blob: ec4adbdcb9b4a9926b4d992906f51787c57dc6a2 [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
Andrey Vagin0f29c762013-03-21 20:33:47 +040092struct netlink_table *nl_table;
93EXPORT_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 Borkmannbcbde0d2013-06-21 19:38:07 +0200126int netlink_add_tap(struct netlink_tap *nt)
127{
128 if (unlikely(nt->dev->type != ARPHRD_NETLINK))
129 return -EINVAL;
130
131 spin_lock(&netlink_tap_lock);
132 list_add_rcu(&nt->list, &netlink_tap_all);
133 spin_unlock(&netlink_tap_lock);
134
Markus Elfringfcd4d352014-11-18 21:03:13 +0100135 __module_get(nt->module);
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200136
137 return 0;
138}
139EXPORT_SYMBOL_GPL(netlink_add_tap);
140
stephen hemminger2173f8d2013-12-30 10:49:22 -0800141static int __netlink_remove_tap(struct netlink_tap *nt)
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200142{
143 bool found = false;
144 struct netlink_tap *tmp;
145
146 spin_lock(&netlink_tap_lock);
147
148 list_for_each_entry(tmp, &netlink_tap_all, list) {
149 if (nt == tmp) {
150 list_del_rcu(&nt->list);
151 found = true;
152 goto out;
153 }
154 }
155
156 pr_warn("__netlink_remove_tap: %p not found\n", nt);
157out:
158 spin_unlock(&netlink_tap_lock);
159
160 if (found && nt->module)
161 module_put(nt->module);
162
163 return found ? 0 : -ENODEV;
164}
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200165
166int netlink_remove_tap(struct netlink_tap *nt)
167{
168 int ret;
169
170 ret = __netlink_remove_tap(nt);
171 synchronize_net();
172
173 return ret;
174}
175EXPORT_SYMBOL_GPL(netlink_remove_tap);
176
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200177static bool netlink_filter_tap(const struct sk_buff *skb)
178{
179 struct sock *sk = skb->sk;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200180
181 /* We take the more conservative approach and
182 * whitelist socket protocols that may pass.
183 */
184 switch (sk->sk_protocol) {
185 case NETLINK_ROUTE:
186 case NETLINK_USERSOCK:
187 case NETLINK_SOCK_DIAG:
188 case NETLINK_NFLOG:
189 case NETLINK_XFRM:
190 case NETLINK_FIB_LOOKUP:
191 case NETLINK_NETFILTER:
192 case NETLINK_GENERIC:
Varka Bhadram498044b2014-07-16 10:59:47 +0530193 return true;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200194 }
195
Varka Bhadram498044b2014-07-16 10:59:47 +0530196 return false;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200197}
198
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200199static int __netlink_deliver_tap_skb(struct sk_buff *skb,
200 struct net_device *dev)
201{
202 struct sk_buff *nskb;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200203 struct sock *sk = skb->sk;
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200204 int ret = -ENOMEM;
205
206 dev_hold(dev);
207 nskb = skb_clone(skb, GFP_ATOMIC);
208 if (nskb) {
209 nskb->dev = dev;
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200210 nskb->protocol = htons((u16) sk->sk_protocol);
Daniel Borkmann604d13c2013-12-23 14:35:56 +0100211 nskb->pkt_type = netlink_is_kernel(sk) ?
212 PACKET_KERNEL : PACKET_USER;
Daniel Borkmann4e48ed82014-08-07 22:22:47 +0200213 skb_reset_network_header(nskb);
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200214 ret = dev_queue_xmit(nskb);
215 if (unlikely(ret > 0))
216 ret = net_xmit_errno(ret);
217 }
218
219 dev_put(dev);
220 return ret;
221}
222
223static void __netlink_deliver_tap(struct sk_buff *skb)
224{
225 int ret;
226 struct netlink_tap *tmp;
227
Daniel Borkmann5ffd5cd2013-09-05 17:48:47 +0200228 if (!netlink_filter_tap(skb))
229 return;
230
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +0200231 list_for_each_entry_rcu(tmp, &netlink_tap_all, list) {
232 ret = __netlink_deliver_tap_skb(skb, tmp->dev);
233 if (unlikely(ret))
234 break;
235 }
236}
237
238static void netlink_deliver_tap(struct sk_buff *skb)
239{
240 rcu_read_lock();
241
242 if (unlikely(!list_empty(&netlink_tap_all)))
243 __netlink_deliver_tap(skb);
244
245 rcu_read_unlock();
246}
247
Daniel Borkmann73bfd372013-12-23 14:35:55 +0100248static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src,
249 struct sk_buff *skb)
250{
251 if (!(netlink_is_kernel(dst) && netlink_is_kernel(src)))
252 netlink_deliver_tap(skb);
253}
254
Patrick McHardycd1df522013-04-17 06:47:05 +0000255static void netlink_overrun(struct sock *sk)
256{
257 struct netlink_sock *nlk = nlk_sk(sk);
258
259 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
260 if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) {
261 sk->sk_err = ENOBUFS;
262 sk->sk_error_report(sk);
263 }
264 }
265 atomic_inc(&sk->sk_drops);
266}
267
268static void netlink_rcv_wake(struct sock *sk)
269{
270 struct netlink_sock *nlk = nlk_sk(sk);
271
272 if (skb_queue_empty(&sk->sk_receive_queue))
273 clear_bit(NETLINK_CONGESTED, &nlk->state);
274 if (!test_bit(NETLINK_CONGESTED, &nlk->state))
275 wake_up_interruptible(&nlk->wait);
276}
277
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000278#ifdef CONFIG_NETLINK_MMAP
Patrick McHardy9652e932013-04-17 06:47:02 +0000279static bool netlink_skb_is_mmaped(const struct sk_buff *skb)
280{
281 return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED;
282}
283
Patrick McHardyf9c22882013-04-17 06:47:04 +0000284static bool netlink_rx_is_mmaped(struct sock *sk)
285{
286 return nlk_sk(sk)->rx_ring.pg_vec != NULL;
287}
288
Patrick McHardy5fd96122013-04-17 06:47:03 +0000289static bool netlink_tx_is_mmaped(struct sock *sk)
290{
291 return nlk_sk(sk)->tx_ring.pg_vec != NULL;
292}
293
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000294static __pure struct page *pgvec_to_page(const void *addr)
295{
296 if (is_vmalloc_addr(addr))
297 return vmalloc_to_page(addr);
298 else
299 return virt_to_page(addr);
300}
301
302static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len)
303{
304 unsigned int i;
305
306 for (i = 0; i < len; i++) {
307 if (pg_vec[i] != NULL) {
308 if (is_vmalloc_addr(pg_vec[i]))
309 vfree(pg_vec[i]);
310 else
311 free_pages((unsigned long)pg_vec[i], order);
312 }
313 }
314 kfree(pg_vec);
315}
316
317static void *alloc_one_pg_vec_page(unsigned long order)
318{
319 void *buffer;
320 gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO |
321 __GFP_NOWARN | __GFP_NORETRY;
322
323 buffer = (void *)__get_free_pages(gfp_flags, order);
324 if (buffer != NULL)
325 return buffer;
326
327 buffer = vzalloc((1 << order) * PAGE_SIZE);
328 if (buffer != NULL)
329 return buffer;
330
331 gfp_flags &= ~__GFP_NORETRY;
332 return (void *)__get_free_pages(gfp_flags, order);
333}
334
335static void **alloc_pg_vec(struct netlink_sock *nlk,
336 struct nl_mmap_req *req, unsigned int order)
337{
338 unsigned int block_nr = req->nm_block_nr;
339 unsigned int i;
Daniel Borkmann8a849bb2013-08-02 17:32:39 +0200340 void **pg_vec;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000341
342 pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL);
343 if (pg_vec == NULL)
344 return NULL;
345
346 for (i = 0; i < block_nr; i++) {
Daniel Borkmann8a849bb2013-08-02 17:32:39 +0200347 pg_vec[i] = alloc_one_pg_vec_page(order);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000348 if (pg_vec[i] == NULL)
349 goto err1;
350 }
351
352 return pg_vec;
353err1:
354 free_pg_vec(pg_vec, order, block_nr);
355 return NULL;
356}
357
358static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req,
359 bool closing, bool tx_ring)
360{
361 struct netlink_sock *nlk = nlk_sk(sk);
362 struct netlink_ring *ring;
363 struct sk_buff_head *queue;
364 void **pg_vec = NULL;
365 unsigned int order = 0;
366 int err;
367
368 ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring;
369 queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue;
370
371 if (!closing) {
372 if (atomic_read(&nlk->mapped))
373 return -EBUSY;
374 if (atomic_read(&ring->pending))
375 return -EBUSY;
376 }
377
378 if (req->nm_block_nr) {
379 if (ring->pg_vec != NULL)
380 return -EBUSY;
381
382 if ((int)req->nm_block_size <= 0)
383 return -EINVAL;
Tobias Klauser74e83b22014-07-31 12:17:08 +0200384 if (!PAGE_ALIGNED(req->nm_block_size))
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000385 return -EINVAL;
386 if (req->nm_frame_size < NL_MMAP_HDRLEN)
387 return -EINVAL;
388 if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT))
389 return -EINVAL;
390
391 ring->frames_per_block = req->nm_block_size /
392 req->nm_frame_size;
393 if (ring->frames_per_block == 0)
394 return -EINVAL;
395 if (ring->frames_per_block * req->nm_block_nr !=
396 req->nm_frame_nr)
397 return -EINVAL;
398
399 order = get_order(req->nm_block_size);
400 pg_vec = alloc_pg_vec(nlk, req, order);
401 if (pg_vec == NULL)
402 return -ENOMEM;
403 } else {
404 if (req->nm_frame_nr)
405 return -EINVAL;
406 }
407
408 err = -EBUSY;
409 mutex_lock(&nlk->pg_vec_lock);
410 if (closing || atomic_read(&nlk->mapped) == 0) {
411 err = 0;
412 spin_lock_bh(&queue->lock);
413
414 ring->frame_max = req->nm_frame_nr - 1;
415 ring->head = 0;
416 ring->frame_size = req->nm_frame_size;
417 ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE;
418
419 swap(ring->pg_vec_len, req->nm_block_nr);
420 swap(ring->pg_vec_order, order);
421 swap(ring->pg_vec, pg_vec);
422
423 __skb_queue_purge(queue);
424 spin_unlock_bh(&queue->lock);
425
426 WARN_ON(atomic_read(&nlk->mapped));
427 }
428 mutex_unlock(&nlk->pg_vec_lock);
429
430 if (pg_vec)
431 free_pg_vec(pg_vec, order, req->nm_block_nr);
432 return err;
433}
434
435static void netlink_mm_open(struct vm_area_struct *vma)
436{
437 struct file *file = vma->vm_file;
438 struct socket *sock = file->private_data;
439 struct sock *sk = sock->sk;
440
441 if (sk)
442 atomic_inc(&nlk_sk(sk)->mapped);
443}
444
445static void netlink_mm_close(struct vm_area_struct *vma)
446{
447 struct file *file = vma->vm_file;
448 struct socket *sock = file->private_data;
449 struct sock *sk = sock->sk;
450
451 if (sk)
452 atomic_dec(&nlk_sk(sk)->mapped);
453}
454
455static const struct vm_operations_struct netlink_mmap_ops = {
456 .open = netlink_mm_open,
457 .close = netlink_mm_close,
458};
459
460static int netlink_mmap(struct file *file, struct socket *sock,
461 struct vm_area_struct *vma)
462{
463 struct sock *sk = sock->sk;
464 struct netlink_sock *nlk = nlk_sk(sk);
465 struct netlink_ring *ring;
466 unsigned long start, size, expected;
467 unsigned int i;
468 int err = -EINVAL;
469
470 if (vma->vm_pgoff)
471 return -EINVAL;
472
473 mutex_lock(&nlk->pg_vec_lock);
474
475 expected = 0;
476 for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) {
477 if (ring->pg_vec == NULL)
478 continue;
479 expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE;
480 }
481
482 if (expected == 0)
483 goto out;
484
485 size = vma->vm_end - vma->vm_start;
486 if (size != expected)
487 goto out;
488
489 start = vma->vm_start;
490 for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) {
491 if (ring->pg_vec == NULL)
492 continue;
493
494 for (i = 0; i < ring->pg_vec_len; i++) {
495 struct page *page;
496 void *kaddr = ring->pg_vec[i];
497 unsigned int pg_num;
498
499 for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) {
500 page = pgvec_to_page(kaddr);
501 err = vm_insert_page(vma, start, page);
502 if (err < 0)
503 goto out;
504 start += PAGE_SIZE;
505 kaddr += PAGE_SIZE;
506 }
507 }
508 }
509
510 atomic_inc(&nlk->mapped);
511 vma->vm_ops = &netlink_mmap_ops;
512 err = 0;
513out:
514 mutex_unlock(&nlk->pg_vec_lock);
Patrick McHardy7cdbac72013-06-11 02:52:47 -0700515 return err;
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000516}
Patrick McHardy9652e932013-04-17 06:47:02 +0000517
David Miller4682a032014-12-16 17:58:17 -0500518static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len)
Patrick McHardy9652e932013-04-17 06:47:02 +0000519{
520#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1
521 struct page *p_start, *p_end;
522
523 /* First page is flushed through netlink_{get,set}_status */
524 p_start = pgvec_to_page(hdr + PAGE_SIZE);
David Miller4682a032014-12-16 17:58:17 -0500525 p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1);
Patrick McHardy9652e932013-04-17 06:47:02 +0000526 while (p_start <= p_end) {
527 flush_dcache_page(p_start);
528 p_start++;
529 }
530#endif
531}
532
533static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr)
534{
535 smp_rmb();
536 flush_dcache_page(pgvec_to_page(hdr));
537 return hdr->nm_status;
538}
539
540static void netlink_set_status(struct nl_mmap_hdr *hdr,
541 enum nl_mmap_status status)
542{
Thomas Grafa18e6a12014-12-18 10:30:26 +0000543 smp_mb();
Patrick McHardy9652e932013-04-17 06:47:02 +0000544 hdr->nm_status = status;
545 flush_dcache_page(pgvec_to_page(hdr));
Patrick McHardy9652e932013-04-17 06:47:02 +0000546}
547
548static struct nl_mmap_hdr *
549__netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos)
550{
551 unsigned int pg_vec_pos, frame_off;
552
553 pg_vec_pos = pos / ring->frames_per_block;
554 frame_off = pos % ring->frames_per_block;
555
556 return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size);
557}
558
559static struct nl_mmap_hdr *
560netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos,
561 enum nl_mmap_status status)
562{
563 struct nl_mmap_hdr *hdr;
564
565 hdr = __netlink_lookup_frame(ring, pos);
566 if (netlink_get_status(hdr) != status)
567 return NULL;
568
569 return hdr;
570}
571
572static struct nl_mmap_hdr *
573netlink_current_frame(const struct netlink_ring *ring,
574 enum nl_mmap_status status)
575{
576 return netlink_lookup_frame(ring, ring->head, status);
577}
578
579static struct nl_mmap_hdr *
580netlink_previous_frame(const struct netlink_ring *ring,
581 enum nl_mmap_status status)
582{
583 unsigned int prev;
584
585 prev = ring->head ? ring->head - 1 : ring->frame_max;
586 return netlink_lookup_frame(ring, prev, status);
587}
588
589static void netlink_increment_head(struct netlink_ring *ring)
590{
591 ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0;
592}
593
594static void netlink_forward_ring(struct netlink_ring *ring)
595{
596 unsigned int head = ring->head, pos = head;
597 const struct nl_mmap_hdr *hdr;
598
599 do {
600 hdr = __netlink_lookup_frame(ring, pos);
601 if (hdr->nm_status == NL_MMAP_STATUS_UNUSED)
602 break;
603 if (hdr->nm_status != NL_MMAP_STATUS_SKIP)
604 break;
605 netlink_increment_head(ring);
606 } while (ring->head != head);
607}
608
Patrick McHardycd1df522013-04-17 06:47:05 +0000609static bool netlink_dump_space(struct netlink_sock *nlk)
610{
611 struct netlink_ring *ring = &nlk->rx_ring;
612 struct nl_mmap_hdr *hdr;
613 unsigned int n;
614
615 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
616 if (hdr == NULL)
617 return false;
618
619 n = ring->head + ring->frame_max / 2;
620 if (n > ring->frame_max)
621 n -= ring->frame_max;
622
623 hdr = __netlink_lookup_frame(ring, n);
624
625 return hdr->nm_status == NL_MMAP_STATUS_UNUSED;
626}
627
Patrick McHardy9652e932013-04-17 06:47:02 +0000628static unsigned int netlink_poll(struct file *file, struct socket *sock,
629 poll_table *wait)
630{
631 struct sock *sk = sock->sk;
632 struct netlink_sock *nlk = nlk_sk(sk);
633 unsigned int mask;
Patrick McHardycd1df522013-04-17 06:47:05 +0000634 int err;
Patrick McHardy9652e932013-04-17 06:47:02 +0000635
Patrick McHardycd1df522013-04-17 06:47:05 +0000636 if (nlk->rx_ring.pg_vec != NULL) {
637 /* Memory mapped sockets don't call recvmsg(), so flow control
638 * for dumps is performed here. A dump is allowed to continue
639 * if at least half the ring is unused.
