Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
113aa838 | 4 | * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk> |
1da177e4 | 5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
cd1df525 | 6 | * Patrick McHardy <kaber@trash.net> |
1da177e4 LT |
7 | * |
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. | |
746fac4d | 12 | * |
1da177e4 LT |
13 | * 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) | |
4fdb3bb7 HW |
17 | * 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 | |
1da177e4 LT |
23 | */ |
24 | ||
1da177e4 LT |
25 | #include <linux/module.h> |
26 | ||
4fc268d2 | 27 | #include <linux/capability.h> |
1da177e4 LT |
28 | #include <linux/kernel.h> |
29 | #include <linux/init.h> | |
1da177e4 LT |
30 | #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> | |
1da177e4 LT |
49 | #include <linux/notifier.h> |
50 | #include <linux/security.h> | |
51 | #include <linux/jhash.h> | |
52 | #include <linux/jiffies.h> | |
53 | #include <linux/random.h> | |
54 | #include <linux/bitops.h> | |
55 | #include <linux/mm.h> | |
56 | #include <linux/types.h> | |
54e0f520 | 57 | #include <linux/audit.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
ccdfcc39 | 59 | #include <linux/vmalloc.h> |
bcbde0d4 | 60 | #include <linux/if_arp.h> |
e341694e | 61 | #include <linux/rhashtable.h> |
9652e931 | 62 | #include <asm/cacheflush.h> |
e341694e | 63 | #include <linux/hash.h> |
54e0f520 | 64 | |
457c4cbc | 65 | #include <net/net_namespace.h> |
1da177e4 LT |
66 | #include <net/sock.h> |
67 | #include <net/scm.h> | |
82ace47a | 68 | #include <net/netlink.h> |
1da177e4 | 69 | |
0f29c768 | 70 | #include "af_netlink.h" |
1da177e4 | 71 | |
5c398dc8 ED |
72 | struct listeners { |
73 | struct rcu_head rcu; | |
74 | unsigned long masks[0]; | |
6c04bb18 JB |
75 | }; |
76 | ||
cd967e05 PM |
77 | /* state bits */ |
78 | #define NETLINK_CONGESTED 0x0 | |
79 | ||
80 | /* flags */ | |
77247bbb | 81 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 82 | #define NETLINK_RECV_PKTINFO 0x2 |
be0c22a4 | 83 | #define NETLINK_BROADCAST_SEND_ERROR 0x4 |
38938bfe | 84 | #define NETLINK_RECV_NO_ENOBUFS 0x8 |
77247bbb | 85 | |
035c4c16 | 86 | static inline int netlink_is_kernel(struct sock *sk) |
aed81560 DL |
87 | { |
88 | return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET; | |
89 | } | |
90 | ||
0f29c768 AV |
91 | struct netlink_table *nl_table; |
92 | EXPORT_SYMBOL_GPL(nl_table); | |
1da177e4 LT |
93 | |
94 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
95 | ||
96 | static int netlink_dump(struct sock *sk); | |
9652e931 | 97 | static void netlink_skb_destructor(struct sk_buff *skb); |
1da177e4 | 98 | |
78fd1d0a TG |
99 | /* nl_table locking explained: |
100 | * Lookup and traversal are protected with nl_sk_hash_lock or nl_table_lock | |
101 | * combined with an RCU read-side lock. Insertion and removal are protected | |
102 | * with nl_sk_hash_lock while using RCU list modification primitives and may | |
103 | * run in parallel to nl_table_lock protected lookups. Destruction of the | |
104 | * Netlink socket may only occur *after* nl_table_lock has been acquired | |
105 | * either during or after the socket has been removed from the list. | |
106 | */ | |
0f29c768 AV |
107 | DEFINE_RWLOCK(nl_table_lock); |
108 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
109 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
110 | ||
6d772ac5 ED |
111 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
112 | ||
e341694e TG |
113 | /* Protects netlink socket hash table mutations */ |
114 | DEFINE_MUTEX(nl_sk_hash_lock); | |
6c8f7e70 | 115 | EXPORT_SYMBOL_GPL(nl_sk_hash_lock); |
e341694e | 116 | |
97127566 | 117 | #ifdef CONFIG_PROVE_LOCKING |
e341694e TG |
118 | static int lockdep_nl_sk_hash_is_held(void) |
119 | { | |
78fd1d0a TG |
120 | if (debug_locks) |
121 | return lockdep_is_held(&nl_sk_hash_lock) || lockdep_is_held(&nl_table_lock); | |
78fd1d0a | 122 | return 1; |
e341694e | 123 | } |
97127566 | 124 | #endif |
e341694e | 125 | |
e041c683 | 126 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 127 | |
bcbde0d4 DB |
128 | static DEFINE_SPINLOCK(netlink_tap_lock); |
129 | static struct list_head netlink_tap_all __read_mostly; | |
130 | ||
b57ef81f | 131 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
132 | { |
133 | return group ? 1 << (group - 1) : 0; | |
134 | } | |
135 | ||
bcbde0d4 DB |
136 | int netlink_add_tap(struct netlink_tap *nt) |
137 | { | |
138 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
139 | return -EINVAL; | |
140 | ||
141 | spin_lock(&netlink_tap_lock); | |
142 | list_add_rcu(&nt->list, &netlink_tap_all); | |
143 | spin_unlock(&netlink_tap_lock); | |
144 | ||
145 | if (nt->module) | |
146 | __module_get(nt->module); | |
147 | ||
148 | return 0; | |
149 | } | |
150 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
151 | ||
2173f8d9 | 152 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
153 | { |
154 | bool found = false; | |
155 | struct netlink_tap *tmp; | |
156 | ||
157 | spin_lock(&netlink_tap_lock); | |
158 | ||
159 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
160 | if (nt == tmp) { | |
161 | list_del_rcu(&nt->list); | |
162 | found = true; | |
163 | goto out; | |
164 | } | |
165 | } | |
166 | ||
167 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
168 | out: | |
169 | spin_unlock(&netlink_tap_lock); | |
170 | ||
171 | if (found && nt->module) | |
172 | module_put(nt->module); | |
173 | ||
174 | return found ? 0 : -ENODEV; | |
175 | } | |
bcbde0d4 DB |
176 | |
177 | int netlink_remove_tap(struct netlink_tap *nt) | |
178 | { | |
179 | int ret; | |
180 | ||
181 | ret = __netlink_remove_tap(nt); | |
182 | synchronize_net(); | |
183 | ||
184 | return ret; | |
185 | } | |
186 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
187 | ||
5ffd5cdd DB |
188 | static bool netlink_filter_tap(const struct sk_buff *skb) |
189 | { | |
190 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
191 | |
192 | /* We take the more conservative approach and | |
193 | * whitelist socket protocols that may pass. | |
194 | */ | |
195 | switch (sk->sk_protocol) { | |
196 | case NETLINK_ROUTE: | |
197 | case NETLINK_USERSOCK: | |
198 | case NETLINK_SOCK_DIAG: | |
199 | case NETLINK_NFLOG: | |
200 | case NETLINK_XFRM: | |
201 | case NETLINK_FIB_LOOKUP: | |
202 | case NETLINK_NETFILTER: | |
203 | case NETLINK_GENERIC: | |
498044bb | 204 | return true; |
5ffd5cdd DB |
205 | } |
206 | ||
498044bb | 207 | return false; |
5ffd5cdd DB |
208 | } |
209 | ||
bcbde0d4 DB |
210 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
211 | struct net_device *dev) | |
212 | { | |
213 | struct sk_buff *nskb; | |
5ffd5cdd | 214 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
215 | int ret = -ENOMEM; |
216 | ||
217 | dev_hold(dev); | |
218 | nskb = skb_clone(skb, GFP_ATOMIC); | |
219 | if (nskb) { | |
220 | nskb->dev = dev; | |
5ffd5cdd | 221 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
222 | nskb->pkt_type = netlink_is_kernel(sk) ? |
223 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 224 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
225 | ret = dev_queue_xmit(nskb); |
226 | if (unlikely(ret > 0)) | |
227 | ret = net_xmit_errno(ret); | |
228 | } | |
229 | ||
230 | dev_put(dev); | |
231 | return ret; | |
232 | } | |
233 | ||
234 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
235 | { | |
236 | int ret; | |
237 | struct netlink_tap *tmp; | |
238 | ||
5ffd5cdd DB |
239 | if (!netlink_filter_tap(skb)) |
240 | return; | |
241 | ||
bcbde0d4 DB |
242 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
243 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
244 | if (unlikely(ret)) | |
245 | break; | |
246 | } | |
247 | } | |
248 | ||
249 | static void netlink_deliver_tap(struct sk_buff *skb) | |
250 | { | |
251 | rcu_read_lock(); | |
252 | ||
253 | if (unlikely(!list_empty(&netlink_tap_all))) | |
254 | __netlink_deliver_tap(skb); | |
255 | ||
256 | rcu_read_unlock(); | |
257 | } | |
258 | ||
73bfd370 DB |
259 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
260 | struct sk_buff *skb) | |
261 | { | |
262 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
263 | netlink_deliver_tap(skb); | |
264 | } | |
265 | ||
cd1df525 PM |
266 | static void netlink_overrun(struct sock *sk) |
267 | { | |
268 | struct netlink_sock *nlk = nlk_sk(sk); | |
269 | ||
270 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
271 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
272 | sk->sk_err = ENOBUFS; | |
273 | sk->sk_error_report(sk); | |
274 | } | |
275 | } | |
276 | atomic_inc(&sk->sk_drops); | |
277 | } | |
278 | ||
279 | static void netlink_rcv_wake(struct sock *sk) | |
280 | { | |
281 | struct netlink_sock *nlk = nlk_sk(sk); | |
282 | ||
283 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
284 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
285 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
286 | wake_up_interruptible(&nlk->wait); | |
287 | } | |
288 | ||
ccdfcc39 | 289 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
290 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
291 | { | |
292 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
293 | } | |
294 | ||
f9c22888 PM |
295 | static bool netlink_rx_is_mmaped(struct sock *sk) |
296 | { | |
297 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
298 | } | |
299 | ||
5fd96123 PM |
300 | static bool netlink_tx_is_mmaped(struct sock *sk) |
301 | { | |
302 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
303 | } | |
304 | ||
ccdfcc39 PM |
305 | static __pure struct page *pgvec_to_page(const void *addr) |
306 | { | |
307 | if (is_vmalloc_addr(addr)) | |
308 | return vmalloc_to_page(addr); | |
309 | else | |
310 | return virt_to_page(addr); | |
311 | } | |
312 | ||
313 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
314 | { | |
315 | unsigned int i; | |
316 | ||
317 | for (i = 0; i < len; i++) { | |
318 | if (pg_vec[i] != NULL) { | |
319 | if (is_vmalloc_addr(pg_vec[i])) | |
320 | vfree(pg_vec[i]); | |
321 | else | |
322 | free_pages((unsigned long)pg_vec[i], order); | |
323 | } | |
324 | } | |
325 | kfree(pg_vec); | |
326 | } | |
327 | ||
328 | static void *alloc_one_pg_vec_page(unsigned long order) | |
329 | { | |
330 | void *buffer; | |
331 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
332 | __GFP_NOWARN | __GFP_NORETRY; | |
333 | ||
334 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
335 | if (buffer != NULL) | |
336 | return buffer; | |
337 | ||
338 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
339 | if (buffer != NULL) | |
340 | return buffer; | |
341 | ||
342 | gfp_flags &= ~__GFP_NORETRY; | |
343 | return (void *)__get_free_pages(gfp_flags, order); | |
344 | } | |
345 | ||
346 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
347 | struct nl_mmap_req *req, unsigned int order) | |
348 | { | |
349 | unsigned int block_nr = req->nm_block_nr; | |
350 | unsigned int i; | |
8a849bb7 | 351 | void **pg_vec; |
ccdfcc39 PM |
352 | |
353 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
354 | if (pg_vec == NULL) | |
355 | return NULL; | |
356 | ||
357 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 358 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
359 | if (pg_vec[i] == NULL) |
360 | goto err1; | |
361 | } | |
362 | ||
363 | return pg_vec; | |
364 | err1: | |
365 | free_pg_vec(pg_vec, order, block_nr); | |
366 | return NULL; | |
367 | } | |
368 | ||
369 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
370 | bool closing, bool tx_ring) | |
371 | { | |
372 | struct netlink_sock *nlk = nlk_sk(sk); | |
373 | struct netlink_ring *ring; | |
374 | struct sk_buff_head *queue; | |
375 | void **pg_vec = NULL; | |
376 | unsigned int order = 0; | |
377 | int err; | |
378 | ||
379 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
380 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
381 | ||
382 | if (!closing) { | |
383 | if (atomic_read(&nlk->mapped)) | |
384 | return -EBUSY; | |
385 | if (atomic_read(&ring->pending)) | |
386 | return -EBUSY; | |
387 | } | |
388 | ||
389 | if (req->nm_block_nr) { | |
390 | if (ring->pg_vec != NULL) | |
391 | return -EBUSY; | |
392 | ||
393 | if ((int)req->nm_block_size <= 0) | |
394 | return -EINVAL; | |
74e83b23 | 395 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
396 | return -EINVAL; |
397 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
398 | return -EINVAL; | |
399 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
400 | return -EINVAL; | |
401 | ||
402 | ring->frames_per_block = req->nm_block_size / | |
403 | req->nm_frame_size; | |
404 | if (ring->frames_per_block == 0) | |
405 | return -EINVAL; | |
406 | if (ring->frames_per_block * req->nm_block_nr != | |
407 | req->nm_frame_nr) | |
408 | return -EINVAL; | |
409 | ||
410 | order = get_order(req->nm_block_size); | |
411 | pg_vec = alloc_pg_vec(nlk, req, order); | |
412 | if (pg_vec == NULL) | |
413 | return -ENOMEM; | |
414 | } else { | |
415 | if (req->nm_frame_nr) | |
416 | return -EINVAL; | |
417 | } | |
418 | ||
419 | err = -EBUSY; | |
420 | mutex_lock(&nlk->pg_vec_lock); | |
421 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
422 | err = 0; | |
423 | spin_lock_bh(&queue->lock); | |
424 | ||
425 | ring->frame_max = req->nm_frame_nr - 1; | |
426 | ring->head = 0; | |
427 | ring->frame_size = req->nm_frame_size; | |
428 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
429 | ||
430 | swap(ring->pg_vec_len, req->nm_block_nr); | |
431 | swap(ring->pg_vec_order, order); | |
432 | swap(ring->pg_vec, pg_vec); | |
433 | ||
434 | __skb_queue_purge(queue); | |
435 | spin_unlock_bh(&queue->lock); | |
436 | ||
437 | WARN_ON(atomic_read(&nlk->mapped)); | |
438 | } | |
