +11003026 struct nl_seq_iter *iter = seq->private;
3027
3028 if (iter->link >= MAX_LINKS)
3029 return;
3030
3031 netlink_walk_stop(iter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003032}
3033
3034
3035static int netlink_seq_show(struct seq_file *seq, void *v)
3036{
Eric Dumazet658cb352012-04-22 21:30:21 +00003037 if (v == SEQ_START_TOKEN) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003038 seq_puts(seq,
3039 "sk Eth Pid Groups "
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00003040 "Rmem Wmem Dump Locks Drops Inode\n");
Eric Dumazet658cb352012-04-22 21:30:21 +00003041 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003042 struct sock *s = v;
3043 struct netlink_sock *nlk = nlk_sk(s);
3044
Pravin B Shelar16b304f2013-08-15 15:31:06 -07003045 seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07003046 s,
3047 s->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00003048 nlk->portid,
Patrick McHardy513c2502005-09-06 15:43:59 -07003049 nlk->groups ? (u32)nlk->groups[0] : 0,
Eric Dumazet31e6d362009-06-17 19:05:41 -07003050 sk_rmem_alloc_get(s),
3051 sk_wmem_alloc_get(s),
Pravin B Shelar16b304f2013-08-15 15:31:06 -07003052 nlk->cb_running,
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07003053 atomic_read(&s->sk_refcnt),
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00003054 atomic_read(&s->sk_drops),
3055 sock_i_ino(s)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003056 );
3057
3058 }
3059 return 0;
3060}
3061
Philippe De Muyter56b3d972007-07-10 23:07:31 -07003062static const struct seq_operations netlink_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003063 .start = netlink_seq_start,
3064 .next = netlink_seq_next,
3065 .stop = netlink_seq_stop,
3066 .show = netlink_seq_show,
3067};
3068
3069
3070static int netlink_seq_open(struct inode *inode, struct file *file)
3071{
Denis V. Luneve372c412007-11-19 22:31:54 -08003072 return seq_open_net(inode, file, &netlink_seq_ops,
3073 sizeof(struct nl_seq_iter));
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003074}
3075
Arjan van de Venda7071d2007-02-12 00:55:36 -08003076static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077 .owner = THIS_MODULE,
3078 .open = netlink_seq_open,
3079 .read = seq_read,
3080 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08003081 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003082};
3083
3084#endif
3085
3086int netlink_register_notifier(struct notifier_block *nb)
3087{
Alan Sterne041c682006-03-27 01:16:30 -08003088 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08003090EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003091
3092int netlink_unregister_notifier(struct notifier_block *nb)
3093{
Alan Sterne041c682006-03-27 01:16:30 -08003094 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003095}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08003096EXPORT_SYMBOL(netlink_unregister_notifier);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09003097
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08003098static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003099 .family = PF_NETLINK,
3100 .owner = THIS_MODULE,
3101 .release = netlink_release,
3102 .bind = netlink_bind,
3103 .connect = netlink_connect,
3104 .socketpair = sock_no_socketpair,
3105 .accept = sock_no_accept,
3106 .getname = netlink_getname,
Patrick McHardy9652e932013-04-17 06:47:02 +00003107 .poll = netlink_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003108 .ioctl = sock_no_ioctl,
3109 .listen = sock_no_listen,
3110 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07003111 .setsockopt = netlink_setsockopt,
3112 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003113 .sendmsg = netlink_sendmsg,
3114 .recvmsg = netlink_recvmsg,
Patrick McHardyccdfcc32013-04-17 06:47:01 +00003115 .mmap = netlink_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003116 .sendpage = sock_no_sendpage,
3117};
3118
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00003119static const struct net_proto_family netlink_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003120 .family = PF_NETLINK,
3121 .create = netlink_create,
3122 .owner = THIS_MODULE, /* for consistency 8) */
3123};
3124
Pavel Emelyanov46650792007-10-08 20:38:39 -07003125static int __net_init netlink_net_init(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003126{
3127#ifdef CONFIG_PROC_FS
Gao fengd4beaa62013-02-18 01:34:54 +00003128 if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003129 return -ENOMEM;
3130#endif
3131 return 0;
3132}
3133