640 */
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700641 while (nlk->cb_running && netlink_dump_space(nlk)) {
Patrick McHardycd1df522013-04-17 06:47:05 +0000642 err = netlink_dump(sk);
643 if (err < 0) {
Ben Pfaffac30ef82014-07-09 10:31:22 -0700644 sk->sk_err = -err;
Patrick McHardycd1df522013-04-17 06:47:05 +0000645 sk->sk_error_report(sk);
646 break;
647 }
648 }
649 netlink_rcv_wake(sk);
650 }
Patrick McHardy5fd96122013-04-17 06:47:03 +0000651
Patrick McHardy9652e932013-04-17 06:47:02 +0000652 mask = datagram_poll(file, sock, wait);
653
654 spin_lock_bh(&sk->sk_receive_queue.lock);
655 if (nlk->rx_ring.pg_vec) {
656 netlink_forward_ring(&nlk->rx_ring);
657 if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED))
658 mask |= POLLIN | POLLRDNORM;
659 }
660 spin_unlock_bh(&sk->sk_receive_queue.lock);
661
662 spin_lock_bh(&sk->sk_write_queue.lock);
663 if (nlk->tx_ring.pg_vec) {
664 if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED))
665 mask |= POLLOUT | POLLWRNORM;
666 }
667 spin_unlock_bh(&sk->sk_write_queue.lock);
668
669 return mask;
670}
671
672static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb)
673{
674 return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN);
675}
676
677static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk,
678 struct netlink_ring *ring,
679 struct nl_mmap_hdr *hdr)
680{
681 unsigned int size;
682 void *data;
683
684 size = ring->frame_size - NL_MMAP_HDRLEN;
685 data = (void *)hdr + NL_MMAP_HDRLEN;
686
687 skb->head = data;
688 skb->data = data;
689 skb_reset_tail_pointer(skb);
690 skb->end = skb->tail + size;
691 skb->len = 0;
692
693 skb->destructor = netlink_skb_destructor;
694 NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED;
695 NETLINK_CB(skb).sk = sk;
696}
Patrick McHardy5fd96122013-04-17 06:47:03 +0000697
698static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg,
699 u32 dst_portid, u32 dst_group,
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100700 struct scm_cookie *scm)
Patrick McHardy5fd96122013-04-17 06:47:03 +0000701{
702 struct netlink_sock *nlk = nlk_sk(sk);
703 struct netlink_ring *ring;
704 struct nl_mmap_hdr *hdr;
705 struct sk_buff *skb;
706 unsigned int maxlen;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000707 int err = 0, len = 0;
708
Patrick McHardy5fd96122013-04-17 06:47:03 +0000709 mutex_lock(&nlk->pg_vec_lock);
710
711 ring = &nlk->tx_ring;
712 maxlen = ring->frame_size - NL_MMAP_HDRLEN;
713
714 do {
David Miller4682a032014-12-16 17:58:17 -0500715 unsigned int nm_len;
716
Patrick McHardy5fd96122013-04-17 06:47:03 +0000717 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID);
718 if (hdr == NULL) {
719 if (!(msg->msg_flags & MSG_DONTWAIT) &&
720 atomic_read(&nlk->tx_ring.pending))
721 schedule();
722 continue;
723 }
David Miller4682a032014-12-16 17:58:17 -0500724
725 nm_len = ACCESS_ONCE(hdr->nm_len);
726 if (nm_len > maxlen) {
Patrick McHardy5fd96122013-04-17 06:47:03 +0000727 err = -EINVAL;
728 goto out;
729 }
730
David Miller4682a032014-12-16 17:58:17 -0500731 netlink_frame_flush_dcache(hdr, nm_len);
Patrick McHardy5fd96122013-04-17 06:47:03 +0000732
David Miller4682a032014-12-16 17:58:17 -0500733 skb = alloc_skb(nm_len, GFP_KERNEL);
734 if (skb == NULL) {
735 err = -ENOBUFS;
736 goto out;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000737 }
David Miller4682a032014-12-16 17:58:17 -0500738 __skb_put(skb, nm_len);
739 memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len);
740 netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED);
Patrick McHardy5fd96122013-04-17 06:47:03 +0000741
742 netlink_increment_head(ring);
743
744 NETLINK_CB(skb).portid = nlk->portid;
745 NETLINK_CB(skb).dst_group = dst_group;
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100746 NETLINK_CB(skb).creds = scm->creds;
Patrick McHardy5fd96122013-04-17 06:47:03 +0000747
748 err = security_netlink_send(sk, skb);
749 if (err) {
750 kfree_skb(skb);
751 goto out;
752 }
753
754 if (unlikely(dst_group)) {
755 atomic_inc(&skb->users);
756 netlink_broadcast(sk, skb, dst_portid, dst_group,
757 GFP_KERNEL);
758 }
759 err = netlink_unicast(sk, skb, dst_portid,
760 msg->msg_flags & MSG_DONTWAIT);
761 if (err < 0)
762 goto out;
763 len += err;
764
765 } while (hdr != NULL ||
766 (!(msg->msg_flags & MSG_DONTWAIT) &&
767 atomic_read(&nlk->tx_ring.pending)));
768
769 if (len > 0)
770 err = len;
771out:
772 mutex_unlock(&nlk->pg_vec_lock);
773 return err;
774}
Patrick McHardyf9c22882013-04-17 06:47:04 +0000775
776static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb)
777{
778 struct nl_mmap_hdr *hdr;
779
780 hdr = netlink_mmap_hdr(skb);
781 hdr->nm_len = skb->len;
782 hdr->nm_group = NETLINK_CB(skb).dst_group;
783 hdr->nm_pid = NETLINK_CB(skb).creds.pid;
Nicolas Dichtel1bf93102013-04-24 10:36:23 +0200784 hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid);
785 hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid);
David Miller4682a032014-12-16 17:58:17 -0500786 netlink_frame_flush_dcache(hdr, hdr->nm_len);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000787 netlink_set_status(hdr, NL_MMAP_STATUS_VALID);
788
789 NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED;
790 kfree_skb(skb);
791}
792
793static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb)
794{
795 struct netlink_sock *nlk = nlk_sk(sk);
796 struct netlink_ring *ring = &nlk->rx_ring;
797 struct nl_mmap_hdr *hdr;
798
799 spin_lock_bh(&sk->sk_receive_queue.lock);
800 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
801 if (hdr == NULL) {
802 spin_unlock_bh(&sk->sk_receive_queue.lock);
803 kfree_skb(skb);
Patrick McHardycd1df522013-04-17 06:47:05 +0000804 netlink_overrun(sk);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000805 return;
806 }
807 netlink_increment_head(ring);
808 __skb_queue_tail(&sk->sk_receive_queue, skb);
809 spin_unlock_bh(&sk->sk_receive_queue.lock);
810
811 hdr->nm_len = skb->len;
812 hdr->nm_group = NETLINK_CB(skb).dst_group;
813 hdr->nm_pid = NETLINK_CB(skb).creds.pid;
Nicolas Dichtel1bf93102013-04-24 10:36:23 +0200814 hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid);
815 hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid);
Patrick McHardyf9c22882013-04-17 06:47:04 +0000816 netlink_set_status(hdr, NL_MMAP_STATUS_COPY);
817}
818
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000819#else /* CONFIG_NETLINK_MMAP */
Patrick McHardy9652e932013-04-17 06:47:02 +0000820#define netlink_skb_is_mmaped(skb) false
Patrick McHardyf9c22882013-04-17 06:47:04 +0000821#define netlink_rx_is_mmaped(sk) false
Patrick McHardy5fd96122013-04-17 06:47:03 +0000822#define netlink_tx_is_mmaped(sk) false
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000823#define netlink_mmap sock_no_mmap
Patrick McHardy9652e932013-04-17 06:47:02 +0000824#define netlink_poll datagram_poll
Christoph Hellwig7cc05662015-01-28 18:04:53 +0100825#define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, scm) 0
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000826#endif /* CONFIG_NETLINK_MMAP */
827
Patrick McHardycf0a0182013-04-17 06:47:00 +0000828static void netlink_skb_destructor(struct sk_buff *skb)
829{
Patrick McHardy9652e932013-04-17 06:47:02 +0000830#ifdef CONFIG_NETLINK_MMAP
831 struct nl_mmap_hdr *hdr;
832 struct netlink_ring *ring;
833 struct sock *sk;
834
835 /* If a packet from the kernel to userspace was freed because of an
836 * error without being delivered to userspace, the kernel must reset
837 * the status. In the direction userspace to kernel, the status is
838 * always reset here after the packet was processed and freed.
839 */
840 if (netlink_skb_is_mmaped(skb)) {
841 hdr = netlink_mmap_hdr(skb);
842 sk = NETLINK_CB(skb).sk;
843
Patrick McHardy5fd96122013-04-17 06:47:03 +0000844 if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) {
845 netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED);
846 ring = &nlk_sk(sk)->tx_ring;
847 } else {
848 if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) {
849 hdr->nm_len = 0;
850 netlink_set_status(hdr, NL_MMAP_STATUS_VALID);
851 }
852 ring = &nlk_sk(sk)->rx_ring;
Patrick McHardy9652e932013-04-17 06:47:02 +0000853 }
Patrick McHardy9652e932013-04-17 06:47:02 +0000854
855 WARN_ON(atomic_read(&ring->pending) == 0);
856 atomic_dec(&ring->pending);
857 sock_put(sk);
858
Pablo Neira5e71d9d2013-06-03 09:28:43 +0000859 skb->head = NULL;
Patrick McHardy9652e932013-04-17 06:47:02 +0000860 }
861#endif
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +0000862 if (is_vmalloc_addr(skb->head)) {
Pablo Neira3a365152013-06-28 03:04:23 +0200863 if (!skb->cloned ||
864 !atomic_dec_return(&(skb_shinfo(skb)->dataref)))
865 vfree(skb->head);
866
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +0000867 skb->head = NULL;
868 }
Patrick McHardy9652e932013-04-17 06:47:02 +0000869 if (skb->sk != NULL)
870 sock_rfree(skb);
Patrick McHardycf0a0182013-04-17 06:47:00 +0000871}
872
873static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
874{
875 WARN_ON(skb->sk != NULL);
876 skb->sk = sk;
877 skb->destructor = netlink_skb_destructor;
878 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
879 sk_mem_charge(sk, skb->truesize);
880}
881
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882static void netlink_sock_destruct(struct sock *sk)
883{
Herbert Xu3f660d62007-05-03 03:17:14 -0700884 struct netlink_sock *nlk = nlk_sk(sk);
885
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700886 if (nlk->cb_running) {
887 if (nlk->cb.done)
888 nlk->cb.done(&nlk->cb);
Gao feng6dc878a2012-10-04 20:15:48 +0000889
Pravin B Shelar16b304f2013-08-15 15:31:06 -0700890 module_put(nlk->cb.module);
891 kfree_skb(nlk->cb.skb);
Herbert Xu3f660d62007-05-03 03:17:14 -0700892 }
893
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 skb_queue_purge(&sk->sk_receive_queue);
Patrick McHardyccdfcc32013-04-17 06:47:01 +0000895#ifdef CONFIG_NETLINK_MMAP
896 if (1) {
897 struct nl_mmap_req req;
898
899 memset(&req, 0, sizeof(req));
900 if (nlk->rx_ring.pg_vec)
901 netlink_set_ring(sk, &req, true, false);
902 memset(&req, 0, sizeof(req));
903 if (nlk->tx_ring.pg_vec)
904 netlink_set_ring(sk, &req, true, true);
905 }
906#endif /* CONFIG_NETLINK_MMAP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907
908 if (!sock_flag(sk, SOCK_DEAD)) {
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800909 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 return;
911 }
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700912
913 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
914 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
915 WARN_ON(nlk_sk(sk)->groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800918/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
919 * SMP. Look, when several writers sleep and reader wakes them up, all but one
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
921 * this, _but_ remember, it adds useless work on UP machines.