439 | mutex_unlock(&nlk->pg_vec_lock); | |
440 | ||
441 | if (pg_vec) | |
442 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
443 | return err; | |
444 | } | |
445 | ||
446 | static void netlink_mm_open(struct vm_area_struct *vma) | |
447 | { | |
448 | struct file *file = vma->vm_file; | |
449 | struct socket *sock = file->private_data; | |
450 | struct sock *sk = sock->sk; | |
451 | ||
452 | if (sk) | |
453 | atomic_inc(&nlk_sk(sk)->mapped); | |
454 | } | |
455 | ||
456 | static void netlink_mm_close(struct vm_area_struct *vma) | |
457 | { | |
458 | struct file *file = vma->vm_file; | |
459 | struct socket *sock = file->private_data; | |
460 | struct sock *sk = sock->sk; | |
461 | ||
462 | if (sk) | |
463 | atomic_dec(&nlk_sk(sk)->mapped); | |
464 | } | |
465 | ||
466 | static const struct vm_operations_struct netlink_mmap_ops = { | |
467 | .open = netlink_mm_open, | |
468 | .close = netlink_mm_close, | |
469 | }; | |
470 | ||
471 | static int netlink_mmap(struct file *file, struct socket *sock, | |
472 | struct vm_area_struct *vma) | |
473 | { | |
474 | struct sock *sk = sock->sk; | |
475 | struct netlink_sock *nlk = nlk_sk(sk); | |
476 | struct netlink_ring *ring; | |
477 | unsigned long start, size, expected; | |
478 | unsigned int i; | |
479 | int err = -EINVAL; | |
480 | ||
481 | if (vma->vm_pgoff) | |
482 | return -EINVAL; | |
483 | ||
484 | mutex_lock(&nlk->pg_vec_lock); | |
485 | ||
486 | expected = 0; | |
487 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
488 | if (ring->pg_vec == NULL) | |
489 | continue; | |
490 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
491 | } | |
492 | ||
493 | if (expected == 0) | |
494 | goto out; | |
495 | ||
496 | size = vma->vm_end - vma->vm_start; | |
497 | if (size != expected) | |
498 | goto out; | |
499 | ||
500 | start = vma->vm_start; | |
501 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
502 | if (ring->pg_vec == NULL) | |
503 | continue; | |
504 | ||
505 | for (i = 0; i < ring->pg_vec_len; i++) { | |
506 | struct page *page; | |
507 | void *kaddr = ring->pg_vec[i]; | |
508 | unsigned int pg_num; | |
509 | ||
510 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
511 | page = pgvec_to_page(kaddr); | |
512 | err = vm_insert_page(vma, start, page); | |
513 | if (err < 0) | |
514 | goto out; | |
515 | start += PAGE_SIZE; | |
516 | kaddr += PAGE_SIZE; | |
517 | } | |
518 | } | |
519 | } | |
520 | ||
521 | atomic_inc(&nlk->mapped); | |
522 | vma->vm_ops = &netlink_mmap_ops; | |
523 | err = 0; | |
524 | out: | |
525 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 526 | return err; |
ccdfcc39 | 527 | } |
9652e931 PM |
528 | |
529 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
530 | { | |
531 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
532 | struct page *p_start, *p_end; | |
533 | ||
534 | /* First page is flushed through netlink_{get,set}_status */ | |
535 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 536 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
537 | while (p_start <= p_end) { |
538 | flush_dcache_page(p_start); | |
539 | p_start++; | |
540 | } | |
541 | #endif | |
542 | } | |
543 | ||
544 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
545 | { | |
546 | smp_rmb(); | |
547 | flush_dcache_page(pgvec_to_page(hdr)); | |
548 | return hdr->nm_status; | |
549 | } | |
550 | ||
551 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
552 | enum nl_mmap_status status) | |
553 | { | |
554 | hdr->nm_status = status; | |
555 | flush_dcache_page(pgvec_to_page(hdr)); | |
556 | smp_wmb(); | |
557 | } | |
558 | ||
559 | static struct nl_mmap_hdr * | |
560 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
561 | { | |
562 | unsigned int pg_vec_pos, frame_off; | |
563 | ||
564 | pg_vec_pos = pos / ring->frames_per_block; | |
565 | frame_off = pos % ring->frames_per_block; | |
566 | ||
567 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
568 | } | |
569 | ||
570 | static struct nl_mmap_hdr * | |
571 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
572 | enum nl_mmap_status status) | |
573 | { | |
574 | struct nl_mmap_hdr *hdr; | |
575 | ||
576 | hdr = __netlink_lookup_frame(ring, pos); | |
577 | if (netlink_get_status(hdr) != status) | |
578 | return NULL; | |
579 | ||
580 | return hdr; | |
581 | } | |
582 | ||
583 | static struct nl_mmap_hdr * | |
584 | netlink_current_frame(const struct netlink_ring *ring, | |
585 | enum nl_mmap_status status) | |
586 | { | |
587 | return netlink_lookup_frame(ring, ring->head, status); | |
588 | } | |
589 | ||
590 | static struct nl_mmap_hdr * | |
591 | netlink_previous_frame(const struct netlink_ring *ring, | |
592 | enum nl_mmap_status status) | |
593 | { | |
594 | unsigned int prev; | |
595 | ||
596 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
597 | return netlink_lookup_frame(ring, prev, status); | |
598 | } | |
599 | ||
600 | static void netlink_increment_head(struct netlink_ring *ring) | |
601 | { | |
602 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
603 | } | |
604 | ||
605 | static void netlink_forward_ring(struct netlink_ring *ring) | |
606 | { | |
607 | unsigned int head = ring->head, pos = head; | |
608 | const struct nl_mmap_hdr *hdr; | |
609 | ||
610 | do { | |
611 | hdr = __netlink_lookup_frame(ring, pos); | |
612 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
613 | break; | |
614 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
615 | break; | |
616 | netlink_increment_head(ring); | |
617 | } while (ring->head != head); | |
618 | } | |
619 | ||
cd1df525 PM |
620 | static bool netlink_dump_space(struct netlink_sock *nlk) |
621 | { | |
622 | struct netlink_ring *ring = &nlk->rx_ring; | |
623 | struct nl_mmap_hdr *hdr; | |
624 | unsigned int n; | |
625 | ||
626 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
627 | if (hdr == NULL) | |
628 | return false; | |
629 | ||
630 | n = ring->head + ring->frame_max / 2; | |
631 | if (n > ring->frame_max) | |
632 | n -= ring->frame_max; | |
633 | ||
634 | hdr = __netlink_lookup_frame(ring, n); | |
635 | ||
636 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
637 | } | |
638 | ||
9652e931 PM |
639 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
640 | poll_table *wait) | |
641 | { | |
642 | struct sock *sk = sock->sk; | |
643 | struct netlink_sock *nlk = nlk_sk(sk); | |
644 | unsigned int mask; | |
cd1df525 | 645 | int err; |
9652e931 | 646 | |
cd1df525 PM |
647 | if (nlk->rx_ring.pg_vec != NULL) { |
648 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
649 | * for dumps is performed here. A dump is allowed to continue | |
650 | * if at least half the ring is unused. | |
651 | */ | |
16b304f3 | 652 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
653 | err = netlink_dump(sk); |
654 | if (err < 0) { | |
ac30ef83 | 655 | sk->sk_err = -err; |
cd1df525 PM |
656 | sk->sk_error_report(sk); |
657 | break; | |
658 | } | |
659 | } | |
660 | netlink_rcv_wake(sk); | |
661 | } | |
5fd96123 | 662 | |
9652e931 PM |
663 | mask = datagram_poll(file, sock, wait); |
664 | ||
665 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
666 | if (nlk->rx_ring.pg_vec) { | |
667 | netlink_forward_ring(&nlk->rx_ring); | |
668 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
669 | mask |= POLLIN | POLLRDNORM; | |
670 | } | |
671 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
672 | ||
673 | spin_lock_bh(&sk->sk_write_queue.lock); | |
674 | if (nlk->tx_ring.pg_vec) { | |
675 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
676 | mask |= POLLOUT | POLLWRNORM; | |
677 | } | |
678 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
679 | ||
680 | return mask; | |
681 | } | |
682 | ||
683 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
684 | { | |
685 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
686 | } | |
687 | ||
688 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
689 | struct netlink_ring *ring, | |
690 | struct nl_mmap_hdr *hdr) | |
691 | { | |
692 | unsigned int size; | |
693 | void *data; | |
694 | ||
695 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
696 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
697 | ||
698 | skb->head = data; | |
699 | skb->data = data; | |
700 | skb_reset_tail_pointer(skb); | |
701 | skb->end = skb->tail + size; | |
702 | skb->len = 0; | |
703 | ||
704 | skb->destructor = netlink_skb_destructor; | |
705 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
706 | NETLINK_CB(skb).sk = sk; | |
707 | } | |
5fd96123 PM |
708 | |
709 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
710 | u32 dst_portid, u32 dst_group, | |
711 | struct sock_iocb *siocb) | |
712 | { | |
713 | struct netlink_sock *nlk = nlk_sk(sk); | |
714 | struct netlink_ring *ring; | |
715 | struct nl_mmap_hdr *hdr; | |
716 | struct sk_buff *skb; | |
717 | unsigned int maxlen; | |
718 | bool excl = true; | |
719 | int err = 0, len = 0; | |
720 | ||
721 | /* Netlink messages are validated by the receiver before processing. | |
722 | * In order to avoid userspace changing the contents of the message | |
723 | * after validation, the socket and the ring may only be used by a | |
724 | * single process, otherwise we fall back to copying. | |
725 | */ | |
24dff96a | 726 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 1 || |
5fd96123 PM |
727 | atomic_read(&nlk->mapped) > 1) |
728 | excl = false; | |
729 | ||
730 | mutex_lock(&nlk->pg_vec_lock); | |
731 | ||
732 | ring = &nlk->tx_ring; | |
733 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
734 | ||
735 | do { | |
736 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
737 | if (hdr == NULL) { | |
738 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
739 | atomic_read(&nlk->tx_ring.pending)) | |
740 | schedule(); | |
741 | continue; | |
742 | } | |
743 | if (hdr->nm_len > maxlen) { | |
744 | err = -EINVAL; | |
745 | goto out; | |
746 | } | |
747 | ||
748 | netlink_frame_flush_dcache(hdr); | |
749 | ||
750 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
751 | skb = alloc_skb_head(GFP_KERNEL); | |
752 | if (skb == NULL) { | |
753 | err = -ENOBUFS; | |
754 | goto out; | |
755 | } | |
756 | sock_hold(sk); | |
757 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
758 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
759 | __skb_put(skb, hdr->nm_len); | |
760 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
761 | atomic_inc(&ring->pending); | |
762 | } else { | |
763 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
764 | if (skb == NULL) { | |
765 | err = -ENOBUFS; | |
766 | goto out; | |
767 | } | |
768 | __skb_put(skb, hdr->nm_len); | |
769 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
770 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
771 | } | |
772 | ||
773 | netlink_increment_head(ring); | |
774 | ||
775 | NETLINK_CB(skb).portid = nlk->portid; | |
776 | NETLINK_CB(skb).dst_group = dst_group; | |
777 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
778 | ||
779 | err = security_netlink_send(sk, skb); | |
780 | if (err) { | |
781 | kfree_skb(skb); | |
782 | goto out; | |
783 | } | |
784 | ||
785 | if (unlikely(dst_group)) { | |
786 | atomic_inc(&skb->users); | |
787 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
788 | GFP_KERNEL); | |
789 | } | |
790 | err = netlink_unicast(sk, skb, dst_portid, | |
791 | msg->msg_flags & MSG_DONTWAIT); | |
792 | if (err < 0) | |
793 | goto out; | |
794 | len += err; | |
795 | ||
796 | } while (hdr != NULL || | |
797 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
798 | atomic_read(&nlk->tx_ring.pending))); | |
799 | ||
800 | if (len > 0) | |
801 | err = len; | |
802 | out: | |
803 | mutex_unlock(&nlk->pg_vec_lock); | |
804 | return err; | |
805 | } | |
f9c22888 PM |
806 | |
807 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
808 | { | |
809 | struct nl_mmap_hdr *hdr; | |
810 | ||
811 | hdr = netlink_mmap_hdr(skb); | |
812 | hdr->nm_len = skb->len; | |
813 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
814 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
815 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
816 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
817 | netlink_frame_flush_dcache(hdr); |
818 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
819 | ||
820 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
821 | kfree_skb(skb); | |
822 | } | |
823 | ||
824 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
825 | { | |
826 | struct netlink_sock *nlk = nlk_sk(sk); | |
827 | struct netlink_ring *ring = &nlk->rx_ring; | |
828 | struct nl_mmap_hdr *hdr; | |
829 | ||
830 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
831 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
832 | if (hdr == NULL) { | |
833 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
834 | kfree_skb(skb); | |
cd1df525 | 835 | netlink_overrun(sk); |
f9c22888 PM |
836 | return; |
837 | } | |
838 | netlink_increment_head(ring); | |
839 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
840 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
841 | ||
842 | hdr->nm_len = skb->len; | |
843 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
844 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
845 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
846 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
847 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
848 | } | |
849 | ||
ccdfcc39 | 850 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 851 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 852 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 853 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 854 | #define netlink_mmap sock_no_mmap |
9652e931 | 855 | #define netlink_poll datagram_poll |