Pavel Emelyanov46650792007-10-08 20:38:39 -07003134static void __net_exit netlink_net_exit(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003135{
3136#ifdef CONFIG_PROC_FS
Gao fengece31ff2013-02-18 01:34:56 +00003137 remove_proc_entry("netlink", net->proc_net);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003138#endif
3139}
3140
David S. Millerb963ea82010-08-30 19:08:01 -07003141static void __init netlink_add_usersock_entry(void)
3142{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003143 struct listeners *listeners;
David S. Millerb963ea82010-08-30 19:08:01 -07003144 int groups = 32;
3145
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003146 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
David S. Millerb963ea82010-08-30 19:08:01 -07003147 if (!listeners)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003148 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
David S. Millerb963ea82010-08-30 19:08:01 -07003149
3150 netlink_table_grab();
3151
3152 nl_table[NETLINK_USERSOCK].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003153 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
David S. Millerb963ea82010-08-30 19:08:01 -07003154 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
3155 nl_table[NETLINK_USERSOCK].registered = 1;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00003156 nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
David S. Millerb963ea82010-08-30 19:08:01 -07003157
3158 netlink_table_ungrab();
3159}
3160
Denis V. Lunev022cbae2007-11-13 03:23:50 -08003161static struct pernet_operations __net_initdata netlink_net_ops = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003162 .init = netlink_net_init,
3163 .exit = netlink_net_exit,
3164};
3165
Patrick McHardy49f7b332015-03-25 13:07:45 +00003166static inline u32 netlink_hash(const void *data, u32 len, u32 seed)
Herbert Xuc428ecd2015-03-20 21:57:01 +11003167{
3168 const struct netlink_sock *nlk = data;
3169 struct netlink_compare_arg arg;
3170
Herbert Xu4e277622015-09-18 19:16:50 +08003171 netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->rhash_portid);
Herbert Xu11b58ba2015-03-24 00:50:22 +11003172 return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed);
Herbert Xuc428ecd2015-03-20 21:57:01 +11003173}
3174
3175static const struct rhashtable_params netlink_rhashtable_params = {
3176 .head_offset = offsetof(struct netlink_sock, node),
3177 .key_len = netlink_compare_arg_len,
Herbert Xuc428ecd2015-03-20 21:57:01 +11003178 .obj_hashfn = netlink_hash,
3179 .obj_cmpfn = netlink_compare,
Thomas Grafb5e2c152015-03-24 20:42:19 +00003180 .automatic_shrinking = true,
Herbert Xuc428ecd2015-03-20 21:57:01 +11003181};
3182
Linus Torvalds1da177e2005-04-16 15:20:36 -07003183static int __init netlink_proto_init(void)
3184{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003185 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003186 int err = proto_register(&netlink_proto, 0);
3187
3188 if (err != 0)
3189 goto out;
3190
YOSHIFUJI Hideaki / 吉藤英明fab25742013-01-09 07:19:48 +00003191 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003192
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07003193 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003194 if (!nl_table)
3195 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003196
Linus Torvalds1da177e2005-04-16 15:20:36 -07003197 for (i = 0; i < MAX_LINKS; i++) {
Herbert Xuc428ecd2015-03-20 21:57:01 +11003198 if (rhashtable_init(&nl_table[i].hash,
3199 &netlink_rhashtable_params) < 0) {
Thomas Grafe3416942014-08-02 11:47:45 +02003200 while (--i > 0)
3201 rhashtable_destroy(&nl_table[i].hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003202 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003203 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003204 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003205 }
3206
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +02003207 INIT_LIST_HEAD(&netlink_tap_all);
3208
David S. Millerb963ea82010-08-30 19:08:01 -07003209 netlink_add_usersock_entry();
3210
Linus Torvalds1da177e2005-04-16 15:20:36 -07003211 sock_register(&netlink_family_ops);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003212 register_pernet_subsys(&netlink_net_ops);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09003213 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214 rtnetlink_init();
3215out:
3216 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003217panic:
3218 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003219}
3220
Linus Torvalds1da177e2005-04-16 15:20:36 -07003221core_initcall(netlink_proto_init);