922 */
923
Johannes Bergd136f1b2009-09-12 03:03:15 +0000924void netlink_table_grab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800925 __acquires(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926{
Johannes Bergd136f1b2009-09-12 03:03:15 +0000927 might_sleep();
928
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700929 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
931 if (atomic_read(&nl_table_users)) {
932 DECLARE_WAITQUEUE(wait, current);
933
934 add_wait_queue_exclusive(&nl_table_wait, &wait);
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800935 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 set_current_state(TASK_UNINTERRUPTIBLE);
937 if (atomic_read(&nl_table_users) == 0)
938 break;
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700939 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 schedule();
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700941 write_lock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 }
943
944 __set_current_state(TASK_RUNNING);
945 remove_wait_queue(&nl_table_wait, &wait);
946 }
947}
948
Johannes Bergd136f1b2009-09-12 03:03:15 +0000949void netlink_table_ungrab(void)
Eric Dumazet9a429c42008-01-01 21:58:02 -0800950 __releases(nl_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951{
Arjan van de Ven6abd2192006-07-03 00:24:07 -0700952 write_unlock_irq(&nl_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 wake_up(&nl_table_wait);
954}
955
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800956static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957netlink_lock_table(void)
958{
959 /* read_lock() synchronizes us to netlink_table_grab */
960
961 read_lock(&nl_table_lock);
962 atomic_inc(&nl_table_users);
963 read_unlock(&nl_table_lock);
964}
965
Patrick McHardy6ac552f2007-12-04 00:19:38 -0800966static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967netlink_unlock_table(void)
968{
969 if (atomic_dec_and_test(&nl_table_users))
970 wake_up(&nl_table_wait);
971}
972
Thomas Grafe3416942014-08-02 11:47:45 +0200973struct netlink_compare_arg
Gao fengda12c902013-06-06 14:49:11 +0800974{
Herbert Xuc428ecd2015-03-20 21:57:01 +1100975 possible_net_t pnet;
Thomas Grafe3416942014-08-02 11:47:45 +0200976 u32 portid;
977};
978
Herbert Xu8f2ddaa2015-03-21 14:14:03 +1100979/* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
980#define netlink_compare_arg_len \
981 (offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
Thomas Grafe3416942014-08-02 11:47:45 +0200982
Herbert Xuc428ecd2015-03-20 21:57:01 +1100983static inline int netlink_compare(struct rhashtable_compare_arg *arg,
984 const void *ptr)
985{
986 const struct netlink_compare_arg *x = arg->key;
987 const struct netlink_sock *nlk = ptr;
988
989 return nlk->portid != x->portid ||
990 !net_eq(sock_net(&nlk->sk), read_pnet(&x->pnet));
991}
992
993static void netlink_compare_arg_init(struct netlink_compare_arg *arg,
994 struct net *net, u32 portid)
995{
996 memset(arg, 0, sizeof(*arg));
997 write_pnet(&arg->pnet, net);
998 arg->portid = portid;
Thomas Grafe3416942014-08-02 11:47:45 +0200999}
1000
1001static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
1002 struct net *net)
1003{
Herbert Xuc428ecd2015-03-20 21:57:01 +11001004 struct netlink_compare_arg arg;
Thomas Grafe3416942014-08-02 11:47:45 +02001005
Herbert Xuc428ecd2015-03-20 21:57:01 +11001006 netlink_compare_arg_init(&arg, net, portid);
1007 return rhashtable_lookup_fast(&table->hash, &arg,
1008 netlink_rhashtable_params);
Gao fengda12c902013-06-06 14:49:11 +08001009}
1010
Herbert Xuc428ecd2015-03-20 21:57:01 +11001011static int __netlink_insert(struct netlink_table *table, struct sock *sk)
Ying Xuec5adde92015-01-12 14:52:23 +08001012{
Herbert Xuc428ecd2015-03-20 21:57:01 +11001013 struct netlink_compare_arg arg;
Ying Xuec5adde92015-01-12 14:52:23 +08001014
Herbert Xuc428ecd2015-03-20 21:57:01 +11001015 netlink_compare_arg_init(&arg, sock_net(sk), nlk_sk(sk)->portid);
1016 return rhashtable_lookup_insert_key(&table->hash, &arg,
1017 &nlk_sk(sk)->node,
1018 netlink_rhashtable_params);
Ying Xuec5adde92015-01-12 14:52:23 +08001019}
1020
Eric W. Biederman15e47302012-09-07 20:12:54 +00001021static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022{
Gao fengda12c902013-06-06 14:49:11 +08001023 struct netlink_table *table = &nl_table[protocol];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
Thomas Grafe3416942014-08-02 11:47:45 +02001026 rcu_read_lock();
1027 sk = __netlink_lookup(table, portid, net);
1028 if (sk)
1029 sock_hold(sk);
1030 rcu_read_unlock();
1031
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 return sk;
1033}
1034
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08001035static const struct proto_ops netlink_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
Patrick McHardy4277a082006-03-20 18:52:01 -08001037static void
1038netlink_update_listeners(struct sock *sk)
1039{
1040 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Patrick McHardy4277a082006-03-20 18:52:01 -08001041 unsigned long mask;
1042 unsigned int i;
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001043 struct listeners *listeners;
1044
1045 listeners = nl_deref_protected(tbl->listeners);
1046 if (!listeners)
1047 return;
Patrick McHardy4277a082006-03-20 18:52:01 -08001048
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001049 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
Patrick McHardy4277a082006-03-20 18:52:01 -08001050 mask = 0;
Sasha Levinb67bfe02013-02-27 17:06:00 -08001051 sk_for_each_bound(sk, &tbl->mc_list) {
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001052 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
1053 mask |= nlk_sk(sk)->groups[i];
1054 }
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001055 listeners->masks[i] = mask;
Patrick McHardy4277a082006-03-20 18:52:01 -08001056 }
1057 /* this function is only called with the netlink table "grabbed", which
1058 * makes sure updates are visible before bind or setsockopt return. */
1059}
1060
Herbert Xu8ea65f42015-01-26 14:02:56 +11001061static int netlink_insert(struct sock *sk, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062{
Gao fengda12c902013-06-06 14:49:11 +08001063 struct netlink_table *table = &nl_table[sk->sk_protocol];
Herbert Xu919d9db2015-01-16 17:23:48 +11001064 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065
Ying Xuec5adde92015-01-12 14:52:23 +08001066 lock_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
1068 err = -EBUSY;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001069 if (nlk_sk(sk)->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 goto err;
1071
1072 err = -ENOMEM;
Thomas Graf97defe12015-01-02 23:00:20 +01001073 if (BITS_PER_LONG > 32 &&
1074 unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 goto err;
1076
Eric W. Biederman15e47302012-09-07 20:12:54 +00001077 nlk_sk(sk)->portid = portid;
Thomas Grafe3416942014-08-02 11:47:45 +02001078 sock_hold(sk);
Herbert Xu919d9db2015-01-16 17:23:48 +11001079
Herbert Xuc428ecd2015-03-20 21:57:01 +11001080 err = __netlink_insert(table, sk);
1081 if (err) {
1082 if (err == -EEXIST)
1083 err = -EADDRINUSE;
Ying Xuec5adde92015-01-12 14:52:23 +08001084 sock_put(sk);
Herbert Xu919d9db2015-01-16 17:23:48 +11001085 }
1086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087err:
Ying Xuec5adde92015-01-12 14:52:23 +08001088 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 return err;
1090}
1091
1092static void netlink_remove(struct sock *sk)
1093{
Thomas Grafe3416942014-08-02 11:47:45 +02001094 struct netlink_table *table;
1095
Thomas Grafe3416942014-08-02 11:47:45 +02001096 table = &nl_table[sk->sk_protocol];
Herbert Xuc428ecd2015-03-20 21:57:01 +11001097 if (!rhashtable_remove_fast(&table->hash, &nlk_sk(sk)->node,
1098 netlink_rhashtable_params)) {
Thomas Grafe3416942014-08-02 11:47:45 +02001099 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
1100 __sock_put(sk);
1101 }
Thomas Grafe3416942014-08-02 11:47:45 +02001102
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 netlink_table_grab();
Johannes Bergb10dcb32014-12-22 18:56:37 +01001104 if (nlk_sk(sk)->subscriptions) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 __sk_del_bind_node(sk);
Johannes Bergb10dcb32014-12-22 18:56:37 +01001106 netlink_update_listeners(sk);
1107 }
Johannes Bergee1c24422015-01-16 11:37:14 +01001108 if (sk->sk_protocol == NETLINK_GENERIC)
1109 atomic_inc(&genl_sk_destructing_cnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 netlink_table_ungrab();
1111}
1112
1113static struct proto netlink_proto = {
1114 .name = "NETLINK",
1115 .owner = THIS_MODULE,
1116 .obj_size = sizeof(struct netlink_sock),
1117};
1118
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001119static int __netlink_create(struct net *net, struct socket *sock,
1120 struct mutex *cb_mutex, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121{
1122 struct sock *sk;
1123 struct netlink_sock *nlk;
Patrick McHardyab33a172005-08-14 19:31:36 -07001124
1125 sock->ops = &netlink_ops;
1126
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001127 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
Patrick McHardyab33a172005-08-14 19:31:36 -07001128 if (!sk)
1129 return -ENOMEM;
1130
1131 sock_init_data(sock, sk);
1132
1133 nlk = nlk_sk(sk);
Eric Dumazet658cb352012-04-22 21:30:21 +00001134 if (cb_mutex) {
Patrick McHardyffa4d722007-04-25 14:01:17 -07001135 nlk->cb_mutex = cb_mutex;
Eric Dumazet658cb352012-04-22 21:30:21 +00001136 } else {
Patrick McHardyffa4d722007-04-25 14:01:17 -07001137 nlk->cb_mutex = &nlk->cb_def_mutex;
1138 mutex_init(nlk->cb_mutex);
1139 }
Patrick McHardyab33a172005-08-14 19:31:36 -07001140 init_waitqueue_head(&nlk->wait);
Patrick McHardyccdfcc32013-04-17 06:47:01 +00001141#ifdef CONFIG_NETLINK_MMAP
1142 mutex_init(&nlk->pg_vec_lock);
1143#endif
Patrick McHardyab33a172005-08-14 19:31:36 -07001144
1145 sk->sk_destruct = netlink_sock_destruct;
1146 sk->sk_protocol = protocol;
1147 return 0;
1148}
1149
Eric Paris3f378b62009-11-05 22:18:14 -08001150static int netlink_create(struct net *net, struct socket *sock, int protocol,
1151 int kern)
Patrick McHardyab33a172005-08-14 19:31:36 -07001152{
1153 struct module *module = NULL;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001154 struct mutex *cb_mutex;
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001155 struct netlink_sock *nlk;
Johannes Berg023e2cf2014-12-23 21:00:06 +01001156 int (*bind)(struct net *net, int group);
1157 void (*unbind)(struct net *net, int group);
Patrick McHardyab33a172005-08-14 19:31:36 -07001158 int err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
1160 sock->state = SS_UNCONNECTED;
1161
1162 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
1163 return -ESOCKTNOSUPPORT;
1164
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001165 if (protocol < 0 || protocol >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 return -EPROTONOSUPPORT;
1167
Patrick McHardy77247bb2005-08-14 19:27:13 -07001168 netlink_lock_table();
Johannes Berg95a5afc2008-10-16 15:24:51 -07001169#ifdef CONFIG_MODULES
Patrick McHardyab33a172005-08-14 19:31:36 -07001170 if (!nl_table[protocol].registered) {
Patrick McHardy77247bb2005-08-14 19:27:13 -07001171 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001172 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
Patrick McHardy77247bb2005-08-14 19:27:13 -07001173 netlink_lock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001174 }
Patrick McHardyab33a172005-08-14 19:31:36 -07001175#endif
1176 if (nl_table[protocol].registered &&
1177 try_module_get(nl_table[protocol].module))
1178 module = nl_table[protocol].module;
Alexey Dobriyan974c37e2010-01-30 10:05:05 +00001179 else
1180 err = -EPROTONOSUPPORT;
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07001181 cb_mutex = nl_table[protocol].cb_mutex;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00001182 bind = nl_table[protocol].bind;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001183 unbind = nl_table[protocol].unbind;
Patrick McHardy77247bb2005-08-14 19:27:13 -07001184 netlink_unlock_table();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001185
Alexey Dobriyan974c37e2010-01-30 10:05:05 +00001186 if (err < 0)
1187 goto out;
1188
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001189 err = __netlink_create(net, sock, cb_mutex, protocol);
1190 if (err < 0)
Patrick McHardyab33a172005-08-14 19:31:36 -07001191 goto out_module;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192
David S. Miller6f756a82008-11-23 17:34:03 -08001193 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -08001194 sock_prot_inuse_add(net, &netlink_proto, 1);
David S. Miller6f756a82008-11-23 17:34:03 -08001195 local_bh_enable();
1196
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001197 nlk = nlk_sk(sock->sk);
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001198 nlk->module = module;
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00001199 nlk->netlink_bind = bind;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001200 nlk->netlink_unbind = unbind;
Patrick McHardyab33a172005-08-14 19:31:36 -07001201out:
1202 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
Patrick McHardyab33a172005-08-14 19:31:36 -07001204out_module:
1205 module_put(module);
1206 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207}
1208
Thomas Graf21e49022015-01-02 23:00:22 +01001209static void deferred_put_nlk_sk(struct rcu_head *head)
1210{
1211 struct netlink_sock *nlk = container_of(head, struct netlink_sock, rcu);
1212
1213 sock_put(&nlk->sk);
1214}
1215
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216static int netlink_release(struct socket *sock)
1217{
1218 struct sock *sk = sock->sk;
1219 struct netlink_sock *nlk;
1220
1221 if (!sk)
1222 return 0;
1223
1224 netlink_remove(sk);
Denis Lunevac57b3a2007-04-18 17:05:58 -07001225 sock_orphan(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 nlk = nlk_sk(sk);
1227
Herbert Xu3f660d62007-05-03 03:17:14 -07001228 /*
1229 * OK. Socket is unlinked, any packets that arrive now
1230 * will be purged.
1231 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232
Johannes Bergee1c24422015-01-16 11:37:14 +01001233 /* must not acquire netlink_table_lock in any way again before unbind
1234 * and notifying genetlink is done as otherwise it might deadlock
1235 */
1236 if (nlk->netlink_unbind) {
1237 int i;
1238
1239 for (i = 0; i < nlk->ngroups; i++)
1240 if (test_bit(i, nlk->groups))
1241 nlk->netlink_unbind(sock_net(sk), i + 1);
1242 }
1243 if (sk->sk_protocol == NETLINK_GENERIC &&
1244 atomic_dec_return(&genl_sk_destructing_cnt) == 0)
1245 wake_up(&genl_sk_destructing_waitq);
1246
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 sock->sk = NULL;
1248 wake_up_interruptible_all(&nlk->wait);
1249
1250 skb_queue_purge(&sk->sk_write_queue);
1251
Eric W. Biederman15e47302012-09-07 20:12:54 +00001252 if (nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 struct netlink_notify n = {
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001254 .net = sock_net(sk),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 .protocol = sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00001256 .portid = nlk->portid,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 };
Alan Sterne041c682006-03-27 01:16:30 -08001258 atomic_notifier_call_chain(&netlink_chain,
1259 NETLINK_URELEASE, &n);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001260 }
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001261
Mariusz Kozlowski5e7c0012007-01-02 15:24:30 -08001262 module_put(nlk->module);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07001263
Denis V. Lunevaed81562007-10-10 21:14:32 -07001264 if (netlink_is_kernel(sk)) {
Johannes Bergb10dcb32014-12-22 18:56:37 +01001265 netlink_table_grab();
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001266 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
1267 if (--nl_table[sk->sk_protocol].registered == 0) {
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001268 struct listeners *old;
1269
1270 old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
1271 RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
1272 kfree_rcu(old, rcu);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001273 nl_table[sk->sk_protocol].module = NULL;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001274 nl_table[sk->sk_protocol].bind = NULL;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001275 nl_table[sk->sk_protocol].unbind = NULL;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001276 nl_table[sk->sk_protocol].flags = 0;
Denis V. Lunev869e58f2008-01-18 23:53:31 -08001277 nl_table[sk->sk_protocol].registered = 0;
1278 }
Johannes Bergb10dcb32014-12-22 18:56:37 +01001279 netlink_table_ungrab();
Eric Dumazet658cb352012-04-22 21:30:21 +00001280 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07001281
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001282 kfree(nlk->groups);
1283 nlk->groups = NULL;
1284
Eric Dumazet37558102008-11-24 14:05:22 -08001285 local_bh_disable();
Eric Dumazetc1fd3b92008-11-23 15:48:22 -08001286 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
Eric Dumazet37558102008-11-24 14:05:22 -08001287 local_bh_enable();
Thomas Graf21e49022015-01-02 23:00:22 +01001288 call_rcu(&nlk->rcu, deferred_put_nlk_sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 return 0;
1290}
1291
1292static int netlink_autobind(struct socket *sock)
1293{
1294 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001295 struct net *net = sock_net(sk);
Gao fengda12c902013-06-06 14:49:11 +08001296 struct netlink_table *table = &nl_table[sk->sk_protocol];
Eric W. Biederman15e47302012-09-07 20:12:54 +00001297 s32 portid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 int err;
1299 static s32 rover = -4097;
1300
1301retry:
1302 cond_resched();
Thomas Grafe3416942014-08-02 11:47:45 +02001303 rcu_read_lock();
1304 if (__netlink_lookup(table, portid, net)) {
1305 /* Bind collision, search negative portid values. */
1306 portid = rover--;
1307 if (rover > -4097)
1308 rover = -4097;
1309 rcu_read_unlock();
1310 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 }
Thomas Grafe3416942014-08-02 11:47:45 +02001312 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
Herbert Xu8ea65f42015-01-26 14:02:56 +11001314 err = netlink_insert(sk, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 if (err == -EADDRINUSE)
1316 goto retry;
David S. Millerd470e3b2005-06-26 15:31:51 -07001317
1318 /* If 2 threads race to autobind, that is fine. */
1319 if (err == -EBUSY)
1320 err = 0;
1321
1322 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323}
1324
Eric W. Biedermanaa4cf942014-04-23 14:28:03 -07001325/**
1326 * __netlink_ns_capable - General netlink message capability test
1327 * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
1328 * @user_ns: The user namespace of the capability to use
1329 * @cap: The capability to use
1330 *
1331 * Test to see if the opener of the socket we received the message
1332 * from had when the netlink socket was created and the sender of the
1333 * message has has the capability @cap in the user namespace @user_ns.