5fd96123 | 856 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
857 | #endif /* CONFIG_NETLINK_MMAP */ |
858 | ||
cf0a018a PM |
859 | static void netlink_skb_destructor(struct sk_buff *skb) |
860 | { | |
9652e931 PM |
861 | #ifdef CONFIG_NETLINK_MMAP |
862 | struct nl_mmap_hdr *hdr; | |
863 | struct netlink_ring *ring; | |
864 | struct sock *sk; | |
865 | ||
866 | /* If a packet from the kernel to userspace was freed because of an | |
867 | * error without being delivered to userspace, the kernel must reset | |
868 | * the status. In the direction userspace to kernel, the status is | |
869 | * always reset here after the packet was processed and freed. | |
870 | */ | |
871 | if (netlink_skb_is_mmaped(skb)) { | |
872 | hdr = netlink_mmap_hdr(skb); | |
873 | sk = NETLINK_CB(skb).sk; | |
874 | ||
5fd96123 PM |
875 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
876 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
877 | ring = &nlk_sk(sk)->tx_ring; | |
878 | } else { | |
879 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
880 | hdr->nm_len = 0; | |
881 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
882 | } | |
883 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 884 | } |
9652e931 PM |
885 | |
886 | WARN_ON(atomic_read(&ring->pending) == 0); | |
887 | atomic_dec(&ring->pending); | |
888 | sock_put(sk); | |
889 | ||
5e71d9d7 | 890 | skb->head = NULL; |
9652e931 PM |
891 | } |
892 | #endif | |
c05cdb1b | 893 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
894 | if (!skb->cloned || |
895 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
896 | vfree(skb->head); | |
897 | ||
c05cdb1b PNA |
898 | skb->head = NULL; |
899 | } | |
9652e931 PM |
900 | if (skb->sk != NULL) |
901 | sock_rfree(skb); | |
cf0a018a PM |
902 | } |
903 | ||
904 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
905 | { | |
906 | WARN_ON(skb->sk != NULL); | |
907 | skb->sk = sk; | |
908 | skb->destructor = netlink_skb_destructor; | |
909 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
910 | sk_mem_charge(sk, skb->truesize); | |
911 | } | |
912 | ||
1da177e4 LT |
913 | static void netlink_sock_destruct(struct sock *sk) |
914 | { | |
3f660d66 HX |
915 | struct netlink_sock *nlk = nlk_sk(sk); |
916 | ||
16b304f3 PS |
917 | if (nlk->cb_running) { |
918 | if (nlk->cb.done) | |
919 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 920 | |
16b304f3 PS |
921 | module_put(nlk->cb.module); |
922 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
923 | } |
924 | ||
1da177e4 | 925 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
926 | #ifdef CONFIG_NETLINK_MMAP |
927 | if (1) { | |
928 | struct nl_mmap_req req; | |
929 | ||
930 | memset(&req, 0, sizeof(req)); | |
931 | if (nlk->rx_ring.pg_vec) | |
932 | netlink_set_ring(sk, &req, true, false); | |
933 | memset(&req, 0, sizeof(req)); | |
934 | if (nlk->tx_ring.pg_vec) | |
935 | netlink_set_ring(sk, &req, true, true); | |
936 | } | |
937 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
938 | |
939 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 940 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
941 | return; |
942 | } | |
547b792c IJ |
943 | |
944 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
945 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
946 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
947 | } |
948 | ||
6ac552fd PM |
949 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
950 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
951 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
952 | * this, _but_ remember, it adds useless work on UP machines. | |
953 | */ | |
954 | ||
d136f1bd | 955 | void netlink_table_grab(void) |
9a429c49 | 956 | __acquires(nl_table_lock) |
1da177e4 | 957 | { |
d136f1bd JB |
958 | might_sleep(); |
959 | ||
6abd219c | 960 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
961 | |
962 | if (atomic_read(&nl_table_users)) { | |
963 | DECLARE_WAITQUEUE(wait, current); | |
964 | ||
965 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 966 | for (;;) { |
1da177e4 LT |
967 | set_current_state(TASK_UNINTERRUPTIBLE); |
968 | if (atomic_read(&nl_table_users) == 0) | |
969 | break; | |
6abd219c | 970 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 971 | schedule(); |
6abd219c | 972 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
973 | } |
974 | ||
975 | __set_current_state(TASK_RUNNING); | |
976 | remove_wait_queue(&nl_table_wait, &wait); | |
977 | } | |
978 | } | |
979 | ||
d136f1bd | 980 | void netlink_table_ungrab(void) |
9a429c49 | 981 | __releases(nl_table_lock) |
1da177e4 | 982 | { |
6abd219c | 983 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
984 | wake_up(&nl_table_wait); |
985 | } | |
986 | ||
6ac552fd | 987 | static inline void |
1da177e4 LT |
988 | netlink_lock_table(void) |
989 | { | |
990 | /* read_lock() synchronizes us to netlink_table_grab */ | |
991 | ||
992 | read_lock(&nl_table_lock); | |
993 | atomic_inc(&nl_table_users); | |
994 | read_unlock(&nl_table_lock); | |
995 | } | |
996 | ||
6ac552fd | 997 | static inline void |
1da177e4 LT |
998 | netlink_unlock_table(void) |
999 | { | |
1000 | if (atomic_dec_and_test(&nl_table_users)) | |
1001 | wake_up(&nl_table_wait); | |
1002 | } | |
1003 | ||
e341694e | 1004 | struct netlink_compare_arg |
1da177e4 | 1005 | { |
e341694e TG |
1006 | struct net *net; |
1007 | u32 portid; | |
1008 | }; | |
1da177e4 | 1009 | |
e341694e | 1010 | static bool netlink_compare(void *ptr, void *arg) |
1da177e4 | 1011 | { |
e341694e TG |
1012 | struct netlink_compare_arg *x = arg; |
1013 | struct sock *sk = ptr; | |
1da177e4 | 1014 | |
e341694e TG |
1015 | return nlk_sk(sk)->portid == x->portid && |
1016 | net_eq(sock_net(sk), x->net); | |
1da177e4 LT |
1017 | } |
1018 | ||
e341694e TG |
1019 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1020 | struct net *net) | |
1da177e4 | 1021 | { |
e341694e TG |
1022 | struct netlink_compare_arg arg = { |
1023 | .net = net, | |
1024 | .portid = portid, | |
1025 | }; | |
1026 | u32 hash; | |
1da177e4 | 1027 | |
e341694e | 1028 | hash = rhashtable_hashfn(&table->hash, &portid, sizeof(portid)); |
1da177e4 | 1029 | |
e341694e TG |
1030 | return rhashtable_lookup_compare(&table->hash, hash, |
1031 | &netlink_compare, &arg); | |
1da177e4 LT |
1032 | } |
1033 | ||
e341694e | 1034 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1035 | { |
e341694e TG |
1036 | struct netlink_table *table = &nl_table[protocol]; |
1037 | struct sock *sk; | |
1da177e4 | 1038 | |
78fd1d0a | 1039 | read_lock(&nl_table_lock); |
e341694e TG |
1040 | rcu_read_lock(); |
1041 | sk = __netlink_lookup(table, portid, net); | |
1042 | if (sk) | |
1043 | sock_hold(sk); | |
1044 | rcu_read_unlock(); | |
78fd1d0a | 1045 | read_unlock(&nl_table_lock); |
1da177e4 | 1046 | |
e341694e | 1047 | return sk; |
1da177e4 LT |
1048 | } |
1049 | ||
90ddc4f0 | 1050 | static const struct proto_ops netlink_ops; |
1da177e4 | 1051 | |
4277a083 PM |
1052 | static void |
1053 | netlink_update_listeners(struct sock *sk) | |
1054 | { | |
1055 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1056 | unsigned long mask; |
1057 | unsigned int i; | |
6d772ac5 ED |
1058 | struct listeners *listeners; |
1059 | ||
1060 | listeners = nl_deref_protected(tbl->listeners); | |
1061 | if (!listeners) | |
1062 | return; | |
4277a083 | 1063 | |
b4ff4f04 | 1064 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1065 | mask = 0; |
b67bfe0d | 1066 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1067 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1068 | mask |= nlk_sk(sk)->groups[i]; | |
1069 | } | |
6d772ac5 | 1070 | listeners->masks[i] = mask; |
4277a083 PM |
1071 | } |
1072 | /* this function is only called with the netlink table "grabbed", which | |
1073 | * makes sure updates are visible before bind or setsockopt return. */ | |
1074 | } | |
1075 | ||
15e47304 | 1076 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1077 | { |
da12c90e | 1078 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1da177e4 | 1079 | int err = -EADDRINUSE; |
1da177e4 | 1080 | |
e341694e TG |
1081 | mutex_lock(&nl_sk_hash_lock); |
1082 | if (__netlink_lookup(table, portid, net)) | |
1da177e4 LT |
1083 | goto err; |
1084 | ||
1085 | err = -EBUSY; | |
15e47304 | 1086 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1087 | goto err; |
1088 | ||
1089 | err = -ENOMEM; | |
e341694e | 1090 | if (BITS_PER_LONG > 32 && unlikely(table->hash.nelems >= UINT_MAX)) |
1da177e4 LT |
1091 | goto err; |
1092 | ||
15e47304 | 1093 | nlk_sk(sk)->portid = portid; |
e341694e TG |
1094 | sock_hold(sk); |
1095 | rhashtable_insert(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL); | |
1da177e4 | 1096 | err = 0; |
1da177e4 | 1097 | err: |
e341694e | 1098 | mutex_unlock(&nl_sk_hash_lock); |
1da177e4 LT |
1099 | return err; |
1100 | } | |
1101 | ||
1102 | static void netlink_remove(struct sock *sk) | |
1103 | { | |
e341694e TG |
1104 | struct netlink_table *table; |
1105 | ||
1106 | mutex_lock(&nl_sk_hash_lock); | |
1107 | table = &nl_table[sk->sk_protocol]; | |
1108 | if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL)) { | |
1109 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
1110 | __sock_put(sk); | |
1111 | } | |
1112 | mutex_unlock(&nl_sk_hash_lock); | |
1113 | ||
1da177e4 | 1114 | netlink_table_grab(); |
f7fa9b10 | 1115 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1116 | __sk_del_bind_node(sk); |
1117 | netlink_table_ungrab(); | |
1118 | } | |
1119 | ||
1120 | static struct proto netlink_proto = { | |
1121 | .name = "NETLINK", | |
1122 | .owner = THIS_MODULE, | |
1123 | .obj_size = sizeof(struct netlink_sock), | |
1124 | }; | |
1125 | ||
1b8d7ae4 EB |
1126 | static int __netlink_create(struct net *net, struct socket *sock, |
1127 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1128 | { |
1129 | struct sock *sk; | |
1130 | struct netlink_sock *nlk; | |
ab33a171 PM |
1131 | |
1132 | sock->ops = &netlink_ops; | |
1133 | ||
6257ff21 | 1134 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1135 | if (!sk) |
1136 | return -ENOMEM; | |
1137 | ||
1138 | sock_init_data(sock, sk); | |
1139 | ||
1140 | nlk = nlk_sk(sk); | |
658cb354 | 1141 | if (cb_mutex) { |
ffa4d721 | 1142 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1143 | } else { |
ffa4d721 PM |
1144 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1145 | mutex_init(nlk->cb_mutex); | |
1146 | } | |
ab33a171 | 1147 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1148 | #ifdef CONFIG_NETLINK_MMAP |
1149 | mutex_init(&nlk->pg_vec_lock); | |
1150 | #endif | |
ab33a171 PM |
1151 | |
1152 | sk->sk_destruct = netlink_sock_destruct; | |
1153 | sk->sk_protocol = protocol; | |
1154 | return 0; | |
1155 | } | |
1156 | ||
3f378b68 EP |
1157 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1158 | int kern) | |
ab33a171 PM |
1159 | { |
1160 | struct module *module = NULL; | |
af65bdfc | 1161 | struct mutex *cb_mutex; |
f7fa9b10 | 1162 | struct netlink_sock *nlk; |
4f520900 RGB |
1163 | int (*bind)(int group); |
1164 | void (*unbind)(int group); | |
ab33a171 | 1165 | int err = 0; |
1da177e4 LT |
1166 | |
1167 | sock->state = SS_UNCONNECTED; | |
1168 | ||
1169 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1170 | return -ESOCKTNOSUPPORT; | |
1171 | ||
6ac552fd | 1172 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1173 | return -EPROTONOSUPPORT; |
1174 | ||
77247bbb | 1175 | netlink_lock_table(); |
95a5afca | 1176 | #ifdef CONFIG_MODULES |
ab33a171 | 1177 | if (!nl_table[protocol].registered) { |
77247bbb | 1178 | netlink_unlock_table(); |
4fdb3bb7 | 1179 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1180 | netlink_lock_table(); |
4fdb3bb7 | 1181 | } |
ab33a171 PM |
1182 | #endif |
1183 | if (nl_table[protocol].registered && | |
1184 | try_module_get(nl_table[protocol].module)) | |
1185 | module = nl_table[protocol].module; | |
974c37e9 AD |
1186 | else |
1187 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1188 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1189 | bind = nl_table[protocol].bind; |
4f520900 | 1190 | unbind = nl_table[protocol].unbind; |
77247bbb | 1191 | netlink_unlock_table(); |
4fdb3bb7 | 1192 | |
974c37e9 AD |
1193 | if (err < 0) |
1194 | goto out; | |
1195 | ||
6ac552fd PM |
1196 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1197 | if (err < 0) | |
f7fa9b10 PM |
1198 | goto out_module; |
1199 | ||
6f756a8c | 1200 | local_bh_disable(); |
c1fd3b94 | 1201 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1202 | local_bh_enable(); |
1203 | ||
f7fa9b10 | 1204 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1205 | nlk->module = module; |
03292745 | 1206 | nlk->netlink_bind = bind; |
4f520900 | 1207 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1208 | out: |
1209 | return err; | |
1da177e4 | 1210 | |
ab33a171 PM |
1211 | out_module: |
1212 | module_put(module); | |
1213 | goto out; | |
1da177e4 LT |
1214 | } |
1215 | ||
1216 | static 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); | |
ac57b3a9 | 1225 | sock_orphan(sk); |
1da177e4 LT |
1226 | nlk = nlk_sk(sk); |
1227 | ||
3f660d66 HX |
1228 | /* |
1229 | * OK. Socket is unlinked, any packets that arrive now | |
1230 | * will be purged. | |
1231 | */ | |
1da177e4 | 1232 | |
1da177e4 LT |
1233 | sock->sk = NULL; |
1234 | wake_up_interruptible_all(&nlk->wait); | |
1235 | ||
1236 | skb_queue_purge(&sk->sk_write_queue); | |
1237 | ||
15e47304 | 1238 | if (nlk->portid) { |
1da177e4 | 1239 | struct netlink_notify n = { |
3b1e0a65 | 1240 | .net = sock_net(sk), |
1da177e4 | 1241 | .protocol = sk->sk_protocol, |
15e47304 | 1242 | .portid = nlk->portid, |
1da177e4 | 1243 | }; |
e041c683 AS |
1244 | atomic_notifier_call_chain(&netlink_chain, |