1334 */
1335bool __netlink_ns_capable(const struct netlink_skb_parms *nsp,
1336 struct user_namespace *user_ns, int cap)
1337{
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07001338 return ((nsp->flags & NETLINK_SKB_DST) ||
1339 file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) &&
1340 ns_capable(user_ns, cap);
Eric W. Biedermanaa4cf942014-04-23 14:28:03 -07001341}
1342EXPORT_SYMBOL(__netlink_ns_capable);
1343
1344/**
1345 * netlink_ns_capable - General netlink message capability test
1346 * @skb: socket buffer holding a netlink command from userspace
1347 * @user_ns: The user namespace of the capability to use
1348 * @cap: The capability to use
1349 *
1350 * Test to see if the opener of the socket we received the message
1351 * from had when the netlink socket was created and the sender of the
1352 * message has has the capability @cap in the user namespace @user_ns.
1353 */
1354bool netlink_ns_capable(const struct sk_buff *skb,
1355 struct user_namespace *user_ns, int cap)
1356{
1357 return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap);
1358}
1359EXPORT_SYMBOL(netlink_ns_capable);
1360
1361/**
1362 * netlink_capable - Netlink global message capability test
1363 * @skb: socket buffer holding a netlink command from userspace
1364 * @cap: The capability to use
1365 *
1366 * Test to see if the opener of the socket we received the message
1367 * from had when the netlink socket was created and the sender of the
1368 * message has has the capability @cap in all user namespaces.
1369 */
1370bool netlink_capable(const struct sk_buff *skb, int cap)
1371{
1372 return netlink_ns_capable(skb, &init_user_ns, cap);
1373}
1374EXPORT_SYMBOL(netlink_capable);
1375
1376/**
1377 * netlink_net_capable - Netlink network namespace message capability test
1378 * @skb: socket buffer holding a netlink command from userspace
1379 * @cap: The capability to use
1380 *
1381 * Test to see if the opener of the socket we received the message
1382 * from had when the netlink socket was created and the sender of the
1383 * message has has the capability @cap over the network namespace of
1384 * the socket we received the message from.
1385 */
1386bool netlink_net_capable(const struct sk_buff *skb, int cap)
1387{
1388 return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap);
1389}
1390EXPORT_SYMBOL(netlink_net_capable);
1391
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001392static inline int netlink_allowed(const struct socket *sock, unsigned int flag)
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001393{
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00001394 return (nl_table[sock->sk->sk_protocol].flags & flag) ||
Eric W. Biedermandf008c92012-11-16 03:03:07 +00001395 ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001396}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001398static void
1399netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
1400{
1401 struct netlink_sock *nlk = nlk_sk(sk);
1402
1403 if (nlk->subscriptions && !subscriptions)
1404 __sk_del_bind_node(sk);
1405 else if (!nlk->subscriptions && subscriptions)
1406 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
1407 nlk->subscriptions = subscriptions;
1408}
1409
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001410static int netlink_realloc_groups(struct sock *sk)
Patrick McHardy513c2502005-09-06 15:43:59 -07001411{
1412 struct netlink_sock *nlk = nlk_sk(sk);
1413 unsigned int groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001414 unsigned long *new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -07001415 int err = 0;
1416
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001417 netlink_table_grab();
1418
Patrick McHardy513c2502005-09-06 15:43:59 -07001419 groups = nl_table[sk->sk_protocol].groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001420 if (!nl_table[sk->sk_protocol].registered) {
Patrick McHardy513c2502005-09-06 15:43:59 -07001421 err = -ENOENT;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001422 goto out_unlock;
1423 }
Patrick McHardy513c2502005-09-06 15:43:59 -07001424
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001425 if (nlk->ngroups >= groups)
1426 goto out_unlock;
Patrick McHardy513c2502005-09-06 15:43:59 -07001427
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001428 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
1429 if (new_groups == NULL) {
1430 err = -ENOMEM;
1431 goto out_unlock;
1432 }
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001433 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001434 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
1435
1436 nlk->groups = new_groups;
Patrick McHardy513c2502005-09-06 15:43:59 -07001437 nlk->ngroups = groups;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001438 out_unlock:
1439 netlink_table_ungrab();
1440 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -07001441}
1442
Johannes Berg02c81ab2014-12-22 18:56:35 +01001443static void netlink_undo_bind(int group, long unsigned int groups,
Johannes Berg023e2cf2014-12-23 21:00:06 +01001444 struct sock *sk)
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001445{
Johannes Berg023e2cf2014-12-23 21:00:06 +01001446 struct netlink_sock *nlk = nlk_sk(sk);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001447 int undo;
1448
1449 if (!nlk->netlink_unbind)
1450 return;
1451
1452 for (undo = 0; undo < group; undo++)
Hiroaki SHIMODA6251edd2014-11-13 04:24:10 +09001453 if (test_bit(undo, &groups))
Pablo Neira8b7c36d2015-01-29 10:51:53 +01001454 nlk->netlink_unbind(sock_net(sk), undo + 1);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001455}
1456
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001457static int netlink_bind(struct socket *sock, struct sockaddr *addr,
1458 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459{
1460 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001461 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 struct netlink_sock *nlk = nlk_sk(sk);
1463 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
1464 int err;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001465 long unsigned int groups = nladdr->nl_groups;
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001466
Hannes Frederic Sowa4e4b5372012-12-15 15:42:19 +00001467 if (addr_len < sizeof(struct sockaddr_nl))
1468 return -EINVAL;
1469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 if (nladdr->nl_family != AF_NETLINK)
1471 return -EINVAL;
1472
1473 /* Only superuser is allowed to listen multicasts */
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001474 if (groups) {
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001475 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy513c2502005-09-06 15:43:59 -07001476 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001477 err = netlink_realloc_groups(sk);
1478 if (err)
1479 return err;
Patrick McHardy513c2502005-09-06 15:43:59 -07001480 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001482 if (nlk->portid)
Eric W. Biederman15e47302012-09-07 20:12:54 +00001483 if (nladdr->nl_pid != nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 return -EINVAL;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001485
1486 if (nlk->netlink_bind && groups) {
1487 int group;
1488
1489 for (group = 0; group < nlk->ngroups; group++) {
1490 if (!test_bit(group, &groups))
1491 continue;
Pablo Neira8b7c36d2015-01-29 10:51:53 +01001492 err = nlk->netlink_bind(net, group + 1);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001493 if (!err)
1494 continue;
Johannes Berg023e2cf2014-12-23 21:00:06 +01001495 netlink_undo_bind(group, groups, sk);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001496 return err;
1497 }
1498 }
1499
1500 if (!nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 err = nladdr->nl_pid ?
Herbert Xu8ea65f42015-01-26 14:02:56 +11001502 netlink_insert(sk, nladdr->nl_pid) :
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 netlink_autobind(sock);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001504 if (err) {
Johannes Berg023e2cf2014-12-23 21:00:06 +01001505 netlink_undo_bind(nlk->ngroups, groups, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 return err;
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001507 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 }
1509
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001510 if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 return 0;
1512
1513 netlink_table_grab();
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001514 netlink_update_subscriptions(sk, nlk->subscriptions +
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001515 hweight32(groups) -
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001516 hweight32(nlk->groups[0]));
Richard Guy Briggs4f520902014-04-22 21:31:54 -04001517 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups;
Patrick McHardy4277a082006-03-20 18:52:01 -08001518 netlink_update_listeners(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 netlink_table_ungrab();
1520
1521 return 0;
1522}
1523
1524static int netlink_connect(struct socket *sock, struct sockaddr *addr,
1525 int alen, int flags)
1526{
1527 int err = 0;
1528 struct sock *sk = sock->sk;
1529 struct netlink_sock *nlk = nlk_sk(sk);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001530 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Changli Gao6503d962010-03-31 22:58:26 +00001532 if (alen < sizeof(addr->sa_family))
1533 return -EINVAL;
1534
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 if (addr->sa_family == AF_UNSPEC) {
1536 sk->sk_state = NETLINK_UNCONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001537 nlk->dst_portid = 0;
Patrick McHardyd629b832005-08-14 19:27:50 -07001538 nlk->dst_group = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return 0;
1540 }
1541 if (addr->sa_family != AF_NETLINK)
1542 return -EINVAL;
1543
Mike Pecovnik46833a82014-02-24 21:11:16 +01001544 if ((nladdr->nl_groups || nladdr->nl_pid) &&
Eric W. Biederman5187cd02014-04-23 14:25:48 -07001545 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 return -EPERM;
1547
Eric W. Biederman15e47302012-09-07 20:12:54 +00001548 if (!nlk->portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 err = netlink_autobind(sock);
1550
1551 if (err == 0) {
1552 sk->sk_state = NETLINK_CONNECTED;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001553 nlk->dst_portid = nladdr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001554 nlk->dst_group = ffs(nladdr->nl_groups);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 }
1556
1557 return err;
1558}
1559
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001560static int netlink_getname(struct socket *sock, struct sockaddr *addr,
1561 int *addr_len, int peer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562{
1563 struct sock *sk = sock->sk;
1564 struct netlink_sock *nlk = nlk_sk(sk);
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001565 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09001566
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 nladdr->nl_family = AF_NETLINK;
1568 nladdr->nl_pad = 0;
1569 *addr_len = sizeof(*nladdr);
1570
1571 if (peer) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001572 nladdr->nl_pid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07001573 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +00001575 nladdr->nl_pid = nlk->portid;
Patrick McHardy513c2502005-09-06 15:43:59 -07001576 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 }
1578 return 0;
1579}
1580
Eric W. Biederman15e47302012-09-07 20:12:54 +00001581static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 struct sock *sock;
1584 struct netlink_sock *nlk;
1585
Eric W. Biederman15e47302012-09-07 20:12:54 +00001586 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 if (!sock)
1588 return ERR_PTR(-ECONNREFUSED);
1589
1590 /* Don't bother queuing skb if kernel socket has no input function */
1591 nlk = nlk_sk(sock);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001592 if (sock->sk_state == NETLINK_CONNECTED &&
Eric W. Biederman15e47302012-09-07 20:12:54 +00001593 nlk->dst_portid != nlk_sk(ssk)->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 sock_put(sock);
1595 return ERR_PTR(-ECONNREFUSED);
1596 }
1597 return sock;
1598}
1599
1600struct sock *netlink_getsockbyfilp(struct file *filp)
1601{
Al Viro496ad9a2013-01-23 17:07:38 -05001602 struct inode *inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 struct sock *sock;
1604
1605 if (!S_ISSOCK(inode->i_mode))
1606 return ERR_PTR(-ENOTSOCK);
1607
1608 sock = SOCKET_I(inode)->sk;
1609 if (sock->sk_family != AF_NETLINK)
1610 return ERR_PTR(-EINVAL);
1611
1612 sock_hold(sock);
1613 return sock;
1614}
1615
Pablo Neira3a365152013-06-28 03:04:23 +02001616static struct sk_buff *netlink_alloc_large_skb(unsigned int size,
1617 int broadcast)
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001618{
1619 struct sk_buff *skb;
1620 void *data;
1621
Pablo Neira3a365152013-06-28 03:04:23 +02001622 if (size <= NLMSG_GOODSIZE || broadcast)
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001623 return alloc_skb(size, GFP_KERNEL);
1624
Pablo Neira3a365152013-06-28 03:04:23 +02001625 size = SKB_DATA_ALIGN(size) +
1626 SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001627
1628 data = vmalloc(size);
1629 if (data == NULL)
Pablo Neira3a365152013-06-28 03:04:23 +02001630 return NULL;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001631
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001632 skb = __build_skb(data, size);
Pablo Neira3a365152013-06-28 03:04:23 +02001633 if (skb == NULL)
1634 vfree(data);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001635 else
Pablo Neira3a365152013-06-28 03:04:23 +02001636 skb->destructor = netlink_skb_destructor;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001637
1638 return skb;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001639}
1640
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641/*
1642 * Attach a skb to a netlink socket.
1643 * The caller must hold a reference to the destination socket. On error, the
1644 * reference is dropped. The skb is not send to the destination, just all
1645 * all error checks are performed and memory in the queue is reserved.
1646 * Return values:
1647 * < 0: error. skb freed, reference to sock dropped.
1648 * 0: continue
1649 * 1: repeat lookup - reference dropped while waiting for socket memory.