1245 | NETLINK_URELEASE, &n); | |
746fac4d | 1246 | } |
4fdb3bb7 | 1247 | |
5e7c001c | 1248 | module_put(nlk->module); |
4fdb3bb7 | 1249 | |
4277a083 | 1250 | netlink_table_grab(); |
aed81560 | 1251 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1252 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1253 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1254 | struct listeners *old; |
1255 | ||
1256 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1257 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1258 | kfree_rcu(old, rcu); | |
869e58f8 | 1259 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1260 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1261 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1262 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1263 | nl_table[sk->sk_protocol].registered = 0; |
1264 | } | |
658cb354 | 1265 | } else if (nlk->subscriptions) { |
4277a083 | 1266 | netlink_update_listeners(sk); |
658cb354 | 1267 | } |
4277a083 | 1268 | netlink_table_ungrab(); |
77247bbb | 1269 | |
f7fa9b10 PM |
1270 | kfree(nlk->groups); |
1271 | nlk->groups = NULL; | |
1272 | ||
3755810c | 1273 | local_bh_disable(); |
c1fd3b94 | 1274 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1275 | local_bh_enable(); |
1da177e4 LT |
1276 | sock_put(sk); |
1277 | return 0; | |
1278 | } | |
1279 | ||
1280 | static int netlink_autobind(struct socket *sock) | |
1281 | { | |
1282 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1283 | struct net *net = sock_net(sk); |
da12c90e | 1284 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1285 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1286 | int err; |
1287 | static s32 rover = -4097; | |
1288 | ||
1289 | retry: | |
1290 | cond_resched(); | |
78fd1d0a | 1291 | netlink_table_grab(); |
e341694e TG |
1292 | rcu_read_lock(); |
1293 | if (__netlink_lookup(table, portid, net)) { | |
1294 | /* Bind collision, search negative portid values. */ | |
1295 | portid = rover--; | |
1296 | if (rover > -4097) | |
1297 | rover = -4097; | |
1298 | rcu_read_unlock(); | |
78fd1d0a | 1299 | netlink_table_ungrab(); |
e341694e | 1300 | goto retry; |
1da177e4 | 1301 | } |
e341694e | 1302 | rcu_read_unlock(); |
78fd1d0a | 1303 | netlink_table_ungrab(); |
1da177e4 | 1304 | |
15e47304 | 1305 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1306 | if (err == -EADDRINUSE) |
1307 | goto retry; | |
d470e3b4 DM |
1308 | |
1309 | /* If 2 threads race to autobind, that is fine. */ | |
1310 | if (err == -EBUSY) | |
1311 | err = 0; | |
1312 | ||
1313 | return err; | |
1da177e4 LT |
1314 | } |
1315 | ||
aa4cf945 EB |
1316 | /** |
1317 | * __netlink_ns_capable - General netlink message capability test | |
1318 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1319 | * @user_ns: The user namespace of the capability to use | |
1320 | * @cap: The capability to use | |
1321 | * | |
1322 | * Test to see if the opener of the socket we received the message | |
1323 | * from had when the netlink socket was created and the sender of the | |
1324 | * message has has the capability @cap in the user namespace @user_ns. | |
1325 | */ | |
1326 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1327 | struct user_namespace *user_ns, int cap) | |
1328 | { | |
2d7a85f4 EB |
1329 | return ((nsp->flags & NETLINK_SKB_DST) || |
1330 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1331 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1332 | } |
1333 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1334 | ||
1335 | /** | |
1336 | * netlink_ns_capable - General netlink message capability test | |
1337 | * @skb: socket buffer holding a netlink command from userspace | |
1338 | * @user_ns: The user namespace of the capability to use | |
1339 | * @cap: The capability to use | |
1340 | * | |
1341 | * Test to see if the opener of the socket we received the message | |
1342 | * from had when the netlink socket was created and the sender of the | |
1343 | * message has has the capability @cap in the user namespace @user_ns. | |
1344 | */ | |
1345 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1346 | struct user_namespace *user_ns, int cap) | |
1347 | { | |
1348 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1349 | } | |
1350 | EXPORT_SYMBOL(netlink_ns_capable); | |
1351 | ||
1352 | /** | |
1353 | * netlink_capable - Netlink global message capability test | |
1354 | * @skb: socket buffer holding a netlink command from userspace | |
1355 | * @cap: The capability to use | |
1356 | * | |
1357 | * Test to see if the opener of the socket we received the message | |
1358 | * from had when the netlink socket was created and the sender of the | |
1359 | * message has has the capability @cap in all user namespaces. | |
1360 | */ | |
1361 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1362 | { | |
1363 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1364 | } | |
1365 | EXPORT_SYMBOL(netlink_capable); | |
1366 | ||
1367 | /** | |
1368 | * netlink_net_capable - Netlink network namespace message capability test | |
1369 | * @skb: socket buffer holding a netlink command from userspace | |
1370 | * @cap: The capability to use | |
1371 | * | |
1372 | * Test to see if the opener of the socket we received the message | |
1373 | * from had when the netlink socket was created and the sender of the | |
1374 | * message has has the capability @cap over the network namespace of | |
1375 | * the socket we received the message from. | |
1376 | */ | |
1377 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1378 | { | |
1379 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1380 | } | |
1381 | EXPORT_SYMBOL(netlink_net_capable); | |
1382 | ||
5187cd05 | 1383 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1384 | { |
9785e10a | 1385 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1386 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1387 | } |
1da177e4 | 1388 | |
f7fa9b10 PM |
1389 | static void |
1390 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1391 | { | |
1392 | struct netlink_sock *nlk = nlk_sk(sk); | |
1393 | ||
1394 | if (nlk->subscriptions && !subscriptions) | |
1395 | __sk_del_bind_node(sk); | |
1396 | else if (!nlk->subscriptions && subscriptions) | |
1397 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1398 | nlk->subscriptions = subscriptions; | |
1399 | } | |
1400 | ||
b4ff4f04 | 1401 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1402 | { |
1403 | struct netlink_sock *nlk = nlk_sk(sk); | |
1404 | unsigned int groups; | |
b4ff4f04 | 1405 | unsigned long *new_groups; |
513c2500 PM |
1406 | int err = 0; |
1407 | ||
b4ff4f04 JB |
1408 | netlink_table_grab(); |
1409 | ||
513c2500 | 1410 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1411 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1412 | err = -ENOENT; |
b4ff4f04 JB |
1413 | goto out_unlock; |
1414 | } | |
513c2500 | 1415 | |
b4ff4f04 JB |
1416 | if (nlk->ngroups >= groups) |
1417 | goto out_unlock; | |
513c2500 | 1418 | |
b4ff4f04 JB |
1419 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1420 | if (new_groups == NULL) { | |
1421 | err = -ENOMEM; | |
1422 | goto out_unlock; | |
1423 | } | |
6ac552fd | 1424 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1425 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1426 | ||
1427 | nlk->groups = new_groups; | |
513c2500 | 1428 | nlk->ngroups = groups; |
b4ff4f04 JB |
1429 | out_unlock: |
1430 | netlink_table_ungrab(); | |
1431 | return err; | |
513c2500 PM |
1432 | } |
1433 | ||
4f520900 RGB |
1434 | static void netlink_unbind(int group, long unsigned int groups, |
1435 | struct netlink_sock *nlk) | |
1436 | { | |
1437 | int undo; | |
1438 | ||
1439 | if (!nlk->netlink_unbind) | |
1440 | return; | |
1441 | ||
1442 | for (undo = 0; undo < group; undo++) | |
1443 | if (test_bit(group, &groups)) | |
1444 | nlk->netlink_unbind(undo); | |
1445 | } | |
1446 | ||
6ac552fd PM |
1447 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1448 | int addr_len) | |
1da177e4 LT |
1449 | { |
1450 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1451 | struct net *net = sock_net(sk); |
1da177e4 LT |
1452 | struct netlink_sock *nlk = nlk_sk(sk); |
1453 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1454 | int err; | |
4f520900 | 1455 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1456 | |
4e4b5376 HFS |
1457 | if (addr_len < sizeof(struct sockaddr_nl)) |
1458 | return -EINVAL; | |
1459 | ||
1da177e4 LT |
1460 | if (nladdr->nl_family != AF_NETLINK) |
1461 | return -EINVAL; | |
1462 | ||
1463 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1464 | if (groups) { |
5187cd05 | 1465 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1466 | return -EPERM; |
b4ff4f04 JB |
1467 | err = netlink_realloc_groups(sk); |
1468 | if (err) | |
1469 | return err; | |
513c2500 | 1470 | } |
1da177e4 | 1471 | |
4f520900 | 1472 | if (nlk->portid) |
15e47304 | 1473 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1474 | return -EINVAL; |
4f520900 RGB |
1475 | |
1476 | if (nlk->netlink_bind && groups) { | |
1477 | int group; | |
1478 | ||
1479 | for (group = 0; group < nlk->ngroups; group++) { | |
1480 | if (!test_bit(group, &groups)) | |
1481 | continue; | |
1482 | err = nlk->netlink_bind(group); | |
1483 | if (!err) | |
1484 | continue; | |
1485 | netlink_unbind(group, groups, nlk); | |
1486 | return err; | |
1487 | } | |
1488 | } | |
1489 | ||
1490 | if (!nlk->portid) { | |
1da177e4 | 1491 | err = nladdr->nl_pid ? |
b4b51029 | 1492 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1493 | netlink_autobind(sock); |
4f520900 RGB |
1494 | if (err) { |
1495 | netlink_unbind(nlk->ngroups - 1, groups, nlk); | |
1da177e4 | 1496 | return err; |
4f520900 | 1497 | } |
1da177e4 LT |
1498 | } |
1499 | ||
4f520900 | 1500 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1501 | return 0; |
1502 | ||
1503 | netlink_table_grab(); | |
f7fa9b10 | 1504 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1505 | hweight32(groups) - |
746fac4d | 1506 | hweight32(nlk->groups[0])); |
4f520900 | 1507 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1508 | netlink_update_listeners(sk); |
1da177e4 LT |
1509 | netlink_table_ungrab(); |
1510 | ||
1511 | return 0; | |
1512 | } | |
1513 | ||
1514 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1515 | int alen, int flags) | |
1516 | { | |
1517 | int err = 0; | |
1518 | struct sock *sk = sock->sk; | |
1519 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1520 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1521 | |
6503d961 CG |
1522 | if (alen < sizeof(addr->sa_family)) |
1523 | return -EINVAL; | |
1524 | ||
1da177e4 LT |
1525 | if (addr->sa_family == AF_UNSPEC) { |
1526 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1527 | nlk->dst_portid = 0; |
d629b836 | 1528 | nlk->dst_group = 0; |
1da177e4 LT |
1529 | return 0; |
1530 | } | |
1531 | if (addr->sa_family != AF_NETLINK) | |
1532 | return -EINVAL; | |
1533 | ||
46833a86 | 1534 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1535 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1536 | return -EPERM; |
1537 | ||
15e47304 | 1538 | if (!nlk->portid) |
1da177e4 LT |
1539 | err = netlink_autobind(sock); |
1540 | ||
1541 | if (err == 0) { | |
1542 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1543 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1544 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1545 | } |
1546 | ||
1547 | return err; | |
1548 | } | |
1549 | ||
6ac552fd PM |
1550 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1551 | int *addr_len, int peer) | |
1da177e4 LT |
1552 | { |
1553 | struct sock *sk = sock->sk; | |
1554 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1555 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1556 | |
1da177e4 LT |
1557 | nladdr->nl_family = AF_NETLINK; |
1558 | nladdr->nl_pad = 0; | |
1559 | *addr_len = sizeof(*nladdr); | |
1560 | ||
1561 | if (peer) { | |
15e47304 | 1562 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1563 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1564 | } else { |
15e47304 | 1565 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1566 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1567 | } |
1568 | return 0; | |
1569 | } | |
1570 | ||
15e47304 | 1571 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1572 | { |
1da177e4 LT |
1573 | struct sock *sock; |
1574 | struct netlink_sock *nlk; | |
1575 | ||
15e47304 | 1576 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1577 | if (!sock) |
1578 | return ERR_PTR(-ECONNREFUSED); | |
1579 | ||
1580 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1581 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1582 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1583 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1584 | sock_put(sock); |
1585 | return ERR_PTR(-ECONNREFUSED); | |
1586 | } | |
1587 | return sock; | |
1588 | } | |
1589 | ||
1590 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1591 | { | |
496ad9aa | 1592 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1593 | struct sock *sock; |
1594 | ||
1595 | if (!S_ISSOCK(inode->i_mode)) | |
1596 | return ERR_PTR(-ENOTSOCK); | |
1597 | ||
1598 | sock = SOCKET_I(inode)->sk; | |
1599 | if (sock->sk_family != AF_NETLINK) | |
1600 | return ERR_PTR(-EINVAL); | |
1601 | ||
1602 | sock_hold(sock); | |
1603 | return sock; | |
1604 | } | |
1605 | ||
3a36515f PN |
1606 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1607 | int broadcast) | |
c05cdb1b PNA |
1608 | { |
1609 | struct sk_buff *skb; | |
1610 | void *data; | |
1611 | ||
3a36515f | 1612 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1613 | return alloc_skb(size, GFP_KERNEL); |
1614 | ||
3a36515f PN |
1615 | size = SKB_DATA_ALIGN(size) + |
1616 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1617 | |
1618 | data = vmalloc(size); | |
1619 | if (data == NULL) | |
3a36515f | 1620 | return NULL; |
c05cdb1b | 1621 | |
3a36515f PN |
1622 | skb = build_skb(data, size); |
1623 | if (skb == NULL) | |
1624 | vfree(data); | |
1625 | else { | |
1626 | skb->head_frag = 0; | |
1627 | skb->destructor = netlink_skb_destructor; | |
1628 | } | |
c05cdb1b PNA |
1629 | |
1630 | return skb; | |
c05cdb1b PNA |
1631 | } |
1632 | ||
1da177e4 LT |
1633 | /* |
1634 | * Attach a skb to a netlink socket. | |
1635 | * The caller must hold a reference to the destination socket. On error, the | |
1636 | * reference is dropped. The skb is not send to the destination, just all | |
1637 | * all error checks are performed and memory in the queue is reserved. | |
1638 | * Return values: | |
1639 | * < 0: error. skb freed, reference to sock dropped. | |
1640 | * 0: continue | |
1641 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1642 | */ | |
9457afee | 1643 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1644 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1645 | { |
1646 | struct netlink_sock *nlk; | |
1647 | ||
1648 | nlk = nlk_sk(sk); | |
1649 | ||
5fd96123 PM |
1650 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1651 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1652 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1653 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1654 | if (!*timeo) { |
aed81560 | 1655 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1656 | netlink_overrun(sk); |
1657 | sock_put(sk); | |
1658 | kfree_skb(skb); | |
1659 | return -EAGAIN; | |
1660 | } | |
1661 | ||
1662 | __set_current_state(TASK_INTERRUPTIBLE); | |
1663 | add_wait_queue(&nlk->wait, &wait); | |
1664 | ||
1665 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1666 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1667 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1668 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1669 | |
1670 | __set_current_state(TASK_RUNNING); | |
1671 | remove_wait_queue(&nlk->wait, &wait); | |
1672 | sock_put(sk); | |
1673 | ||
1674 | if (signal_pending(current)) { | |
1675 | kfree_skb(skb); | |
c3d8d1e3 | 1676 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1677 | } |
1678 | return 1; | |
1679 | } | |
cf0a018a | 1680 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1681 | return 0; |
1682 | } | |
1683 | ||
4a7e7c2a | 1684 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1685 | { |
1da177e4 LT |
1686 | int len = skb->len; |
1687 | ||
bcbde0d4 DB |
1688 | netlink_deliver_tap(skb); |
1689 | ||
f9c22888 PM |
1690 | #ifdef CONFIG_NETLINK_MMAP |
1691 | if (netlink_skb_is_mmaped(skb)) | |
1692 | netlink_queue_mmaped_skb(sk, skb); | |
1693 | else if (netlink_rx_is_mmaped(sk)) | |
1694 | netlink_ring_set_copied(sk, skb); | |
1695 | else | |
1696 | #endif /* CONFIG_NETLINK_MMAP */ | |
1697 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1698 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1699 | return len; |
1700 | } | |
1701 | ||
1702 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1703 | { | |
1704 | int len = __netlink_sendskb(sk, skb); | |
1705 | ||
1da177e4 LT |
1706 | sock_put(sk); |
1707 | return len; | |
1708 | } | |
1709 | ||
1710 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1711 | { | |
1712 | kfree_skb(skb); | |
1713 | sock_put(sk); | |
1714 | } | |
1715 | ||
b57ef81f | 1716 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1717 | { |
1718 | int delta; | |
1719 | ||
1298ca46 | 1720 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1721 | if (netlink_skb_is_mmaped(skb)) |
1722 | return skb; | |
1da177e4 | 1723 | |
4305b541 | 1724 | delta = skb->end - skb->tail; |
c05cdb1b | 1725 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1726 | return skb; |
1727 | ||
1728 | if (skb_shared(skb)) { | |
1729 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1730 | if (!nskb) | |
1731 | return skb; | |
8460c00f | 1732 | consume_skb(skb); |
1da177e4 LT |
1733 | skb = nskb; |
1734 | } | |
1735 | ||
1736 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1737 | skb->truesize -= delta; | |
1738 | ||
1739 | return skb; | |
1740 | } | |
1741 | ||
3fbc2905 EB |
1742 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1743 | struct sock *ssk) | |
cd40b7d3 DL |
1744 | { |
1745 | int ret; | |
1746 | struct netlink_sock *nlk = nlk_sk(sk); | |
1747 | ||
1748 | ret = -ECONNREFUSED; | |
1749 | if (nlk->netlink_rcv != NULL) { | |
1750 | ret = skb->len; | |
cf0a018a | 1751 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1752 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1753 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1754 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1755 | consume_skb(skb); |
1756 | } else { | |
1757 | kfree_skb(skb); | |
cd40b7d3 | 1758 | } |
cd40b7d3 DL |
1759 | sock_put(sk); |
1760 | return ret; | |
1761 | } | |
1762 | ||
1763 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1764 | u32 portid, int nonblock) |
1da177e4 LT |
1765 | { |
1766 | struct sock *sk; | |
1767 | int err; | |
1768 | long timeo; | |
1769 | ||
1770 | skb = netlink_trim(skb, gfp_any()); | |
1771 | ||
1772 | timeo = sock_sndtimeo(ssk, nonblock); | |
1773 | retry: | |
15e47304 | 1774 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1775 | if (IS_ERR(sk)) { |
1776 | kfree_skb(skb); | |
1777 | return PTR_ERR(sk); | |
1778 | } | |
cd40b7d3 | 1779 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1780 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1781 | |
b1153f29 | 1782 | if (sk_filter(sk, skb)) { |
84874607 | 1783 | err = skb->len; |
b1153f29 SH |
1784 | kfree_skb(skb); |
1785 | sock_put(sk); | |
1786 | return err; | |
1787 | } | |
1788 | ||
9457afee | 1789 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1790 | if (err == 1) |
1791 | goto retry; | |
1792 | if (err) | |
1793 | return err; | |
1794 | ||
7ee015e0 | 1795 | return netlink_sendskb(sk, skb); |
1da177e4 | 1796 | } |
6ac552fd | 1797 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1798 | |
f9c22888 PM |
1799 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1800 | u32 dst_portid, gfp_t gfp_mask) | |
1801 | { | |
1802 | #ifdef CONFIG_NETLINK_MMAP | |
1803 | struct sock *sk = NULL; | |
1804 | struct sk_buff *skb; | |
1805 | struct netlink_ring *ring; | |
1806 | struct nl_mmap_hdr *hdr; | |
1807 | unsigned int maxlen; | |
1808 | ||
1809 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1810 | if (IS_ERR(sk)) | |
1811 | goto out; | |
1812 | ||
1813 | ring = &nlk_sk(sk)->rx_ring; | |
1814 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1815 | if (ring->pg_vec == NULL) | |
1816 | goto out_put; | |
1817 | ||
aae9f0e2 TG |
1818 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1819 | goto out_put; | |
1820 | ||
f9c22888 PM |
1821 | skb = alloc_skb_head(gfp_mask); |
1822 | if (skb == NULL) | |
1823 | goto err1; | |
1824 | ||
1825 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1826 | /* check again under lock */ | |
1827 | if (ring->pg_vec == NULL) | |
1828 | goto out_free; | |
1829 | ||
aae9f0e2 | 1830 | /* check again under lock */ |
f9c22888 PM |
1831 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1832 | if (maxlen < size) | |
1833 | goto out_free; | |
1834 | ||
1835 | netlink_forward_ring(ring); | |
1836 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1837 | if (hdr == NULL) | |
1838 | goto err2; | |
1839 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1840 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1841 | atomic_inc(&ring->pending); | |
1842 | netlink_increment_head(ring); | |
1843 | ||
1844 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1845 | return skb; | |
1846 | ||
1847 | err2: | |
1848 | kfree_skb(skb); | |
1849 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1850 | netlink_overrun(sk); |
f9c22888 PM |
1851 | err1: |
1852 | sock_put(sk); | |
1853 | return NULL; | |
1854 | ||
1855 | out_free: | |
1856 | kfree_skb(skb); | |
1857 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1858 | out_put: | |
1859 | sock_put(sk); | |
1860 | out: | |
1861 | #endif | |
1862 | return alloc_skb(size, gfp_mask); | |
1863 | } | |
1864 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1865 | ||
4277a083 PM |
1866 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1867 | { | |
1868 | int res = 0; | |
5c398dc8 | 1869 | struct listeners *listeners; |
4277a083 | 1870 | |
aed81560 | 1871 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1872 | |
1873 | rcu_read_lock(); | |
1874 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1875 | ||
6d772ac5 | 1876 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1877 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1878 | |
1879 | rcu_read_unlock(); | |
1880 | ||
4277a083 PM |
1881 | return res; |
1882 | } | |
1883 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1884 | ||
b57ef81f | 1885 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1886 | { |
1887 | struct netlink_sock *nlk = nlk_sk(sk); | |
1888 | ||
1889 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1890 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1891 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1892 | __netlink_sendskb(sk, skb); |
2c645800 | 1893 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1894 | } |
1895 | return -1; | |
1896 | } | |
1897 | ||
1898 | struct netlink_broadcast_data { | |
1899 | struct sock *exclude_sk; | |
b4b51029 | 1900 | struct net *net; |
15e47304 | 1901 | u32 portid; |
1da177e4 LT |
1902 | u32 group; |
1903 | int failure; | |
ff491a73 | 1904 | int delivery_failure; |
1da177e4 LT |
1905 | int congested; |
1906 | int delivered; | |
7d877f3b | 1907 | gfp_t allocation; |
1da177e4 | 1908 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1909 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1910 | void *tx_data; | |
1da177e4 LT |
1911 | }; |
1912 | ||
46c9521f RR |
1913 | static void do_one_broadcast(struct sock *sk, |
1914 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1915 | { |
1916 | struct netlink_sock *nlk = nlk_sk(sk); | |
1917 | int val; | |
1918 | ||
1919 | if (p->exclude_sk == sk) | |
46c9521f | 1920 | return; |
1da177e4 | 1921 | |
15e47304 | 1922 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1923 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1924 | return; |
1da177e4 | 1925 | |
878628fb | 1926 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1927 | return; |
b4b51029 | 1928 | |
1da177e4 LT |
1929 | if (p->failure) { |
1930 | netlink_overrun(sk); | |
46c9521f | 1931 | return; |
1da177e4 LT |
1932 | } |
1933 | ||
1934 | sock_hold(sk); | |
1935 | if (p->skb2 == NULL) { | |
68acc024 | 1936 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1937 | p->skb2 = skb_clone(p->skb, p->allocation); |
1938 | } else { | |
68acc024 TC |
1939 | p->skb2 = skb_get(p->skb); |
1940 | /* | |
1941 | * skb ownership may have been set when | |
1942 | * delivered to a previous socket. | |
1943 | */ | |
1944 | skb_orphan(p->skb2); | |
1da177e4 LT |
1945 | } |
1946 | } | |
1947 | if (p->skb2 == NULL) { | |
1948 | netlink_overrun(sk); | |
1949 | /* Clone failed. Notify ALL listeners. */ | |
1950 | p->failure = 1; | |
be0c22a4 PNA |
1951 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1952 | p->delivery_failure = 1; | |
910a7e90 EB |
1953 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1954 | kfree_skb(p->skb2); | |
1955 | p->skb2 = NULL; | |
b1153f29 SH |
1956 | } else if (sk_filter(sk, p->skb2)) { |
1957 | kfree_skb(p->skb2); | |
1958 | p->skb2 = NULL; | |
1da177e4 LT |
1959 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1960 | netlink_overrun(sk); | |
be0c22a4 PNA |
1961 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1962 | p->delivery_failure = 1; | |
1da177e4 LT |
1963 | } else { |
1964 | p->congested |= val; | |
1965 | p->delivered = 1; | |
1966 | p->skb2 = NULL; | |
1967 | } | |
1968 | sock_put(sk); | |
1da177e4 LT |
1969 | } |
1970 | ||
15e47304 | 1971 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1972 | u32 group, gfp_t allocation, |
1973 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1974 | void *filter_data) | |
1da177e4 | 1975 | { |
3b1e0a65 | 1976 | struct net *net = sock_net(ssk); |
1da177e4 | 1977 | struct netlink_broadcast_data info; |
1da177e4 LT |
1978 | struct sock *sk; |
1979 | ||
1980 | skb = netlink_trim(skb, allocation); | |
1981 | ||
1982 | info.exclude_sk = ssk; | |
b4b51029 | 1983 | info.net = net; |
15e47304 | 1984 | info.portid = portid; |
1da177e4 LT |
1985 | info.group = group; |
1986 | info.failure = 0; | |
ff491a73 | 1987 | info.delivery_failure = 0; |
1da177e4 LT |
1988 | info.congested = 0; |
1989 | info.delivered = 0; | |
1990 | info.allocation = allocation; | |
1991 | info.skb = skb; | |
1992 | info.skb2 = NULL; | |
910a7e90 EB |
1993 | info.tx_filter = filter; |
1994 | info.tx_data = filter_data; | |
1da177e4 LT |
1995 | |
1996 | /* While we sleep in clone, do not allow to change socket list */ | |
1997 | ||
1998 | netlink_lock_table(); | |
1999 | ||
b67bfe0d | 2000 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
2001 | do_one_broadcast(sk, &info); |
2002 | ||
70d4bf6d | 2003 | consume_skb(skb); |
aa1c6a6f | 2004 | |
1da177e4 LT |
2005 | netlink_unlock_table(); |
2006 | ||
70d4bf6d NH |
2007 | if (info.delivery_failure) { |
2008 | kfree_skb(info.skb2); | |
ff491a73 | 2009 | return -ENOBUFS; |
658cb354 ED |
2010 | } |
2011 | consume_skb(info.skb2); | |
ff491a73 | 2012 | |
1da177e4 LT |
2013 | if (info.delivered) { |
2014 | if (info.congested && (allocation & __GFP_WAIT)) | |