1650 */
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001651int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001652 long *timeo, struct sock *ssk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653{
1654 struct netlink_sock *nlk;
1655
1656 nlk = nlk_sk(sk);
1657
Patrick McHardy5fd96122013-04-17 06:47:03 +00001658 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
1659 test_bit(NETLINK_CONGESTED, &nlk->state)) &&
1660 !netlink_skb_is_mmaped(skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 DECLARE_WAITQUEUE(wait, current);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001662 if (!*timeo) {
Denis V. Lunevaed81562007-10-10 21:14:32 -07001663 if (!ssk || netlink_is_kernel(ssk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 netlink_overrun(sk);
1665 sock_put(sk);
1666 kfree_skb(skb);
1667 return -EAGAIN;
1668 }
1669
1670 __set_current_state(TASK_INTERRUPTIBLE);
1671 add_wait_queue(&nlk->wait, &wait);
1672
1673 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
Patrick McHardycd967e02013-04-17 06:46:56 +00001674 test_bit(NETLINK_CONGESTED, &nlk->state)) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 !sock_flag(sk, SOCK_DEAD))
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001676 *timeo = schedule_timeout(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
1678 __set_current_state(TASK_RUNNING);
1679 remove_wait_queue(&nlk->wait, &wait);
1680 sock_put(sk);
1681
1682 if (signal_pending(current)) {
1683 kfree_skb(skb);
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001684 return sock_intr_errno(*timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 }
1686 return 1;
1687 }
Patrick McHardycf0a0182013-04-17 06:47:00 +00001688 netlink_skb_set_owner_r(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 return 0;
1690}
1691
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001692static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 int len = skb->len;
1695
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +02001696 netlink_deliver_tap(skb);
1697
Patrick McHardyf9c22882013-04-17 06:47:04 +00001698#ifdef CONFIG_NETLINK_MMAP
1699 if (netlink_skb_is_mmaped(skb))
1700 netlink_queue_mmaped_skb(sk, skb);
1701 else if (netlink_rx_is_mmaped(sk))
1702 netlink_ring_set_copied(sk, skb);
1703 else
1704#endif /* CONFIG_NETLINK_MMAP */
1705 skb_queue_tail(&sk->sk_receive_queue, skb);
David S. Miller676d2362014-04-11 16:15:36 -04001706 sk->sk_data_ready(sk);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001707 return len;
1708}
1709
1710int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
1711{
1712 int len = __netlink_sendskb(sk, skb);
1713
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 sock_put(sk);
1715 return len;
1716}
1717
1718void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
1719{
1720 kfree_skb(skb);
1721 sock_put(sk);
1722}
1723
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001724static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
1726 int delta;
1727
Patrick McHardy1298ca42013-04-17 06:46:59 +00001728 WARN_ON(skb->sk != NULL);
Patrick McHardy5fd96122013-04-17 06:47:03 +00001729 if (netlink_skb_is_mmaped(skb))
1730 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001732 delta = skb->end - skb->tail;
Pablo Neira Ayusoc05cdb12013-06-03 09:46:28 +00001733 if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 return skb;
1735
1736 if (skb_shared(skb)) {
1737 struct sk_buff *nskb = skb_clone(skb, allocation);
1738 if (!nskb)
1739 return skb;
Eric Dumazet8460c002012-04-19 02:24:28 +00001740 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 skb = nskb;
1742 }
1743
1744 if (!pskb_expand_head(skb, 0, -delta, allocation))
1745 skb->truesize -= delta;
1746
1747 return skb;
1748}
1749
Eric W. Biederman3fbc2902012-05-24 17:21:27 -06001750static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb,
1751 struct sock *ssk)
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001752{
1753 int ret;
1754 struct netlink_sock *nlk = nlk_sk(sk);
1755
1756 ret = -ECONNREFUSED;
1757 if (nlk->netlink_rcv != NULL) {
1758 ret = skb->len;
Patrick McHardycf0a0182013-04-17 06:47:00 +00001759 netlink_skb_set_owner_r(skb, sk);
Patrick McHardye32123e2013-04-17 06:46:57 +00001760 NETLINK_CB(skb).sk = ssk;
Daniel Borkmann73bfd372013-12-23 14:35:55 +01001761 netlink_deliver_tap_kernel(sk, ssk, skb);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001762 nlk->netlink_rcv(skb);
Eric Dumazetbfb253c2012-04-22 21:30:29 +00001763 consume_skb(skb);
1764 } else {
1765 kfree_skb(skb);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001766 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001767 sock_put(sk);
1768 return ret;
1769}
1770
1771int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
Eric W. Biederman15e47302012-09-07 20:12:54 +00001772 u32 portid, int nonblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773{
1774 struct sock *sk;
1775 int err;
1776 long timeo;
1777
1778 skb = netlink_trim(skb, gfp_any());
1779
1780 timeo = sock_sndtimeo(ssk, nonblock);
1781retry:
Eric W. Biederman15e47302012-09-07 20:12:54 +00001782 sk = netlink_getsockbyportid(ssk, portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 if (IS_ERR(sk)) {
1784 kfree_skb(skb);
1785 return PTR_ERR(sk);
1786 }
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001787 if (netlink_is_kernel(sk))
Eric W. Biederman3fbc2902012-05-24 17:21:27 -06001788 return netlink_unicast_kernel(sk, skb, ssk);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07001789
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001790 if (sk_filter(sk, skb)) {
Wang Chen84874602008-07-01 19:55:09 -07001791 err = skb->len;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001792 kfree_skb(skb);
1793 sock_put(sk);
1794 return err;
1795 }
1796
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001797 err = netlink_attachskb(sk, skb, &timeo, ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 if (err == 1)
1799 goto retry;
1800 if (err)
1801 return err;
1802
Denis V. Lunev7ee015e2007-10-10 21:14:03 -07001803 return netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08001805EXPORT_SYMBOL(netlink_unicast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806
Patrick McHardyf9c22882013-04-17 06:47:04 +00001807struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size,
1808 u32 dst_portid, gfp_t gfp_mask)
1809{
1810#ifdef CONFIG_NETLINK_MMAP
1811 struct sock *sk = NULL;
1812 struct sk_buff *skb;
1813 struct netlink_ring *ring;
1814 struct nl_mmap_hdr *hdr;
1815 unsigned int maxlen;
1816
1817 sk = netlink_getsockbyportid(ssk, dst_portid);
1818 if (IS_ERR(sk))
1819 goto out;
1820
1821 ring = &nlk_sk(sk)->rx_ring;
1822 /* fast-path without atomic ops for common case: non-mmaped receiver */
1823 if (ring->pg_vec == NULL)
1824 goto out_put;
1825
Thomas Grafaae9f0e2013-11-30 13:21:31 +01001826 if (ring->frame_size - NL_MMAP_HDRLEN < size)
1827 goto out_put;
1828
Patrick McHardyf9c22882013-04-17 06:47:04 +00001829 skb = alloc_skb_head(gfp_mask);
1830 if (skb == NULL)
1831 goto err1;
1832
1833 spin_lock_bh(&sk->sk_receive_queue.lock);
1834 /* check again under lock */
1835 if (ring->pg_vec == NULL)
1836 goto out_free;
1837
Thomas Grafaae9f0e2013-11-30 13:21:31 +01001838 /* check again under lock */
Patrick McHardyf9c22882013-04-17 06:47:04 +00001839 maxlen = ring->frame_size - NL_MMAP_HDRLEN;
1840 if (maxlen < size)
1841 goto out_free;
1842
1843 netlink_forward_ring(ring);
1844 hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED);
1845 if (hdr == NULL)
1846 goto err2;
1847 netlink_ring_setup_skb(skb, sk, ring, hdr);
1848 netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED);
1849 atomic_inc(&ring->pending);
1850 netlink_increment_head(ring);
1851
1852 spin_unlock_bh(&sk->sk_receive_queue.lock);
1853 return skb;
1854
1855err2:
1856 kfree_skb(skb);
1857 spin_unlock_bh(&sk->sk_receive_queue.lock);
Patrick McHardycd1df522013-04-17 06:47:05 +00001858 netlink_overrun(sk);
Patrick McHardyf9c22882013-04-17 06:47:04 +00001859err1:
1860 sock_put(sk);
1861 return NULL;
1862
1863out_free:
1864 kfree_skb(skb);
1865 spin_unlock_bh(&sk->sk_receive_queue.lock);
1866out_put:
1867 sock_put(sk);
1868out:
1869#endif
1870 return alloc_skb(size, gfp_mask);
1871}
1872EXPORT_SYMBOL_GPL(netlink_alloc_skb);
1873
Patrick McHardy4277a082006-03-20 18:52:01 -08001874int netlink_has_listeners(struct sock *sk, unsigned int group)
1875{
1876 int res = 0;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001877 struct listeners *listeners;
Patrick McHardy4277a082006-03-20 18:52:01 -08001878
Denis V. Lunevaed81562007-10-10 21:14:32 -07001879 BUG_ON(!netlink_is_kernel(sk));
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001880
1881 rcu_read_lock();
1882 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
1883
Eric Dumazet6d772ac2012-10-18 03:21:55 +00001884 if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00001885 res = test_bit(group - 1, listeners->masks);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07001886
1887 rcu_read_unlock();
1888
Patrick McHardy4277a082006-03-20 18:52:01 -08001889 return res;
1890}
1891EXPORT_SYMBOL_GPL(netlink_has_listeners);
1892
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00001893static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894{
1895 struct netlink_sock *nlk = nlk_sk(sk);
1896
1897 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
Patrick McHardycd967e02013-04-17 06:46:56 +00001898 !test_bit(NETLINK_CONGESTED, &nlk->state)) {
Patrick McHardycf0a0182013-04-17 06:47:00 +00001899 netlink_skb_set_owner_r(skb, sk);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00001900 __netlink_sendskb(sk, skb);
stephen hemminger2c6458002011-12-22 08:52:03 +00001901 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902 }
1903 return -1;
1904}
1905
1906struct netlink_broadcast_data {
1907 struct sock *exclude_sk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001908 struct net *net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001909 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 u32 group;
1911 int failure;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001912 int delivery_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 int congested;
1914 int delivered;
Al Viro7d877f32005-10-21 03:20:43 -04001915 gfp_t allocation;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 struct sk_buff *skb, *skb2;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001917 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
1918 void *tx_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919};
1920
Rami Rosen46c95212014-07-01 21:17:35 +03001921static void do_one_broadcast(struct sock *sk,
1922 struct netlink_broadcast_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923{
1924 struct netlink_sock *nlk = nlk_sk(sk);
1925 int val;
1926
1927 if (p->exclude_sk == sk)
Rami Rosen46c95212014-07-01 21:17:35 +03001928 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929
Eric W. Biederman15e47302012-09-07 20:12:54 +00001930 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07001931 !test_bit(p->group - 1, nlk->groups))
Rami Rosen46c95212014-07-01 21:17:35 +03001932 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +09001934 if (!net_eq(sock_net(sk), p->net))
Rami Rosen46c95212014-07-01 21:17:35 +03001935 return;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001936
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 if (p->failure) {
1938 netlink_overrun(sk);
Rami Rosen46c95212014-07-01 21:17:35 +03001939 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 }
1941
1942 sock_hold(sk);
1943 if (p->skb2 == NULL) {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001944 if (skb_shared(p->skb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 p->skb2 = skb_clone(p->skb, p->allocation);
1946 } else {
Tommy S. Christensen68acc022005-05-19 13:06:35 -07001947 p->skb2 = skb_get(p->skb);
1948 /*
1949 * skb ownership may have been set when
1950 * delivered to a previous socket.
1951 */
1952 skb_orphan(p->skb2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 }
1954 }
1955 if (p->skb2 == NULL) {
1956 netlink_overrun(sk);
1957 /* Clone failed. Notify ALL listeners. */
1958 p->failure = 1;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001959 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1960 p->delivery_failure = 1;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001961 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1962 kfree_skb(p->skb2);
1963 p->skb2 = NULL;
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07001964 } else if (sk_filter(sk, p->skb2)) {
1965 kfree_skb(p->skb2);
1966 p->skb2 = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1968 netlink_overrun(sk);
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00001969 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1970 p->delivery_failure = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 } else {
1972 p->congested |= val;
1973 p->delivered = 1;
1974 p->skb2 = NULL;
1975 }
1976 sock_put(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977}
1978
Eric W. Biederman15e47302012-09-07 20:12:54 +00001979int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07001980 u32 group, gfp_t allocation,
1981 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1982 void *filter_data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001984 struct net *net = sock_net(ssk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 struct netlink_broadcast_data info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 struct sock *sk;
1987
1988 skb = netlink_trim(skb, allocation);
1989
1990 info.exclude_sk = ssk;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02001991 info.net = net;
Eric W. Biederman15e47302012-09-07 20:12:54 +00001992 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 info.group = group;
1994 info.failure = 0;
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08001995 info.delivery_failure = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 info.congested = 0;
1997 info.delivered = 0;
1998 info.allocation = allocation;
1999 info.skb = skb;
2000 info.skb2 = NULL;
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002001 info.tx_filter = filter;
2002 info.tx_data = filter_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003
2004 /* While we sleep in clone, do not allow to change socket list */
2005
2006 netlink_lock_table();
2007
Sasha Levinb67bfe02013-02-27 17:06:00 -08002008 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 do_one_broadcast(sk, &info);
2010
Neil Horman70d4bf62010-07-20 06:45:56 +00002011 consume_skb(skb);
Tommy S. Christensenaa1c6a62005-05-19 13:07:32 -07002012
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 netlink_unlock_table();
2014
Neil Horman70d4bf62010-07-20 06:45:56 +00002015 if (info.delivery_failure) {
2016 kfree_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08002017 return -ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00002018 }
2019 consume_skb(info.skb2);
Pablo Neira Ayusoff491a72009-02-05 23:56:36 -08002020
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 if (info.delivered) {
2022 if (info.congested && (allocation & __GFP_WAIT))
2023 yield();
2024 return 0;
2025 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 return -ESRCH;
2027}
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002028EXPORT_SYMBOL(netlink_broadcast_filtered);
2029
Eric W. Biederman15e47302012-09-07 20:12:54 +00002030int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002031 u32 group, gfp_t allocation)
2032{
Eric W. Biederman15e47302012-09-07 20:12:54 +00002033 return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
Eric W. Biederman910a7e92010-05-04 17:36:46 -07002034 NULL, NULL);
2035}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002036EXPORT_SYMBOL(netlink_broadcast);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037
2038struct netlink_set_err_data {
2039 struct sock *exclude_sk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002040 u32 portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 u32 group;
2042 int code;
2043};
2044
stephen hemmingerb57ef81f2011-12-22 08:52:02 +00002045static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046{
2047 struct netlink_sock *nlk = nlk_sk(sk);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002048 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
2050 if (sk == p->exclude_sk)
2051 goto out;
2052
Octavian Purdila09ad9bc2009-11-25 15:14:13 -08002053 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002054 goto out;
2055
Eric W. Biederman15e47302012-09-07 20:12:54 +00002056 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
Patrick McHardyf7fa9b12005-08-15 12:29:13 -07002057 !test_bit(p->group - 1, nlk->groups))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 goto out;
2059
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002060 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
2061 ret = 1;
2062 goto out;
2063 }
2064
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 sk->sk_err = p->code;
2066 sk->sk_error_report(sk);
2067out:
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002068 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069}
2070
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002071/**
2072 * netlink_set_err - report error to broadcast listeners
2073 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
Eric W. Biederman15e47302012-09-07 20:12:54 +00002074 * @portid: the PORTID of a process that we want to skip (if any)
Johannes Berg840e93f22013-11-19 10:35:40 +01002075 * @group: the broadcast group that will notice the error
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002076 * @code: error code, must be negative (as usual in kernelspace)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002077 *
2078 * This function returns the number of broadcast listeners that have set the
2079 * NETLINK_RECV_NO_ENOBUFS socket option.