2015 | yield(); | |
2016 | return 0; | |
2017 | } | |
1da177e4 LT |
2018 | return -ESRCH; |
2019 | } | |
910a7e90 EB |
2020 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2021 | ||
15e47304 | 2022 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2023 | u32 group, gfp_t allocation) |
2024 | { | |
15e47304 | 2025 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2026 | NULL, NULL); |
2027 | } | |
6ac552fd | 2028 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2029 | |
2030 | struct netlink_set_err_data { | |
2031 | struct sock *exclude_sk; | |
15e47304 | 2032 | u32 portid; |
1da177e4 LT |
2033 | u32 group; |
2034 | int code; | |
2035 | }; | |
2036 | ||
b57ef81f | 2037 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2038 | { |
2039 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2040 | int ret = 0; |
1da177e4 LT |
2041 | |
2042 | if (sk == p->exclude_sk) | |
2043 | goto out; | |
2044 | ||
09ad9bc7 | 2045 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2046 | goto out; |
2047 | ||
15e47304 | 2048 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2049 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2050 | goto out; |
2051 | ||
1a50307b PNA |
2052 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2053 | ret = 1; | |
2054 | goto out; | |
2055 | } | |
2056 | ||
1da177e4 LT |
2057 | sk->sk_err = p->code; |
2058 | sk->sk_error_report(sk); | |
2059 | out: | |
1a50307b | 2060 | return ret; |
1da177e4 LT |
2061 | } |
2062 | ||
4843b93c PNA |
2063 | /** |
2064 | * netlink_set_err - report error to broadcast listeners | |
2065 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2066 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2067 | * @group: the broadcast group that will notice the error |
4843b93c | 2068 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2069 | * |
2070 | * This function returns the number of broadcast listeners that have set the | |
2071 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2072 | */ |
15e47304 | 2073 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2074 | { |
2075 | struct netlink_set_err_data info; | |
1da177e4 | 2076 | struct sock *sk; |
1a50307b | 2077 | int ret = 0; |
1da177e4 LT |
2078 | |
2079 | info.exclude_sk = ssk; | |
15e47304 | 2080 | info.portid = portid; |
1da177e4 | 2081 | info.group = group; |
4843b93c PNA |
2082 | /* sk->sk_err wants a positive error value */ |
2083 | info.code = -code; | |
1da177e4 LT |
2084 | |
2085 | read_lock(&nl_table_lock); | |
2086 | ||
b67bfe0d | 2087 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2088 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2089 | |
2090 | read_unlock(&nl_table_lock); | |
1a50307b | 2091 | return ret; |
1da177e4 | 2092 | } |
dd5b6ce6 | 2093 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2094 | |
84659eb5 JB |
2095 | /* must be called with netlink table grabbed */ |
2096 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2097 | unsigned int group, | |
2098 | int is_new) | |
2099 | { | |
2100 | int old, new = !!is_new, subscriptions; | |
2101 | ||
2102 | old = test_bit(group - 1, nlk->groups); | |
2103 | subscriptions = nlk->subscriptions - old + new; | |
2104 | if (new) | |
2105 | __set_bit(group - 1, nlk->groups); | |
2106 | else | |
2107 | __clear_bit(group - 1, nlk->groups); | |
2108 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2109 | netlink_update_listeners(&nlk->sk); | |
2110 | } | |
2111 | ||
9a4595bc | 2112 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2113 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2114 | { |
2115 | struct sock *sk = sock->sk; | |
2116 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2117 | unsigned int val = 0; |
2118 | int err; | |
9a4595bc PM |
2119 | |
2120 | if (level != SOL_NETLINK) | |
2121 | return -ENOPROTOOPT; | |
2122 | ||
ccdfcc39 PM |
2123 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2124 | optlen >= sizeof(int) && | |
eb496534 | 2125 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2126 | return -EFAULT; |
2127 | ||
2128 | switch (optname) { | |
2129 | case NETLINK_PKTINFO: | |
2130 | if (val) | |
2131 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2132 | else | |
2133 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2134 | err = 0; | |
2135 | break; | |
2136 | case NETLINK_ADD_MEMBERSHIP: | |
2137 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2138 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2139 | return -EPERM; |
b4ff4f04 JB |
2140 | err = netlink_realloc_groups(sk); |
2141 | if (err) | |
2142 | return err; | |
9a4595bc PM |
2143 | if (!val || val - 1 >= nlk->ngroups) |
2144 | return -EINVAL; | |
7774d5e0 | 2145 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
4f520900 RGB |
2146 | err = nlk->netlink_bind(val); |
2147 | if (err) | |
2148 | return err; | |
2149 | } | |
9a4595bc | 2150 | netlink_table_grab(); |
84659eb5 JB |
2151 | netlink_update_socket_mc(nlk, val, |
2152 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2153 | netlink_table_ungrab(); |
7774d5e0 RGB |
2154 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
2155 | nlk->netlink_unbind(val); | |
03292745 | 2156 | |
9a4595bc PM |
2157 | err = 0; |
2158 | break; | |
2159 | } | |
be0c22a4 PNA |
2160 | case NETLINK_BROADCAST_ERROR: |
2161 | if (val) | |
2162 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2163 | else | |
2164 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2165 | err = 0; | |
2166 | break; | |
38938bfe PNA |
2167 | case NETLINK_NO_ENOBUFS: |
2168 | if (val) { | |
2169 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2170 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2171 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2172 | } else { |
38938bfe | 2173 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2174 | } |
38938bfe PNA |
2175 | err = 0; |
2176 | break; | |
ccdfcc39 PM |
2177 | #ifdef CONFIG_NETLINK_MMAP |
2178 | case NETLINK_RX_RING: | |
2179 | case NETLINK_TX_RING: { | |
2180 | struct nl_mmap_req req; | |
2181 | ||
2182 | /* Rings might consume more memory than queue limits, require | |
2183 | * CAP_NET_ADMIN. | |
2184 | */ | |
2185 | if (!capable(CAP_NET_ADMIN)) | |
2186 | return -EPERM; | |
2187 | if (optlen < sizeof(req)) | |
2188 | return -EINVAL; | |
2189 | if (copy_from_user(&req, optval, sizeof(req))) | |
2190 | return -EFAULT; | |
2191 | err = netlink_set_ring(sk, &req, false, | |
2192 | optname == NETLINK_TX_RING); | |
2193 | break; | |
2194 | } | |
2195 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2196 | default: |
2197 | err = -ENOPROTOOPT; | |
2198 | } | |
2199 | return err; | |
2200 | } | |
2201 | ||
2202 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2203 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2204 | { |
2205 | struct sock *sk = sock->sk; | |
2206 | struct netlink_sock *nlk = nlk_sk(sk); | |
2207 | int len, val, err; | |
2208 | ||
2209 | if (level != SOL_NETLINK) | |
2210 | return -ENOPROTOOPT; | |
2211 | ||
2212 | if (get_user(len, optlen)) | |
2213 | return -EFAULT; | |
2214 | if (len < 0) | |
2215 | return -EINVAL; | |
2216 | ||
2217 | switch (optname) { | |
2218 | case NETLINK_PKTINFO: | |
2219 | if (len < sizeof(int)) | |
2220 | return -EINVAL; | |
2221 | len = sizeof(int); | |
2222 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2223 | if (put_user(len, optlen) || |
2224 | put_user(val, optval)) | |
2225 | return -EFAULT; | |
9a4595bc PM |
2226 | err = 0; |
2227 | break; | |
be0c22a4 PNA |
2228 | case NETLINK_BROADCAST_ERROR: |
2229 | if (len < sizeof(int)) | |
2230 | return -EINVAL; | |
2231 | len = sizeof(int); | |
2232 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2233 | if (put_user(len, optlen) || | |
2234 | put_user(val, optval)) | |
2235 | return -EFAULT; | |
2236 | err = 0; | |
2237 | break; | |
38938bfe PNA |
2238 | case NETLINK_NO_ENOBUFS: |
2239 | if (len < sizeof(int)) | |
2240 | return -EINVAL; | |
2241 | len = sizeof(int); | |
2242 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2243 | if (put_user(len, optlen) || | |
2244 | put_user(val, optval)) | |
2245 | return -EFAULT; | |
2246 | err = 0; | |
2247 | break; | |
9a4595bc PM |
2248 | default: |
2249 | err = -ENOPROTOOPT; | |
2250 | } | |
2251 | return err; | |
2252 | } | |
2253 | ||
2254 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2255 | { | |
2256 | struct nl_pktinfo info; | |
2257 | ||
2258 | info.group = NETLINK_CB(skb).dst_group; | |
2259 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2260 | } | |
2261 | ||
1da177e4 LT |
2262 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2263 | struct msghdr *msg, size_t len) | |
2264 | { | |
2265 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2266 | struct sock *sk = sock->sk; | |
2267 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2268 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2269 | u32 dst_portid; |
d629b836 | 2270 | u32 dst_group; |
1da177e4 LT |
2271 | struct sk_buff *skb; |
2272 | int err; | |
2273 | struct scm_cookie scm; | |
2d7a85f4 | 2274 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2275 | |
2276 | if (msg->msg_flags&MSG_OOB) | |
2277 | return -EOPNOTSUPP; | |
2278 | ||
16e57262 | 2279 | if (NULL == siocb->scm) |
1da177e4 | 2280 | siocb->scm = &scm; |
16e57262 | 2281 | |
e0e3cea4 | 2282 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2283 | if (err < 0) |
2284 | return err; | |
2285 | ||
2286 | if (msg->msg_namelen) { | |
b47030c7 | 2287 | err = -EINVAL; |
1da177e4 | 2288 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2289 | goto out; |
15e47304 | 2290 | dst_portid = addr->nl_pid; |
d629b836 | 2291 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2292 | err = -EPERM; |
15e47304 | 2293 | if ((dst_group || dst_portid) && |
5187cd05 | 2294 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2295 | goto out; |
2d7a85f4 | 2296 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2297 | } else { |
15e47304 | 2298 | dst_portid = nlk->dst_portid; |
d629b836 | 2299 | dst_group = nlk->dst_group; |
1da177e4 LT |
2300 | } |
2301 | ||
15e47304 | 2302 | if (!nlk->portid) { |
1da177e4 LT |
2303 | err = netlink_autobind(sock); |
2304 | if (err) | |
2305 | goto out; | |
2306 | } | |
2307 | ||
5fd96123 PM |
2308 | if (netlink_tx_is_mmaped(sk) && |
2309 | msg->msg_iov->iov_base == NULL) { | |
2310 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2311 | siocb); | |
2312 | goto out; | |
2313 | } | |
2314 | ||
1da177e4 LT |
2315 | err = -EMSGSIZE; |
2316 | if (len > sk->sk_sndbuf - 32) | |
2317 | goto out; | |
2318 | err = -ENOBUFS; | |
3a36515f | 2319 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2320 | if (skb == NULL) |
1da177e4 LT |
2321 | goto out; |
2322 | ||
15e47304 | 2323 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2324 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2325 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2326 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2327 | |
1da177e4 | 2328 | err = -EFAULT; |
6ac552fd | 2329 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2330 | kfree_skb(skb); |
2331 | goto out; | |
2332 | } | |
2333 | ||
2334 | err = security_netlink_send(sk, skb); | |
2335 | if (err) { | |
2336 | kfree_skb(skb); | |
2337 | goto out; | |
2338 | } | |
2339 | ||
d629b836 | 2340 | if (dst_group) { |
1da177e4 | 2341 | atomic_inc(&skb->users); |
15e47304 | 2342 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2343 | } |
15e47304 | 2344 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2345 | |
2346 | out: | |
b47030c7 | 2347 | scm_destroy(siocb->scm); |
1da177e4 LT |
2348 | return err; |
2349 | } | |
2350 | ||
2351 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2352 | struct msghdr *msg, size_t len, | |
2353 | int flags) | |
2354 | { | |
2355 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2356 | struct scm_cookie scm; | |
2357 | struct sock *sk = sock->sk; | |
2358 | struct netlink_sock *nlk = nlk_sk(sk); | |
2359 | int noblock = flags&MSG_DONTWAIT; | |
2360 | size_t copied; | |
68d6ac6d | 2361 | struct sk_buff *skb, *data_skb; |
b44d211e | 2362 | int err, ret; |
1da177e4 LT |
2363 | |
2364 | if (flags&MSG_OOB) | |
2365 | return -EOPNOTSUPP; | |
2366 | ||
2367 | copied = 0; | |
2368 | ||
6ac552fd PM |
2369 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2370 | if (skb == NULL) | |
1da177e4 LT |
2371 | goto out; |
2372 | ||
68d6ac6d JB |
2373 | data_skb = skb; |
2374 | ||
1dacc76d JB |
2375 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2376 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2377 | /* |
68d6ac6d JB |
2378 | * If this skb has a frag_list, then here that means that we |
2379 | * will have to use the frag_list skb's data for compat tasks | |
2380 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2381 | * |
68d6ac6d JB |
2382 | * If we need to send the compat skb, assign it to the |
2383 | * 'data_skb' variable so that it will be used below for data | |
2384 | * copying. We keep 'skb' for everything else, including | |
2385 | * freeing both later. | |
1dacc76d | 2386 | */ |
68d6ac6d JB |
2387 | if (flags & MSG_CMSG_COMPAT) |
2388 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2389 | } |
2390 | #endif | |
2391 | ||
9063e21f ED |
2392 | /* Record the max length of recvmsg() calls for future allocations */ |
2393 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2394 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2395 | 16384); | |
2396 | ||
68d6ac6d | 2397 | copied = data_skb->len; |
1da177e4 LT |
2398 | if (len < copied) { |
2399 | msg->msg_flags |= MSG_TRUNC; | |
2400 | copied = len; | |
2401 | } | |
2402 | ||
68d6ac6d | 2403 | skb_reset_transport_header(data_skb); |