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002080 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002081int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082{
2083 struct netlink_set_err_data info;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 struct sock *sk;
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002085 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086
2087 info.exclude_sk = ssk;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002088 info.portid = portid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 info.group = group;
Pablo Neira Ayuso4843b932009-03-03 23:37:30 -08002090 /* sk->sk_err wants a positive error value */
2091 info.code = -code;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
2093 read_lock(&nl_table_lock);
2094
Sasha Levinb67bfe02013-02-27 17:06:00 -08002095 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002096 ret += do_one_set_err(sk, &info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
2098 read_unlock(&nl_table_lock);
Pablo Neira Ayuso1a503072010-03-18 14:24:42 +00002099 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100}
Pablo Neira Ayusodd5b6ce2009-03-23 13:21:06 +01002101EXPORT_SYMBOL(netlink_set_err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102
Johannes Berg84659eb2007-07-18 15:47:05 -07002103/* must be called with netlink table grabbed */
2104static void netlink_update_socket_mc(struct netlink_sock *nlk,
2105 unsigned int group,
2106 int is_new)
2107{
2108 int old, new = !!is_new, subscriptions;
2109
2110 old = test_bit(group - 1, nlk->groups);
2111 subscriptions = nlk->subscriptions - old + new;
2112 if (new)
2113 __set_bit(group - 1, nlk->groups);
2114 else
2115 __clear_bit(group - 1, nlk->groups);
2116 netlink_update_subscriptions(&nlk->sk, subscriptions);
2117 netlink_update_listeners(&nlk->sk);
2118}
2119
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002120static int netlink_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002121 char __user *optval, unsigned int optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002122{
2123 struct sock *sk = sock->sk;
2124 struct netlink_sock *nlk = nlk_sk(sk);
Johannes Bergeb496532007-07-18 02:07:51 -07002125 unsigned int val = 0;
2126 int err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002127
2128 if (level != SOL_NETLINK)
2129 return -ENOPROTOOPT;
2130
Patrick McHardyccdfcc32013-04-17 06:47:01 +00002131 if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING &&
2132 optlen >= sizeof(int) &&
Johannes Bergeb496532007-07-18 02:07:51 -07002133 get_user(val, (unsigned int __user *)optval))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002134 return -EFAULT;
2135
2136 switch (optname) {
2137 case NETLINK_PKTINFO:
2138 if (val)
2139 nlk->flags |= NETLINK_RECV_PKTINFO;
2140 else
2141 nlk->flags &= ~NETLINK_RECV_PKTINFO;
2142 err = 0;
2143 break;
2144 case NETLINK_ADD_MEMBERSHIP:
2145 case NETLINK_DROP_MEMBERSHIP: {
Eric W. Biederman5187cd02014-04-23 14:25:48 -07002146 if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV))
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002147 return -EPERM;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002148 err = netlink_realloc_groups(sk);
2149 if (err)
2150 return err;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002151 if (!val || val - 1 >= nlk->ngroups)
2152 return -EINVAL;
Richard Guy Briggs7774d5e2014-04-22 21:31:55 -04002153 if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) {
Johannes Berg023e2cf2014-12-23 21:00:06 +01002154 err = nlk->netlink_bind(sock_net(sk), val);
Richard Guy Briggs4f520902014-04-22 21:31:54 -04002155 if (err)
2156 return err;
2157 }
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002158 netlink_table_grab();
Johannes Berg84659eb2007-07-18 15:47:05 -07002159 netlink_update_socket_mc(nlk, val,
2160 optname == NETLINK_ADD_MEMBERSHIP);
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002161 netlink_table_ungrab();
Richard Guy Briggs7774d5e2014-04-22 21:31:55 -04002162 if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind)
Johannes Berg023e2cf2014-12-23 21:00:06 +01002163 nlk->netlink_unbind(sock_net(sk), val);
Pablo Neira Ayuso03292742012-06-29 06:15:22 +00002164
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002165 err = 0;
2166 break;
2167 }
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00002168 case NETLINK_BROADCAST_ERROR:
2169 if (val)
2170 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
2171 else
2172 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
2173 err = 0;
2174 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002175 case NETLINK_NO_ENOBUFS:
2176 if (val) {
2177 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
Patrick McHardycd967e02013-04-17 06:46:56 +00002178 clear_bit(NETLINK_CONGESTED, &nlk->state);
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002179 wake_up_interruptible(&nlk->wait);
Eric Dumazet658cb352012-04-22 21:30:21 +00002180 } else {
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002181 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
Eric Dumazet658cb352012-04-22 21:30:21 +00002182 }
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002183 err = 0;
2184 break;
Patrick McHardyccdfcc32013-04-17 06:47:01 +00002185#ifdef CONFIG_NETLINK_MMAP
2186 case NETLINK_RX_RING:
2187 case NETLINK_TX_RING: {
2188 struct nl_mmap_req req;
2189
2190 /* Rings might consume more memory than queue limits, require
2191 * CAP_NET_ADMIN.
2192 */
2193 if (!capable(CAP_NET_ADMIN))
2194 return -EPERM;
2195 if (optlen < sizeof(req))
2196 return -EINVAL;
2197 if (copy_from_user(&req, optval, sizeof(req)))
2198 return -EFAULT;
2199 err = netlink_set_ring(sk, &req, false,
2200 optname == NETLINK_TX_RING);
2201 break;
2202 }
2203#endif /* CONFIG_NETLINK_MMAP */
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002204 default:
2205 err = -ENOPROTOOPT;
2206 }
2207 return err;
2208}
2209
2210static int netlink_getsockopt(struct socket *sock, int level, int optname,
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002211 char __user *optval, int __user *optlen)
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002212{
2213 struct sock *sk = sock->sk;
2214 struct netlink_sock *nlk = nlk_sk(sk);
2215 int len, val, err;
2216
2217 if (level != SOL_NETLINK)
2218 return -ENOPROTOOPT;
2219
2220 if (get_user(len, optlen))
2221 return -EFAULT;
2222 if (len < 0)
2223 return -EINVAL;
2224
2225 switch (optname) {
2226 case NETLINK_PKTINFO:
2227 if (len < sizeof(int))
2228 return -EINVAL;
2229 len = sizeof(int);
2230 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
Heiko Carstensa27b58f2006-10-30 15:06:12 -08002231 if (put_user(len, optlen) ||
2232 put_user(val, optval))
2233 return -EFAULT;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002234 err = 0;
2235 break;
Pablo Neira Ayusobe0c22a2009-02-18 01:40:43 +00002236 case NETLINK_BROADCAST_ERROR:
2237 if (len < sizeof(int))
2238 return -EINVAL;
2239 len = sizeof(int);
2240 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
2241 if (put_user(len, optlen) ||
2242 put_user(val, optval))
2243 return -EFAULT;
2244 err = 0;
2245 break;
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07002246 case NETLINK_NO_ENOBUFS:
2247 if (len < sizeof(int))
2248 return -EINVAL;
2249 len = sizeof(int);
2250 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
2251 if (put_user(len, optlen) ||
2252 put_user(val, optval))
2253 return -EFAULT;
2254 err = 0;
2255 break;
Patrick McHardy9a4595b2005-08-15 12:32:15 -07002256 default:
2257 err = -ENOPROTOOPT;
2258 }
2259 return err;
2260}
2261
2262static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
2263{
2264 struct nl_pktinfo info;
2265
2266 info.group = NETLINK_CB(skb).dst_group;
2267 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
2268}
2269
Ying Xue1b784142015-03-02 15:37:48 +08002270static int netlink_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002272 struct sock *sk = sock->sk;
2273 struct netlink_sock *nlk = nlk_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01002274 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002275 u32 dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002276 u32 dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277 struct sk_buff *skb;
2278 int err;
2279 struct scm_cookie scm;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002280 u32 netlink_skb_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
2282 if (msg->msg_flags&MSG_OOB)
2283 return -EOPNOTSUPP;
2284
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002285 err = scm_send(sock, msg, &scm, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 if (err < 0)
2287 return err;
2288
2289 if (msg->msg_namelen) {
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002290 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 if (addr->nl_family != AF_NETLINK)
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002292 goto out;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002293 dst_portid = addr->nl_pid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002294 dst_group = ffs(addr->nl_groups);
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002295 err = -EPERM;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002296 if ((dst_group || dst_portid) &&
Eric W. Biederman5187cd02014-04-23 14:25:48 -07002297 !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND))
Eric W. Biedermanb47030c2010-06-13 03:31:06 +00002298 goto out;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002299 netlink_skb_flags |= NETLINK_SKB_DST;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002300 } else {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002301 dst_portid = nlk->dst_portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002302 dst_group = nlk->dst_group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 }
2304
Eric W. Biederman15e47302012-09-07 20:12:54 +00002305 if (!nlk->portid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 err = netlink_autobind(sock);
2307 if (err)
2308 goto out;
2309 }
2310
Al Viroa8866ff2014-12-12 23:02:36 -05002311 /* It's a really convoluted way for userland to ask for mmaped
2312 * sendmsg(), but that's what we've got...
2313 */
Patrick McHardy5fd96122013-04-17 06:47:03 +00002314 if (netlink_tx_is_mmaped(sk) &&
Al Viroa8866ff2014-12-12 23:02:36 -05002315 msg->msg_iter.type == ITER_IOVEC &&
2316 msg->msg_iter.nr_segs == 1 &&
Al Viroc0371da2014-11-24 10:42:55 -05002317 msg->msg_iter.iov->iov_base == NULL) {
Patrick McHardy5fd96122013-04-17 06:47:03 +00002318 err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group,
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002319 &scm);
Patrick McHardy5fd96122013-04-17 06:47:03 +00002320 goto out;
2321 }
2322
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 err = -EMSGSIZE;
2324 if (len > sk->sk_sndbuf - 32)
2325 goto out;
2326 err = -ENOBUFS;
Pablo Neira3a365152013-06-28 03:04:23 +02002327 skb = netlink_alloc_large_skb(len, dst_group);
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002328 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 goto out;
2330
Eric W. Biederman15e47302012-09-07 20:12:54 +00002331 NETLINK_CB(skb).portid = nlk->portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002332 NETLINK_CB(skb).dst_group = dst_group;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002333 NETLINK_CB(skb).creds = scm.creds;
Eric W. Biederman2d7a85f2014-05-30 11:04:00 -07002334 NETLINK_CB(skb).flags = netlink_skb_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336 err = -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002337 if (memcpy_from_msg(skb_put(skb, len), msg, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 kfree_skb(skb);
2339 goto out;
2340 }
2341
2342 err = security_netlink_send(sk, skb);
2343 if (err) {
2344 kfree_skb(skb);
2345 goto out;
2346 }
2347
Patrick McHardyd629b832005-08-14 19:27:50 -07002348 if (dst_group) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002349 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002350 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 }
Eric W. Biederman15e47302012-09-07 20:12:54 +00002352 err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353
2354out:
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002355 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356 return err;
2357}
2358
Ying Xue1b784142015-03-02 15:37:48 +08002359static int netlink_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 int flags)
2361{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362 struct scm_cookie scm;
2363 struct sock *sk = sock->sk;
2364 struct netlink_sock *nlk = nlk_sk(sk);
2365 int noblock = flags&MSG_DONTWAIT;
2366 size_t copied;
Johannes Berg68d6ac62010-08-15 21:20:44 +00002367 struct sk_buff *skb, *data_skb;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002368 int err, ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369
2370 if (flags&MSG_OOB)
2371 return -EOPNOTSUPP;
2372
2373 copied = 0;
2374
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002375 skb = skb_recv_datagram(sk, flags, noblock, &err);
2376 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 goto out;
2378
Johannes Berg68d6ac62010-08-15 21:20:44 +00002379 data_skb = skb;
2380
Johannes Berg1dacc762009-07-01 11:26:02 +00002381#ifdef CONFIG_COMPAT_NETLINK_MESSAGES
2382 if (unlikely(skb_shinfo(skb)->frag_list)) {
Johannes Berg1dacc762009-07-01 11:26:02 +00002383 /*
Johannes Berg68d6ac62010-08-15 21:20:44 +00002384 * If this skb has a frag_list, then here that means that we
2385 * will have to use the frag_list skb's data for compat tasks
2386 * and the regular skb's data for normal (non-compat) tasks.
Johannes Berg1dacc762009-07-01 11:26:02 +00002387 *
Johannes Berg68d6ac62010-08-15 21:20:44 +00002388 * If we need to send the compat skb, assign it to the
2389 * 'data_skb' variable so that it will be used below for data
2390 * copying. We keep 'skb' for everything else, including
2391 * freeing both later.
Johannes Berg1dacc762009-07-01 11:26:02 +00002392 */
Johannes Berg68d6ac62010-08-15 21:20:44 +00002393 if (flags & MSG_CMSG_COMPAT)
2394 data_skb = skb_shinfo(skb)->frag_list;
Johannes Berg1dacc762009-07-01 11:26:02 +00002395 }
2396#endif
2397
Eric Dumazet9063e212014-03-07 12:02:33 -08002398 /* Record the max length of recvmsg() calls for future allocations */
2399 nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len);
2400 nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len,
2401 16384);
2402
Johannes Berg68d6ac62010-08-15 21:20:44 +00002403 copied = data_skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 if (len < copied) {
2405 msg->msg_flags |= MSG_TRUNC;
2406 copied = len;
2407 }
2408
Johannes Berg68d6ac62010-08-15 21:20:44 +00002409 skb_reset_transport_header(data_skb);
David S. Miller51f3d022014-11-05 16:46:40 -05002410 err = skb_copy_datagram_msg(data_skb, 0, msg, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411
2412 if (msg->msg_name) {
Steffen Hurrle342dfc32014-01-17 22:53:15 +01002413 DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414 addr->nl_family = AF_NETLINK;
2415 addr->nl_pad = 0;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002416 addr->nl_pid = NETLINK_CB(skb).portid;
Patrick McHardyd629b832005-08-14 19:27:50 -07002417 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418 msg->msg_namelen = sizeof(*addr);
2419 }
2420
Patrick McHardycc9a06c2006-03-12 20:34:27 -08002421 if (nlk->flags & NETLINK_RECV_PKTINFO)
2422 netlink_cmsg_recv_pktinfo(msg, skb);
2423
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002424 memset(&scm, 0, sizeof(scm));
2425 scm.creds = *NETLINK_CREDS(skb);
Patrick McHardy188ccb52007-05-03 03:27:01 -07002426 if (flags & MSG_TRUNC)
Johannes Berg68d6ac62010-08-15 21:20:44 +00002427 copied = data_skb->len;
David S. Millerdaa37662010-08-15 23:21:50 -07002428
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429 skb_free_datagram(sk, skb);
2430
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002431 if (nlk->cb_running &&
2432 atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
Andrey Vaginb44d2112011-02-21 02:40:47 +00002433 ret = netlink_dump(sk);
2434 if (ret) {
Ben Pfaffac30ef82014-07-09 10:31:22 -07002435 sk->sk_err = -ret;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002436 sk->sk_error_report(sk);
2437 }
2438 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002440 scm_recv(sock, msg, &scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441out:
2442 netlink_rcv_wake(sk);
2443 return err ? : copied;
2444}
2445
David S. Miller676d2362014-04-11 16:15:36 -04002446static void netlink_data_ready(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447{
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002448 BUG();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449}
2450
2451/*
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09002452 * We export these functions to other modules. They provide a
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 * complete set of kernel non-blocking support for message
2454 * queueing.
2455 */
2456
2457struct sock *
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00002458__netlink_kernel_create(struct net *net, int unit, struct module *module,
2459 struct netlink_kernel_cfg *cfg)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460{
2461 struct socket *sock;
2462 struct sock *sk;
Patrick McHardy77247bb2005-08-14 19:27:13 -07002463 struct netlink_sock *nlk;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002464 struct listeners *listeners = NULL;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002465 struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
2466 unsigned int groups;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467
Akinobu Mitafab2caf2006-08-29 02:15:24 -07002468 BUG_ON(!nl_table);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002470 if (unit < 0 || unit >= MAX_LINKS)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471 return NULL;
2472
2473 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
2474 return NULL;
2475
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08002476 /*
2477 * We have to just have a reference on the net from sk, but don't
2478 * get_net it. Besides, we cannot get and then put the net here.
2479 * So we create one inside init_net and the move it to net.
2480 */
2481
2482 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
2483 goto out_sock_release_nosk;
2484
2485 sk = sock->sk;
Denis V. Lunevedf02082008-02-29 11:18:32 -08002486 sk_change_net(sk, net);
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002487
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002488 if (!cfg || cfg->groups < 32)
Patrick McHardy4277a082006-03-20 18:52:01 -08002489 groups = 32;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002490 else
2491 groups = cfg->groups;
Patrick McHardy4277a082006-03-20 18:52:01 -08002492
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002493 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
Patrick McHardy4277a082006-03-20 18:52:01 -08002494 if (!listeners)
2495 goto out_sock_release;
2496
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497 sk->sk_data_ready = netlink_data_ready;
Pablo Neira Ayusoa31f2d12012-06-29 06:15:21 +00002498 if (cfg && cfg->input)
2499 nlk_sk(sk)->netlink_rcv = cfg->input;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002500
Herbert Xu8ea65f42015-01-26 14:02:56 +11002501 if (netlink_insert(sk, 0))
Patrick McHardy77247bb2005-08-14 19:27:13 -07002502 goto out_sock_release;
2503
2504 nlk = nlk_sk(sk);
2505 nlk->flags |= NETLINK_KERNEL_SOCKET;
2506
2507 netlink_table_grab();
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002508 if (!nl_table[unit].registered) {
2509 nl_table[unit].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002510 rcu_assign_pointer(nl_table[unit].listeners, listeners);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002511 nl_table[unit].cb_mutex = cb_mutex;
2512 nl_table[unit].module = module;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002513 if (cfg) {
2514 nl_table[unit].bind = cfg->bind;
Hiroaki SHIMODA6251edd2014-11-13 04:24:10 +09002515 nl_table[unit].unbind = cfg->unbind;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002516 nl_table[unit].flags = cfg->flags;
Gao fengda12c902013-06-06 14:49:11 +08002517 if (cfg->compare)
2518 nl_table[unit].compare = cfg->compare;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00002519 }
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002520 nl_table[unit].registered = 1;
Jesper Juhlf937f1f462007-10-15 01:39:12 -07002521 } else {
2522 kfree(listeners);
Denis V. Lunev869e58f2008-01-18 23:53:31 -08002523 nl_table[unit].registered++;
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002524 }
Patrick McHardy77247bb2005-08-14 19:27:13 -07002525 netlink_table_ungrab();
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002526 return sk;
2527
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002528out_sock_release:
Patrick McHardy4277a082006-03-20 18:52:01 -08002529 kfree(listeners);
Denis V. Lunev9dfbec12008-02-29 11:17:56 -08002530 netlink_kernel_release(sk);
Pavel Emelyanov23fe1862008-01-30 19:31:06 -08002531 return NULL;
2532
2533out_sock_release_nosk:
Harald Welte4fdb3bb2005-08-09 19:40:55 -07002534 sock_release(sock);
Patrick McHardy77247bb2005-08-14 19:27:13 -07002535 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536}
Pablo Neira Ayuso9f00d972012-09-08 02:53:54 +00002537EXPORT_SYMBOL(__netlink_kernel_create);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08002538
2539void
2540netlink_kernel_release(struct sock *sk)
2541{
Denis V. Lunevedf02082008-02-29 11:18:32 -08002542 sk_release_kernel(sk);
Denis V. Lunevb7c6ba62008-01-28 14:41:19 -08002543}
2544EXPORT_SYMBOL(netlink_kernel_release);
2545
Johannes Bergd136f1b2009-09-12 03:03:15 +00002546int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002547{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002548 struct listeners *new, *old;
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002549 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002550
2551 if (groups < 32)
2552 groups = 32;
2553
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002554 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002555 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
2556 if (!new)
Johannes Bergd136f1b2009-09-12 03:03:15 +00002557 return -ENOMEM;
Eric Dumazet6d772ac2012-10-18 03:21:55 +00002558 old = nl_deref_protected(tbl->listeners);
Eric Dumazet5c398dc2010-10-24 04:27:10 +00002559 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
2560 rcu_assign_pointer(tbl->listeners, new);
2561
Lai Jiangshan37b6b932011-03-15 18:01:42 +08002562 kfree_rcu(old, rcu);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002563 }
2564 tbl->groups = groups;
2565
Johannes Bergd136f1b2009-09-12 03:03:15 +00002566 return 0;
2567}
2568
2569/**
2570 * netlink_change_ngroups - change number of multicast groups
2571 *
2572 * This changes the number of multicast groups that are available
2573 * on a certain netlink family. Note that it is not possible to
2574 * change the number of groups to below 32. Also note that it does
2575 * not implicitly call netlink_clear_multicast_users() when the
2576 * number of groups is reduced.
2577 *
2578 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
2579 * @groups: The new number of groups.
2580 */
2581int netlink_change_ngroups(struct sock *sk, unsigned int groups)
2582{
2583 int err;
2584
2585 netlink_table_grab();
2586 err = __netlink_change_ngroups(sk, groups);
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002587 netlink_table_ungrab();
Johannes Bergd136f1b2009-09-12 03:03:15 +00002588
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002589 return err;
2590}
Johannes Bergb4ff4f02007-07-18 15:46:06 -07002591
Johannes Bergb8273572009-09-24 15:44:05 -07002592void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
2593{
2594 struct sock *sk;
Johannes Bergb8273572009-09-24 15:44:05 -07002595 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
2596
Sasha Levinb67bfe02013-02-27 17:06:00 -08002597 sk_for_each_bound(sk, &tbl->mc_list)
Johannes Bergb8273572009-09-24 15:44:05 -07002598 netlink_update_socket_mc(nlk_sk(sk), group, 0);
2599}
2600
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002601struct nlmsghdr *
Eric W. Biederman15e47302012-09-07 20:12:54 +00002602__nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002603{
2604 struct nlmsghdr *nlh;
Hong zhi guo573ce262013-03-27 06:47:04 +00002605 int size = nlmsg_msg_size(len);
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002606
Wang Yufen23b45672014-02-17 16:53:32 +08002607 nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size));
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002608 nlh->nlmsg_type = type;
2609 nlh->nlmsg_len = size;
2610 nlh->nlmsg_flags = flags;
Eric W. Biederman15e47302012-09-07 20:12:54 +00002611 nlh->nlmsg_pid = portid;
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002612 nlh->nlmsg_seq = seq;
2613 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
Hong zhi guo573ce262013-03-27 06:47:04 +00002614 memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
Denys Vlasenkoa46621a2012-01-30 15:22:06 -05002615 return nlh;
2616}
2617EXPORT_SYMBOL(__nlmsg_put);
2618
Linus Torvalds1da177e2005-04-16 15:20:36 -07002619/*
2620 * It looks a bit ugly.
2621 * It would be better to create kernel thread.
2622 */
2623
2624static int netlink_dump(struct sock *sk)
2625{
2626 struct netlink_sock *nlk = nlk_sk(sk);
2627 struct netlink_callback *cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00002628 struct sk_buff *skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 struct nlmsghdr *nlh;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002630 int len, err = -ENOBUFS;
Greg Rosec7ac8672011-06-10 01:27:09 +00002631 int alloc_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002633 mutex_lock(nlk->cb_mutex);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002634 if (!nlk->cb_running) {
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002635 err = -EINVAL;
2636 goto errout_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637 }
2638
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002639 cb = &nlk->cb;
Greg Rosec7ac8672011-06-10 01:27:09 +00002640 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
2641
Patrick McHardyf9c22882013-04-17 06:47:04 +00002642 if (!netlink_rx_is_mmaped(sk) &&
2643 atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf)
2644 goto errout_skb;
Eric Dumazet9063e212014-03-07 12:02:33 -08002645
2646 /* NLMSG_GOODSIZE is small to avoid high order allocations being
2647 * required, but it makes sense to _attempt_ a 16K bytes allocation
2648 * to reduce number of system calls on dump operations, if user
2649 * ever provided a big enough buffer.
2650 */
2651 if (alloc_size < nlk->max_recvmsg_len) {
2652 skb = netlink_alloc_skb(sk,
2653 nlk->max_recvmsg_len,
2654 nlk->portid,
2655 GFP_KERNEL |
2656 __GFP_NOWARN |
2657 __GFP_NORETRY);
2658 /* available room should be exact amount to avoid MSG_TRUNC */
2659 if (skb)
2660 skb_reserve(skb, skb_tailroom(skb) -
2661 nlk->max_recvmsg_len);
2662 }
2663 if (!skb)
2664 skb = netlink_alloc_skb(sk, alloc_size, nlk->portid,
2665 GFP_KERNEL);
Greg Rosec7ac8672011-06-10 01:27:09 +00002666 if (!skb)
Dan Carpenterc63d6ea2011-06-15 03:11:42 +00002667 goto errout_skb;
Patrick McHardyf9c22882013-04-17 06:47:04 +00002668 netlink_skb_set_owner_r(skb, sk);
Greg Rosec7ac8672011-06-10 01:27:09 +00002669
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670 len = cb->dump(skb, cb);
2671
2672 if (len > 0) {
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002673 mutex_unlock(nlk->cb_mutex);
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07002674
2675 if (sk_filter(sk, skb))
2676 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00002677 else
2678 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002679 return 0;
2680 }
2681
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002682 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
2683 if (!nlh)
2684 goto errout_skb;
2685
Johannes Berg670dc282011-06-20 13:40:46 +02002686 nl_dump_check_consistent(cb, nlh);
2687
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002688 memcpy(nlmsg_data(nlh), &len, sizeof(len));
2689
Stephen Hemmingerb1153f22008-03-21 15:46:12 -07002690 if (sk_filter(sk, skb))
2691 kfree_skb(skb);
Eric Dumazet4a7e7c22012-04-05 22:17:46 +00002692 else
2693 __netlink_sendskb(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002694
Thomas Grafa8f74b22005-11-10 02:25:52 +01002695 if (cb->done)
2696 cb->done(cb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002698 nlk->cb_running = false;
2699 mutex_unlock(nlk->cb_mutex);
Gao feng6dc878a2012-10-04 20:15:48 +00002700 module_put(cb->module);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002701 consume_skb(cb->skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002702 return 0;
Thomas Graf17977542005-06-18 22:53:48 -07002703
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002704errout_skb:
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002705 mutex_unlock(nlk->cb_mutex);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002706 kfree_skb(skb);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002707 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708}
2709
Gao feng6dc878a2012-10-04 20:15:48 +00002710int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
2711 const struct nlmsghdr *nlh,
2712 struct netlink_dump_control *control)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713{
2714 struct netlink_callback *cb;
2715 struct sock *sk;
2716 struct netlink_sock *nlk;
Andrey Vaginb44d2112011-02-21 02:40:47 +00002717 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002718
Patrick McHardyf9c22882013-04-17 06:47:04 +00002719 /* Memory mapped dump requests need to be copied to avoid looping
2720 * on the pending state in netlink_mmap_sendmsg() while the CB hold
2721 * a reference to the skb.
2722 */
2723 if (netlink_skb_is_mmaped(skb)) {
2724 skb = skb_copy(skb, GFP_KERNEL);
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002725 if (skb == NULL)
Patrick McHardyf9c22882013-04-17 06:47:04 +00002726 return -ENOBUFS;
Patrick McHardyf9c22882013-04-17 06:47:04 +00002727 } else
2728 atomic_inc(&skb->users);
2729
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002730 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
2731 if (sk == NULL) {
2732 ret = -ECONNREFUSED;
2733 goto error_free;
2734 }
2735
2736 nlk = nlk_sk(sk);
2737 mutex_lock(nlk->cb_mutex);
2738 /* A dump is in progress... */
2739 if (nlk->cb_running) {
2740 ret = -EBUSY;
2741 goto error_unlock;
2742 }
2743 /* add reference of module which cb->dump belongs to */
2744 if (!try_module_get(control->module)) {
2745 ret = -EPROTONOSUPPORT;
2746 goto error_unlock;
2747 }
2748
2749 cb = &nlk->cb;
2750 memset(cb, 0, sizeof(*cb));
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00002751 cb->dump = control->dump;
2752 cb->done = control->done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002753 cb->nlh = nlh;
Pablo Neira Ayuso7175c882012-02-24 14:30:16 +00002754 cb->data = control->data;
Gao feng6dc878a2012-10-04 20:15:48 +00002755 cb->module = control->module;
Pablo Neira Ayuso80d326f2012-02-24 14:30:15 +00002756 cb->min_dump_alloc = control->min_dump_alloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002757 cb->skb = skb;
2758
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002759 nlk->cb_running = true;
Gao feng6dc878a2012-10-04 20:15:48 +00002760
Patrick McHardyaf65bdf2007-04-20 14:14:21 -07002761 mutex_unlock(nlk->cb_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762
Andrey Vaginb44d2112011-02-21 02:40:47 +00002763 ret = netlink_dump(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764 sock_put(sk);
Denis V. Lunev5c582982007-10-23 20:29:25 -07002765
Andrey Vaginb44d2112011-02-21 02:40:47 +00002766 if (ret)
2767 return ret;
2768
Denis V. Lunev5c582982007-10-23 20:29:25 -07002769 /* We successfully started a dump, by returning -EINTR we
2770 * signal not to send ACK even if it was requested.