51f3d02b | 2404 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2405 | |
2406 | if (msg->msg_name) { | |
342dfc30 | 2407 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2408 | addr->nl_family = AF_NETLINK; |
2409 | addr->nl_pad = 0; | |
15e47304 | 2410 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2411 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2412 | msg->msg_namelen = sizeof(*addr); |
2413 | } | |
2414 | ||
cc9a06cd PM |
2415 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2416 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2417 | ||
1da177e4 LT |
2418 | if (NULL == siocb->scm) { |
2419 | memset(&scm, 0, sizeof(scm)); | |
2420 | siocb->scm = &scm; | |
2421 | } | |
2422 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2423 | if (flags & MSG_TRUNC) |
68d6ac6d | 2424 | copied = data_skb->len; |
daa3766e | 2425 | |
1da177e4 LT |
2426 | skb_free_datagram(sk, skb); |
2427 | ||
16b304f3 PS |
2428 | if (nlk->cb_running && |
2429 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2430 | ret = netlink_dump(sk); |
2431 | if (ret) { | |
ac30ef83 | 2432 | sk->sk_err = -ret; |
b44d211e AV |
2433 | sk->sk_error_report(sk); |
2434 | } | |
2435 | } | |
1da177e4 LT |
2436 | |
2437 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2438 | out: |
2439 | netlink_rcv_wake(sk); | |
2440 | return err ? : copied; | |
2441 | } | |
2442 | ||
676d2369 | 2443 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2444 | { |
cd40b7d3 | 2445 | BUG(); |
1da177e4 LT |
2446 | } |
2447 | ||
2448 | /* | |
746fac4d | 2449 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2450 | * complete set of kernel non-blocking support for message |
2451 | * queueing. | |
2452 | */ | |
2453 | ||
2454 | struct sock * | |
9f00d977 PNA |
2455 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2456 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2457 | { |
2458 | struct socket *sock; | |
2459 | struct sock *sk; | |
77247bbb | 2460 | struct netlink_sock *nlk; |
5c398dc8 | 2461 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2462 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2463 | unsigned int groups; | |
1da177e4 | 2464 | |
fab2caf6 | 2465 | BUG_ON(!nl_table); |
1da177e4 | 2466 | |
6ac552fd | 2467 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2468 | return NULL; |
2469 | ||
2470 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2471 | return NULL; | |
2472 | ||
23fe1866 PE |
2473 | /* |
2474 | * We have to just have a reference on the net from sk, but don't | |
2475 | * get_net it. Besides, we cannot get and then put the net here. | |
2476 | * So we create one inside init_net and the move it to net. | |
2477 | */ | |
2478 | ||
2479 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2480 | goto out_sock_release_nosk; | |
2481 | ||
2482 | sk = sock->sk; | |
edf02087 | 2483 | sk_change_net(sk, net); |
4fdb3bb7 | 2484 | |
a31f2d17 | 2485 | if (!cfg || cfg->groups < 32) |
4277a083 | 2486 | groups = 32; |
a31f2d17 PNA |
2487 | else |
2488 | groups = cfg->groups; | |
4277a083 | 2489 | |
5c398dc8 | 2490 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2491 | if (!listeners) |
2492 | goto out_sock_release; | |
2493 | ||
1da177e4 | 2494 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2495 | if (cfg && cfg->input) |
2496 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2497 | |
b4b51029 | 2498 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2499 | goto out_sock_release; |
4fdb3bb7 | 2500 | |
77247bbb PM |
2501 | nlk = nlk_sk(sk); |
2502 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2503 | |
4fdb3bb7 | 2504 | netlink_table_grab(); |
b4b51029 EB |
2505 | if (!nl_table[unit].registered) { |
2506 | nl_table[unit].groups = groups; | |
5c398dc8 | 2507 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2508 | nl_table[unit].cb_mutex = cb_mutex; |
2509 | nl_table[unit].module = module; | |
9785e10a PNA |
2510 | if (cfg) { |
2511 | nl_table[unit].bind = cfg->bind; | |
2512 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2513 | if (cfg->compare) |
2514 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2515 | } |
b4b51029 | 2516 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2517 | } else { |
2518 | kfree(listeners); | |
869e58f8 | 2519 | nl_table[unit].registered++; |
b4b51029 | 2520 | } |
4fdb3bb7 | 2521 | netlink_table_ungrab(); |
77247bbb PM |
2522 | return sk; |
2523 | ||
4fdb3bb7 | 2524 | out_sock_release: |
4277a083 | 2525 | kfree(listeners); |
9dfbec1f | 2526 | netlink_kernel_release(sk); |
23fe1866 PE |
2527 | return NULL; |
2528 | ||
2529 | out_sock_release_nosk: | |
4fdb3bb7 | 2530 | sock_release(sock); |
77247bbb | 2531 | return NULL; |
1da177e4 | 2532 | } |
9f00d977 | 2533 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2534 | |
2535 | void | |
2536 | netlink_kernel_release(struct sock *sk) | |
2537 | { | |
edf02087 | 2538 | sk_release_kernel(sk); |
b7c6ba6e DL |
2539 | } |
2540 | EXPORT_SYMBOL(netlink_kernel_release); | |
2541 | ||
d136f1bd | 2542 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2543 | { |
5c398dc8 | 2544 | struct listeners *new, *old; |
b4ff4f04 | 2545 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2546 | |
2547 | if (groups < 32) | |
2548 | groups = 32; | |
2549 | ||
b4ff4f04 | 2550 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2551 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2552 | if (!new) | |
d136f1bd | 2553 | return -ENOMEM; |
6d772ac5 | 2554 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2555 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2556 | rcu_assign_pointer(tbl->listeners, new); | |
2557 | ||
37b6b935 | 2558 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2559 | } |
2560 | tbl->groups = groups; | |
2561 | ||
d136f1bd JB |
2562 | return 0; |
2563 | } | |
2564 | ||
2565 | /** | |
2566 | * netlink_change_ngroups - change number of multicast groups | |
2567 | * | |
2568 | * This changes the number of multicast groups that are available | |
2569 | * on a certain netlink family. Note that it is not possible to | |
2570 | * change the number of groups to below 32. Also note that it does | |
2571 | * not implicitly call netlink_clear_multicast_users() when the | |
2572 | * number of groups is reduced. | |
2573 | * | |
2574 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2575 | * @groups: The new number of groups. | |
2576 | */ | |
2577 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2578 | { | |
2579 | int err; | |
2580 | ||
2581 | netlink_table_grab(); | |
2582 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2583 | netlink_table_ungrab(); |
d136f1bd | 2584 | |
b4ff4f04 JB |
2585 | return err; |
2586 | } | |
b4ff4f04 | 2587 | |
b8273570 JB |
2588 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2589 | { | |
2590 | struct sock *sk; | |
b8273570 JB |
2591 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2592 | ||
b67bfe0d | 2593 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2594 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2595 | } | |
2596 | ||
a46621a3 | 2597 | struct nlmsghdr * |
15e47304 | 2598 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2599 | { |
2600 | struct nlmsghdr *nlh; | |
573ce260 | 2601 | int size = nlmsg_msg_size(len); |
a46621a3 | 2602 | |
23b45672 | 2603 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2604 | nlh->nlmsg_type = type; |
2605 | nlh->nlmsg_len = size; | |
2606 | nlh->nlmsg_flags = flags; | |
15e47304 | 2607 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2608 | nlh->nlmsg_seq = seq; |
2609 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2610 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2611 | return nlh; |
2612 | } | |
2613 | EXPORT_SYMBOL(__nlmsg_put); | |
2614 | ||
1da177e4 LT |
2615 | /* |
2616 | * It looks a bit ugly. | |
2617 | * It would be better to create kernel thread. | |
2618 | */ | |
2619 | ||
2620 | static int netlink_dump(struct sock *sk) | |
2621 | { | |
2622 | struct netlink_sock *nlk = nlk_sk(sk); | |
2623 | struct netlink_callback *cb; | |
c7ac8679 | 2624 | struct sk_buff *skb = NULL; |
1da177e4 | 2625 | struct nlmsghdr *nlh; |
bf8b79e4 | 2626 | int len, err = -ENOBUFS; |
c7ac8679 | 2627 | int alloc_size; |
1da177e4 | 2628 | |
af65bdfc | 2629 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2630 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2631 | err = -EINVAL; |
2632 | goto errout_skb; | |
1da177e4 LT |
2633 | } |
2634 | ||
16b304f3 | 2635 | cb = &nlk->cb; |
c7ac8679 GR |
2636 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2637 | ||
f9c22888 PM |
2638 | if (!netlink_rx_is_mmaped(sk) && |
2639 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2640 | goto errout_skb; | |
9063e21f ED |
2641 | |
2642 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2643 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2644 | * to reduce number of system calls on dump operations, if user | |
2645 | * ever provided a big enough buffer. | |
2646 | */ | |
2647 | if (alloc_size < nlk->max_recvmsg_len) { | |
2648 | skb = netlink_alloc_skb(sk, | |
2649 | nlk->max_recvmsg_len, | |
2650 | nlk->portid, | |
2651 | GFP_KERNEL | | |
2652 | __GFP_NOWARN | | |
2653 | __GFP_NORETRY); | |
2654 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2655 | if (skb) | |
2656 | skb_reserve(skb, skb_tailroom(skb) - | |
2657 | nlk->max_recvmsg_len); | |
2658 | } | |
2659 | if (!skb) | |
2660 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2661 | GFP_KERNEL); | |
c7ac8679 | 2662 | if (!skb) |
c63d6ea3 | 2663 | goto errout_skb; |
f9c22888 | 2664 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2665 | |
1da177e4 LT |
2666 | len = cb->dump(skb, cb); |
2667 | ||
2668 | if (len > 0) { | |
af65bdfc | 2669 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2670 | |
2671 | if (sk_filter(sk, skb)) | |
2672 | kfree_skb(skb); | |
4a7e7c2a ED |
2673 | else |
2674 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2675 | return 0; |
2676 | } | |
2677 | ||
bf8b79e4 TG |
2678 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2679 | if (!nlh) | |
2680 | goto errout_skb; | |
2681 | ||
670dc283 JB |
2682 | nl_dump_check_consistent(cb, nlh); |
2683 | ||
bf8b79e4 TG |
2684 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2685 | ||
b1153f29 SH |
2686 | if (sk_filter(sk, skb)) |
2687 | kfree_skb(skb); | |
4a7e7c2a ED |
2688 | else |
2689 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2690 | |
a8f74b22 TG |
2691 | if (cb->done) |
2692 | cb->done(cb); | |
1da177e4 | 2693 | |
16b304f3 PS |
2694 | nlk->cb_running = false; |
2695 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2696 | module_put(cb->module); |
16b304f3 | 2697 | consume_skb(cb->skb); |
1da177e4 | 2698 | return 0; |
1797754e | 2699 | |
bf8b79e4 | 2700 | errout_skb: |
af65bdfc | 2701 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2702 | kfree_skb(skb); |
bf8b79e4 | 2703 | return err; |
1da177e4 LT |
2704 | } |
2705 | ||
6dc878a8 G |
2706 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2707 | const struct nlmsghdr *nlh, | |
2708 | struct netlink_dump_control *control) | |
1da177e4 LT |
2709 | { |
2710 | struct netlink_callback *cb; | |
2711 | struct sock *sk; | |
2712 | struct netlink_sock *nlk; | |
b44d211e | 2713 | int ret; |
1da177e4 | 2714 | |
f9c22888 PM |
2715 | /* Memory mapped dump requests need to be copied to avoid looping |
2716 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2717 | * a reference to the skb. | |
2718 | */ | |
2719 | if (netlink_skb_is_mmaped(skb)) { | |
2720 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2721 | if (skb == NULL) |
f9c22888 | 2722 | return -ENOBUFS; |
f9c22888 PM |
2723 | } else |
2724 | atomic_inc(&skb->users); | |
2725 | ||
15e47304 | 2726 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2727 | if (sk == NULL) { |
16b304f3 PS |
2728 | ret = -ECONNREFUSED; |
2729 | goto error_free; | |
1da177e4 | 2730 | } |
6dc878a8 | 2731 | |
16b304f3 | 2732 | nlk = nlk_sk(sk); |
af65bdfc | 2733 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2734 | /* A dump is in progress... */ |
16b304f3 | 2735 | if (nlk->cb_running) { |
6dc878a8 | 2736 | ret = -EBUSY; |
16b304f3 | 2737 | goto error_unlock; |
1da177e4 | 2738 | } |
6dc878a8 | 2739 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2740 | if (!try_module_get(control->module)) { |
6dc878a8 | 2741 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2742 | goto error_unlock; |
6dc878a8 G |
2743 | } |
2744 | ||
16b304f3 PS |
2745 | cb = &nlk->cb; |
2746 | memset(cb, 0, sizeof(*cb)); | |
2747 | cb->dump = control->dump; | |
2748 | cb->done = control->done; | |
2749 | cb->nlh = nlh; | |
2750 | cb->data = control->data; | |
2751 | cb->module = control->module; | |
2752 | cb->min_dump_alloc = control->min_dump_alloc; | |
2753 | cb->skb = skb; | |
2754 | ||
2755 | nlk->cb_running = true; | |
2756 | ||
af65bdfc | 2757 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2758 | |
b44d211e | 2759 | ret = netlink_dump(sk); |
1da177e4 | 2760 | sock_put(sk); |
5c58298c | 2761 | |
b44d211e AV |
2762 | if (ret) |
2763 | return ret; | |
2764 | ||
5c58298c DL |
2765 | /* We successfully started a dump, by returning -EINTR we |
2766 | * signal not to send ACK even if it was requested. | |
2767 | */ | |
2768 | return -EINTR; | |
16b304f3 PS |
2769 | |
2770 | error_unlock: | |
2771 | sock_put(sk); | |
2772 | mutex_unlock(nlk->cb_mutex); | |
2773 | error_free: | |
2774 | kfree_skb(skb); | |
2775 | return ret; | |
1da177e4 | 2776 | } |
6dc878a8 | 2777 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2778 | |
2779 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2780 | { | |
2781 | struct sk_buff *skb; | |
2782 | struct nlmsghdr *rep; | |
2783 | struct nlmsgerr *errmsg; | |