2771 */
2772 return -EINTR;
Pravin B Shelar16b304f2013-08-15 15:31:06 -07002773
2774error_unlock:
2775 sock_put(sk);
2776 mutex_unlock(nlk->cb_mutex);
2777error_free:
2778 kfree_skb(skb);
2779 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780}
Gao feng6dc878a2012-10-04 20:15:48 +00002781EXPORT_SYMBOL(__netlink_dump_start);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002782
2783void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
2784{
2785 struct sk_buff *skb;
2786 struct nlmsghdr *rep;
2787 struct nlmsgerr *errmsg;
Thomas Graf339bf982006-11-10 14:10:15 -08002788 size_t payload = sizeof(*errmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789
Thomas Graf339bf982006-11-10 14:10:15 -08002790 /* error messages get the original request appened */
2791 if (err)
2792 payload += nlmsg_len(nlh);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002793
Patrick McHardyf9c22882013-04-17 06:47:04 +00002794 skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload),
2795 NETLINK_CB(in_skb).portid, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796 if (!skb) {
2797 struct sock *sk;
2798
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002799 sk = netlink_lookup(sock_net(in_skb->sk),
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02002800 in_skb->sk->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00002801 NETLINK_CB(in_skb).portid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802 if (sk) {
2803 sk->sk_err = ENOBUFS;
2804 sk->sk_error_report(sk);
2805 sock_put(sk);
2806 }
2807 return;
2808 }
2809
Eric W. Biederman15e47302012-09-07 20:12:54 +00002810 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
John Fastabend5dba93a2009-09-25 13:11:44 +00002811 NLMSG_ERROR, payload, 0);
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002812 errmsg = nlmsg_data(rep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813 errmsg->error = err;
Thomas Grafbf8b79e2006-08-04 23:03:29 -07002814 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
Eric W. Biederman15e47302012-09-07 20:12:54 +00002815 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002816}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002817EXPORT_SYMBOL(netlink_ack);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002819int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
Thomas Graf1d00a4e2007-03-22 23:30:12 -07002820 struct nlmsghdr *))
Thomas Graf82ace472005-11-10 02:25:53 +01002821{
Thomas Graf82ace472005-11-10 02:25:53 +01002822 struct nlmsghdr *nlh;
2823 int err;
2824
2825 while (skb->len >= nlmsg_total_size(0)) {
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002826 int msglen;
2827
Arnaldo Carvalho de Melob529ccf2007-04-25 19:08:35 -07002828 nlh = nlmsg_hdr(skb);
Thomas Grafd35b6852007-03-22 23:28:46 -07002829 err = 0;
Thomas Graf82ace472005-11-10 02:25:53 +01002830
Martin Murrayad8e4b72006-01-10 13:02:29 -08002831 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
Thomas Graf82ace472005-11-10 02:25:53 +01002832 return 0;
2833
Thomas Grafd35b6852007-03-22 23:28:46 -07002834 /* Only requests are handled by the kernel */
2835 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
Denis V. Lunev5c582982007-10-23 20:29:25 -07002836 goto ack;
Thomas Grafd35b6852007-03-22 23:28:46 -07002837
Thomas Graf45e7ae72007-03-22 23:29:10 -07002838 /* Skip control messages */
2839 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
Denis V. Lunev5c582982007-10-23 20:29:25 -07002840 goto ack;
Thomas Graf45e7ae72007-03-22 23:29:10 -07002841
Thomas Graf1d00a4e2007-03-22 23:30:12 -07002842 err = cb(skb, nlh);
Denis V. Lunev5c582982007-10-23 20:29:25 -07002843 if (err == -EINTR)
2844 goto skip;
2845
2846ack:
Thomas Grafd35b6852007-03-22 23:28:46 -07002847 if (nlh->nlmsg_flags & NLM_F_ACK || err)
Thomas Graf82ace472005-11-10 02:25:53 +01002848 netlink_ack(skb, nlh, err);
Thomas Graf82ace472005-11-10 02:25:53 +01002849
Denis V. Lunev5c582982007-10-23 20:29:25 -07002850skip:
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002851 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
Denis V. Lunevcd40b7d2007-10-10 21:15:29 -07002852 if (msglen > skb->len)
2853 msglen = skb->len;
2854 skb_pull(skb, msglen);
Thomas Graf82ace472005-11-10 02:25:53 +01002855 }
2856
2857 return 0;
2858}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002859EXPORT_SYMBOL(netlink_rcv_skb);
Thomas Graf82ace472005-11-10 02:25:53 +01002860
2861/**
Thomas Grafd387f6a2006-08-15 00:31:06 -07002862 * nlmsg_notify - send a notification netlink message
2863 * @sk: netlink socket to use
2864 * @skb: notification message
Eric W. Biederman15e47302012-09-07 20:12:54 +00002865 * @portid: destination netlink portid for reports or 0
Thomas Grafd387f6a2006-08-15 00:31:06 -07002866 * @group: destination multicast group or 0
2867 * @report: 1 to report back, 0 to disable
2868 * @flags: allocation flags
2869 */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002870int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
Thomas Grafd387f6a2006-08-15 00:31:06 -07002871 unsigned int group, int report, gfp_t flags)
2872{
2873 int err = 0;
2874
2875 if (group) {
Eric W. Biederman15e47302012-09-07 20:12:54 +00002876 int exclude_portid = 0;
Thomas Grafd387f6a2006-08-15 00:31:06 -07002877
2878 if (report) {
2879 atomic_inc(&skb->users);
Eric W. Biederman15e47302012-09-07 20:12:54 +00002880 exclude_portid = portid;
Thomas Grafd387f6a2006-08-15 00:31:06 -07002881 }
2882
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002883 /* errors reported via destination sk->sk_err, but propagate
2884 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
Eric W. Biederman15e47302012-09-07 20:12:54 +00002885 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
Thomas Grafd387f6a2006-08-15 00:31:06 -07002886 }
2887
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002888 if (report) {
2889 int err2;
2890
Eric W. Biederman15e47302012-09-07 20:12:54 +00002891 err2 = nlmsg_unicast(sk, skb, portid);
Pablo Neira Ayuso1ce85fe2009-02-24 23:18:28 -08002892 if (!err || err == -ESRCH)
2893 err = err2;
2894 }
Thomas Grafd387f6a2006-08-15 00:31:06 -07002895
2896 return err;
2897}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08002898EXPORT_SYMBOL(nlmsg_notify);
Thomas Grafd387f6a2006-08-15 00:31:06 -07002899
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900#ifdef CONFIG_PROC_FS
2901struct nl_seq_iter {
Denis V. Luneve372c412007-11-19 22:31:54 -08002902 struct seq_net_private p;
Herbert Xu56d28b12015-02-04 07:33:24 +11002903 struct rhashtable_iter hti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904 int link;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905};
2906
Herbert Xu56d28b12015-02-04 07:33:24 +11002907static int netlink_walk_start(struct nl_seq_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908{
Herbert Xu56d28b12015-02-04 07:33:24 +11002909 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910
Herbert Xu56d28b12015-02-04 07:33:24 +11002911 err = rhashtable_walk_init(&nl_table[iter->link].hash, &iter->hti);
2912 if (err) {
2913 iter->link = MAX_LINKS;
2914 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915 }
Herbert Xu56d28b12015-02-04 07:33:24 +11002916
2917 err = rhashtable_walk_start(&iter->hti);
2918 return err == -EAGAIN ? 0 : err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002919}
2920
Herbert Xu56d28b12015-02-04 07:33:24 +11002921static void netlink_walk_stop(struct nl_seq_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002922{
Herbert Xu56d28b12015-02-04 07:33:24 +11002923 rhashtable_walk_stop(&iter->hti);
2924 rhashtable_walk_exit(&iter->hti);
2925}
2926
2927static void *__netlink_seq_next(struct seq_file *seq)
2928{
2929 struct nl_seq_iter *iter = seq->private;
2930 struct netlink_sock *nlk;
2931
2932 do {
2933 for (;;) {
2934 int err;
2935
2936 nlk = rhashtable_walk_next(&iter->hti);
2937
2938 if (IS_ERR(nlk)) {
2939 if (PTR_ERR(nlk) == -EAGAIN)
2940 continue;
2941
2942 return nlk;
2943 }
2944
2945 if (nlk)
2946 break;
2947
2948 netlink_walk_stop(iter);
2949 if (++iter->link >= MAX_LINKS)
2950 return NULL;
2951
2952 err = netlink_walk_start(iter);
2953 if (err)
2954 return ERR_PTR(err);
2955 }
2956 } while (sock_net(&nlk->sk) != seq_file_net(seq));
2957
2958 return nlk;
2959}
2960
2961static void *netlink_seq_start(struct seq_file *seq, loff_t *posp)
2962{
2963 struct nl_seq_iter *iter = seq->private;
2964 void *obj = SEQ_START_TOKEN;
2965 loff_t pos;
2966 int err;
2967
2968 iter->link = 0;
2969
2970 err = netlink_walk_start(iter);
2971 if (err)
2972 return ERR_PTR(err);
2973
2974 for (pos = *posp; pos && obj && !IS_ERR(obj); pos--)
2975 obj = __netlink_seq_next(seq);
2976
2977 return obj;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002978}
2979
2980static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2981{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002982 ++*pos;
Herbert Xu56d28b12015-02-04 07:33:24 +11002983 return __netlink_seq_next(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002984}
2985
2986static void netlink_seq_stop(struct seq_file *seq, void *v)
2987{
Herbert Xu56d28b12015-02-04 07:33:24 +11002988 struct nl_seq_iter *iter = seq->private;
2989
2990 if (iter->link >= MAX_LINKS)
2991 return;
2992
2993 netlink_walk_stop(iter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002994}
2995
2996
2997static int netlink_seq_show(struct seq_file *seq, void *v)
2998{
Eric Dumazet658cb352012-04-22 21:30:21 +00002999 if (v == SEQ_START_TOKEN) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000 seq_puts(seq,
3001 "sk Eth Pid Groups "
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00003002 "Rmem Wmem Dump Locks Drops Inode\n");
Eric Dumazet658cb352012-04-22 21:30:21 +00003003 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003004 struct sock *s = v;
3005 struct netlink_sock *nlk = nlk_sk(s);
3006
Pravin B Shelar16b304f2013-08-15 15:31:06 -07003007 seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07003008 s,
3009 s->sk_protocol,
Eric W. Biederman15e47302012-09-07 20:12:54 +00003010 nlk->portid,
Patrick McHardy513c2502005-09-06 15:43:59 -07003011 nlk->groups ? (u32)nlk->groups[0] : 0,
Eric Dumazet31e6d362009-06-17 19:05:41 -07003012 sk_rmem_alloc_get(s),
3013 sk_wmem_alloc_get(s),
Pravin B Shelar16b304f2013-08-15 15:31:06 -07003014 nlk->cb_running,
Pablo Neira Ayuso38938bf2009-03-24 16:37:55 -07003015 atomic_read(&s->sk_refcnt),
Masatake YAMATOcf0aa4e2010-02-27 19:45:37 +00003016 atomic_read(&s->sk_drops),
3017 sock_i_ino(s)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003018 );
3019
3020 }
3021 return 0;
3022}
3023
Philippe De Muyter56b3d972007-07-10 23:07:31 -07003024static const struct seq_operations netlink_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003025 .start = netlink_seq_start,
3026 .next = netlink_seq_next,
3027 .stop = netlink_seq_stop,
3028 .show = netlink_seq_show,
3029};
3030
3031
3032static int netlink_seq_open(struct inode *inode, struct file *file)
3033{
Denis V. Luneve372c412007-11-19 22:31:54 -08003034 return seq_open_net(inode, file, &netlink_seq_ops,
3035 sizeof(struct nl_seq_iter));
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003036}
3037
Arjan van de Venda7071d2007-02-12 00:55:36 -08003038static const struct file_operations netlink_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003039 .owner = THIS_MODULE,
3040 .open = netlink_seq_open,
3041 .read = seq_read,
3042 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08003043 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044};
3045
3046#endif
3047
3048int netlink_register_notifier(struct notifier_block *nb)
3049{
Alan Sterne041c682006-03-27 01:16:30 -08003050 return atomic_notifier_chain_register(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08003052EXPORT_SYMBOL(netlink_register_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003053
3054int netlink_unregister_notifier(struct notifier_block *nb)
3055{
Alan Sterne041c682006-03-27 01:16:30 -08003056 return atomic_notifier_chain_unregister(&netlink_chain, nb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057}
Patrick McHardy6ac552f2007-12-04 00:19:38 -08003058EXPORT_SYMBOL(netlink_unregister_notifier);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09003059
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08003060static const struct proto_ops netlink_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003061 .family = PF_NETLINK,
3062 .owner = THIS_MODULE,
3063 .release = netlink_release,
3064 .bind = netlink_bind,
3065 .connect = netlink_connect,
3066 .socketpair = sock_no_socketpair,
3067 .accept = sock_no_accept,
3068 .getname = netlink_getname,
Patrick McHardy9652e932013-04-17 06:47:02 +00003069 .poll = netlink_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070 .ioctl = sock_no_ioctl,
3071 .listen = sock_no_listen,
3072 .shutdown = sock_no_shutdown,
Patrick McHardy9a4595b2005-08-15 12:32:15 -07003073 .setsockopt = netlink_setsockopt,
3074 .getsockopt = netlink_getsockopt,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003075 .sendmsg = netlink_sendmsg,
3076 .recvmsg = netlink_recvmsg,
Patrick McHardyccdfcc32013-04-17 06:47:01 +00003077 .mmap = netlink_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -07003078 .sendpage = sock_no_sendpage,
3079};
3080
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00003081static const struct net_proto_family netlink_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003082 .family = PF_NETLINK,
3083 .create = netlink_create,
3084 .owner = THIS_MODULE, /* for consistency 8) */
3085};
3086
Pavel Emelyanov46650792007-10-08 20:38:39 -07003087static int __net_init netlink_net_init(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003088{
3089#ifdef CONFIG_PROC_FS
Gao fengd4beaa62013-02-18 01:34:54 +00003090 if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops))
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003091 return -ENOMEM;
3092#endif
3093 return 0;
3094}
3095
Pavel Emelyanov46650792007-10-08 20:38:39 -07003096static void __net_exit netlink_net_exit(struct net *net)
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003097{
3098#ifdef CONFIG_PROC_FS
Gao fengece31ff2013-02-18 01:34:56 +00003099 remove_proc_entry("netlink", net->proc_net);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003100#endif
3101}
3102
David S. Millerb963ea82010-08-30 19:08:01 -07003103static void __init netlink_add_usersock_entry(void)
3104{
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003105 struct listeners *listeners;
David S. Millerb963ea82010-08-30 19:08:01 -07003106 int groups = 32;
3107
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003108 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
David S. Millerb963ea82010-08-30 19:08:01 -07003109 if (!listeners)
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003110 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
David S. Millerb963ea82010-08-30 19:08:01 -07003111
3112 netlink_table_grab();
3113
3114 nl_table[NETLINK_USERSOCK].groups = groups;
Eric Dumazet5c398dc2010-10-24 04:27:10 +00003115 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
David S. Millerb963ea82010-08-30 19:08:01 -07003116 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
3117 nl_table[NETLINK_USERSOCK].registered = 1;
Pablo Neira Ayuso9785e102012-09-08 02:53:53 +00003118 nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
David S. Millerb963ea82010-08-30 19:08:01 -07003119
3120 netlink_table_ungrab();
3121}
3122
Denis V. Lunev022cbae2007-11-13 03:23:50 -08003123static struct pernet_operations __net_initdata netlink_net_ops = {
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003124 .init = netlink_net_init,
3125 .exit = netlink_net_exit,
3126};
3127
Patrick McHardy49f7b332015-03-25 13:07:45 +00003128static inline u32 netlink_hash(const void *data, u32 len, u32 seed)
Herbert Xuc428ecd2015-03-20 21:57:01 +11003129{
3130 const struct netlink_sock *nlk = data;
3131 struct netlink_compare_arg arg;
3132
3133 netlink_compare_arg_init(&arg, sock_net(&nlk->sk), nlk->portid);
Herbert Xu11b58ba2015-03-24 00:50:22 +11003134 return jhash2((u32 *)&arg, netlink_compare_arg_len / sizeof(u32), seed);
Herbert Xuc428ecd2015-03-20 21:57:01 +11003135}
3136
3137static const struct rhashtable_params netlink_rhashtable_params = {
3138 .head_offset = offsetof(struct netlink_sock, node),
3139 .key_len = netlink_compare_arg_len,
Herbert Xuc428ecd2015-03-20 21:57:01 +11003140 .obj_hashfn = netlink_hash,
3141 .obj_cmpfn = netlink_compare,
3142 .max_size = 65536,
Thomas Grafb5e2c152015-03-24 20:42:19 +00003143 .automatic_shrinking = true,
Herbert Xuc428ecd2015-03-20 21:57:01 +11003144};
3145
Linus Torvalds1da177e2005-04-16 15:20:36 -07003146static int __init netlink_proto_init(void)
3147{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003148 int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003149 int err = proto_register(&netlink_proto, 0);
3150
3151 if (err != 0)
3152 goto out;
3153
YOSHIFUJI Hideaki / 吉藤英明fab25742013-01-09 07:19:48 +00003154 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003155
Panagiotis Issaris0da974f2006-07-21 14:51:30 -07003156 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003157 if (!nl_table)
3158 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003159
Linus Torvalds1da177e2005-04-16 15:20:36 -07003160 for (i = 0; i < MAX_LINKS; i++) {
Herbert Xuc428ecd2015-03-20 21:57:01 +11003161 if (rhashtable_init(&nl_table[i].hash,
3162 &netlink_rhashtable_params) < 0) {
Thomas Grafe3416942014-08-02 11:47:45 +02003163 while (--i > 0)
3164 rhashtable_destroy(&nl_table[i].hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003165 kfree(nl_table);
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003166 goto panic;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003167 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003168 }
3169
Daniel Borkmannbcbde0d2013-06-21 19:38:07 +02003170 INIT_LIST_HEAD(&netlink_tap_all);
3171
David S. Millerb963ea82010-08-30 19:08:01 -07003172 netlink_add_usersock_entry();
3173
Linus Torvalds1da177e2005-04-16 15:20:36 -07003174 sock_register(&netlink_family_ops);
Eric W. Biedermanb4b51022007-09-12 13:05:38 +02003175 register_pernet_subsys(&netlink_net_ops);
YOSHIFUJI Hideaki746fac42007-02-09 23:25:07 +09003176 /* The netlink device handler may be needed early. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003177 rtnetlink_init();
3178out:
3179 return err;
Akinobu Mitafab2caf2006-08-29 02:15:24 -07003180panic:
3181 panic("netlink_init: Cannot allocate nl_table\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07003182}
3183
Linus Torvalds1da177e2005-04-16 15:20:36 -07003184core_initcall(netlink_proto_init);