339bf98f | 2784 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2785 | |
339bf98f TG |
2786 | /* error messages get the original request appened */ |
2787 | if (err) | |
2788 | payload += nlmsg_len(nlh); | |
1da177e4 | 2789 | |
f9c22888 PM |
2790 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2791 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2792 | if (!skb) { |
2793 | struct sock *sk; | |
2794 | ||
3b1e0a65 | 2795 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2796 | in_skb->sk->sk_protocol, |
15e47304 | 2797 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2798 | if (sk) { |
2799 | sk->sk_err = ENOBUFS; | |
2800 | sk->sk_error_report(sk); | |
2801 | sock_put(sk); | |
2802 | } | |
2803 | return; | |
2804 | } | |
2805 | ||
15e47304 | 2806 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2807 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2808 | errmsg = nlmsg_data(rep); |
1da177e4 | 2809 | errmsg->error = err; |
bf8b79e4 | 2810 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2811 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2812 | } |
6ac552fd | 2813 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2814 | |
cd40b7d3 | 2815 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2816 | struct nlmsghdr *)) |
82ace47a | 2817 | { |
82ace47a TG |
2818 | struct nlmsghdr *nlh; |
2819 | int err; | |
2820 | ||
2821 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2822 | int msglen; |
2823 | ||
b529ccf2 | 2824 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2825 | err = 0; |
82ace47a | 2826 | |
ad8e4b75 | 2827 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2828 | return 0; |
2829 | ||
d35b6856 TG |
2830 | /* Only requests are handled by the kernel */ |
2831 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2832 | goto ack; |
45e7ae7f TG |
2833 | |
2834 | /* Skip control messages */ | |
2835 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2836 | goto ack; |
d35b6856 | 2837 | |
1d00a4eb | 2838 | err = cb(skb, nlh); |
5c58298c DL |
2839 | if (err == -EINTR) |
2840 | goto skip; | |
2841 | ||
2842 | ack: | |
d35b6856 | 2843 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2844 | netlink_ack(skb, nlh, err); |
82ace47a | 2845 | |
5c58298c | 2846 | skip: |
6ac552fd | 2847 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2848 | if (msglen > skb->len) |
2849 | msglen = skb->len; | |
2850 | skb_pull(skb, msglen); | |
82ace47a TG |
2851 | } |
2852 | ||
2853 | return 0; | |
2854 | } | |
6ac552fd | 2855 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2856 | |
d387f6ad TG |
2857 | /** |
2858 | * nlmsg_notify - send a notification netlink message | |
2859 | * @sk: netlink socket to use | |
2860 | * @skb: notification message | |
15e47304 | 2861 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2862 | * @group: destination multicast group or 0 |
2863 | * @report: 1 to report back, 0 to disable | |
2864 | * @flags: allocation flags | |
2865 | */ | |
15e47304 | 2866 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2867 | unsigned int group, int report, gfp_t flags) |
2868 | { | |
2869 | int err = 0; | |
2870 | ||
2871 | if (group) { | |
15e47304 | 2872 | int exclude_portid = 0; |
d387f6ad TG |
2873 | |
2874 | if (report) { | |
2875 | atomic_inc(&skb->users); | |
15e47304 | 2876 | exclude_portid = portid; |
d387f6ad TG |
2877 | } |
2878 | ||
1ce85fe4 PNA |
2879 | /* errors reported via destination sk->sk_err, but propagate |
2880 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2881 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2882 | } |
2883 | ||
1ce85fe4 PNA |
2884 | if (report) { |
2885 | int err2; | |
2886 | ||
15e47304 | 2887 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2888 | if (!err || err == -ESRCH) |
2889 | err = err2; | |
2890 | } | |
d387f6ad TG |
2891 | |
2892 | return err; | |
2893 | } | |
6ac552fd | 2894 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2895 | |
1da177e4 LT |
2896 | #ifdef CONFIG_PROC_FS |
2897 | struct nl_seq_iter { | |
e372c414 | 2898 | struct seq_net_private p; |
1da177e4 LT |
2899 | int link; |
2900 | int hash_idx; | |
2901 | }; | |
2902 | ||
2903 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2904 | { | |
2905 | struct nl_seq_iter *iter = seq->private; | |
2906 | int i, j; | |
e341694e | 2907 | struct netlink_sock *nlk; |
1da177e4 | 2908 | struct sock *s; |
1da177e4 LT |
2909 | loff_t off = 0; |
2910 | ||
6ac552fd | 2911 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e | 2912 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2913 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
e341694e TG |
2914 | |
2915 | for (j = 0; j < tbl->size; j++) { | |
2916 | rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) { | |
2917 | s = (struct sock *)nlk; | |
1da177e4 | 2918 | |
1218854a | 2919 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2920 | continue; |
1da177e4 LT |
2921 | if (off == pos) { |
2922 | iter->link = i; | |
2923 | iter->hash_idx = j; | |
2924 | return s; | |
2925 | } | |
2926 | ++off; | |
2927 | } | |
2928 | } | |
2929 | } | |
2930 | return NULL; | |
2931 | } | |
2932 | ||
2933 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
78fd1d0a | 2934 | __acquires(nl_table_lock) __acquires(RCU) |
1da177e4 | 2935 | { |
78fd1d0a | 2936 | read_lock(&nl_table_lock); |
e341694e | 2937 | rcu_read_lock(); |
1da177e4 LT |
2938 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
2939 | } | |
2940 | ||
2941 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2942 | { | |
78fd1d0a | 2943 | struct rhashtable *ht; |
e341694e | 2944 | struct netlink_sock *nlk; |
1da177e4 | 2945 | struct nl_seq_iter *iter; |
da12c90e | 2946 | struct net *net; |
1da177e4 LT |
2947 | int i, j; |
2948 | ||
2949 | ++*pos; | |
2950 | ||
2951 | if (v == SEQ_START_TOKEN) | |
2952 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2953 | |
da12c90e | 2954 | net = seq_file_net(seq); |
b4b51029 | 2955 | iter = seq->private; |
e341694e TG |
2956 | nlk = v; |
2957 | ||
78fd1d0a TG |
2958 | i = iter->link; |
2959 | ht = &nl_table[i].hash; | |
2960 | rht_for_each_entry(nlk, nlk->node.next, ht, node) | |
e341694e TG |
2961 | if (net_eq(sock_net((struct sock *)nlk), net)) |
2962 | return nlk; | |
1da177e4 | 2963 | |
1da177e4 LT |
2964 | j = iter->hash_idx + 1; |
2965 | ||
2966 | do { | |
67a24ac1 | 2967 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
da12c90e | 2968 | |
e341694e | 2969 | for (; j < tbl->size; j++) { |
78fd1d0a | 2970 | rht_for_each_entry(nlk, tbl->buckets[j], ht, node) { |
e341694e TG |
2971 | if (net_eq(sock_net((struct sock *)nlk), net)) { |
2972 | iter->link = i; | |
2973 | iter->hash_idx = j; | |
2974 | return nlk; | |
2975 | } | |
1da177e4 LT |
2976 | } |
2977 | } | |
2978 | ||
2979 | j = 0; | |
2980 | } while (++i < MAX_LINKS); | |
2981 | ||
2982 | return NULL; | |
2983 | } | |
2984 | ||
2985 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
78fd1d0a | 2986 | __releases(RCU) __releases(nl_table_lock) |
1da177e4 | 2987 | { |
e341694e | 2988 | rcu_read_unlock(); |
78fd1d0a | 2989 | read_unlock(&nl_table_lock); |
1da177e4 LT |
2990 | } |
2991 | ||
2992 | ||
2993 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2994 | { | |
658cb354 | 2995 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2996 | seq_puts(seq, |
2997 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2998 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2999 | } else { |
1da177e4 LT |
3000 | struct sock *s = v; |
3001 | struct netlink_sock *nlk = nlk_sk(s); | |
3002 | ||
16b304f3 | 3003 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
3004 | s, |
3005 | s->sk_protocol, | |
15e47304 | 3006 | nlk->portid, |
513c2500 | 3007 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
3008 | sk_rmem_alloc_get(s), |
3009 | sk_wmem_alloc_get(s), | |
16b304f3 | 3010 | nlk->cb_running, |
38938bfe | 3011 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3012 | atomic_read(&s->sk_drops), |
3013 | sock_i_ino(s) | |
1da177e4 LT |
3014 | ); |
3015 | ||
3016 | } | |
3017 | return 0; | |
3018 | } | |
3019 | ||
56b3d975 | 3020 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3021 | .start = netlink_seq_start, |
3022 | .next = netlink_seq_next, | |
3023 | .stop = netlink_seq_stop, | |
3024 | .show = netlink_seq_show, | |
3025 | }; | |
3026 | ||
3027 | ||
3028 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3029 | { | |
e372c414 DL |
3030 | return seq_open_net(inode, file, &netlink_seq_ops, |
3031 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3032 | } |
3033 | ||
da7071d7 | 3034 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3035 | .owner = THIS_MODULE, |
3036 | .open = netlink_seq_open, | |
3037 | .read = seq_read, | |
3038 | .llseek = seq_lseek, | |
e372c414 | 3039 | .release = seq_release_net, |
1da177e4 LT |
3040 | }; |
3041 | ||
3042 | #endif | |
3043 | ||
3044 | int netlink_register_notifier(struct notifier_block *nb) | |
3045 | { | |
e041c683 | 3046 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3047 | } |
6ac552fd | 3048 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3049 | |
3050 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3051 | { | |
e041c683 | 3052 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3053 | } |
6ac552fd | 3054 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3055 | |
90ddc4f0 | 3056 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3057 | .family = PF_NETLINK, |
3058 | .owner = THIS_MODULE, | |
3059 | .release = netlink_release, | |
3060 | .bind = netlink_bind, | |
3061 | .connect = netlink_connect, | |
3062 | .socketpair = sock_no_socketpair, | |
3063 | .accept = sock_no_accept, | |
3064 | .getname = netlink_getname, | |
9652e931 | 3065 | .poll = netlink_poll, |
1da177e4 LT |
3066 | .ioctl = sock_no_ioctl, |
3067 | .listen = sock_no_listen, | |
3068 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3069 | .setsockopt = netlink_setsockopt, |
3070 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3071 | .sendmsg = netlink_sendmsg, |
3072 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3073 | .mmap = netlink_mmap, |
1da177e4 LT |
3074 | .sendpage = sock_no_sendpage, |
3075 | }; | |
3076 | ||
ec1b4cf7 | 3077 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3078 | .family = PF_NETLINK, |
3079 | .create = netlink_create, | |
3080 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3081 | }; | |
3082 | ||
4665079c | 3083 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3084 | { |
3085 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3086 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3087 | return -ENOMEM; |
3088 | #endif | |
3089 | return 0; | |
3090 | } | |
3091 | ||
4665079c | 3092 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3093 | { |
3094 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3095 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3096 | #endif |
3097 | } | |
3098 | ||
b963ea89 DM |
3099 | static void __init netlink_add_usersock_entry(void) |
3100 | { | |
5c398dc8 | 3101 | struct listeners *listeners; |
b963ea89 DM |
3102 | int groups = 32; |
3103 | ||
5c398dc8 | 3104 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3105 | if (!listeners) |
5c398dc8 | 3106 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3107 | |
3108 | netlink_table_grab(); | |
3109 | ||
3110 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3111 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3112 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3113 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3114 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3115 | |
3116 | netlink_table_ungrab(); | |
3117 | } | |
3118 | ||
022cbae6 | 3119 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3120 | .init = netlink_net_init, |
3121 | .exit = netlink_net_exit, | |
3122 | }; | |
3123 | ||
1da177e4 LT |
3124 | static int __init netlink_proto_init(void) |
3125 | { | |
1da177e4 | 3126 | int i; |
1da177e4 | 3127 | int err = proto_register(&netlink_proto, 0); |
e341694e TG |
3128 | struct rhashtable_params ht_params = { |
3129 | .head_offset = offsetof(struct netlink_sock, node), | |
3130 | .key_offset = offsetof(struct netlink_sock, portid), | |
3131 | .key_len = sizeof(u32), /* portid */ | |
3132 | .hashfn = arch_fast_hash, | |
3133 | .max_shift = 16, /* 64K */ | |
3134 | .grow_decision = rht_grow_above_75, | |
3135 | .shrink_decision = rht_shrink_below_30, | |
97127566 | 3136 | #ifdef CONFIG_PROVE_LOCKING |
e341694e | 3137 | .mutex_is_held = lockdep_nl_sk_hash_is_held, |
97127566 | 3138 | #endif |
e341694e | 3139 | }; |
1da177e4 LT |
3140 | |
3141 | if (err != 0) | |
3142 | goto out; | |
3143 | ||
fab25745 | 3144 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3145 | |
0da974f4 | 3146 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3147 | if (!nl_table) |
3148 | goto panic; | |
1da177e4 | 3149 | |
1da177e4 | 3150 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e TG |
3151 | if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) { |
3152 | while (--i > 0) | |
3153 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3154 | kfree(nl_table); |
fab2caf6 | 3155 | goto panic; |
1da177e4 | 3156 | } |
1da177e4 LT |
3157 | } |
3158 | ||
bcbde0d4 DB |
3159 | INIT_LIST_HEAD(&netlink_tap_all); |
3160 | ||
b963ea89 DM |
3161 | netlink_add_usersock_entry(); |
3162 | ||
1da177e4 | 3163 | sock_register(&netlink_family_ops); |
b4b51029 | 3164 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3165 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3166 | rtnetlink_init(); |
3167 | out: | |
3168 | return err; | |
fab2caf6 AM |
3169 | panic: |
3170 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3171 | } |
3172 | ||
1da177e4 | 3173 | core_initcall(netlink_proto_init); |