]>
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 |
7b4ce235 | 118 | static int lockdep_nl_sk_hash_is_held(void *parent) |
e341694e | 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 | ||
fcd4d35e | 145 | __module_get(nt->module); |
bcbde0d4 DB |
146 | |
147 | return 0; | |
148 | } | |
149 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
150 | ||
2173f8d9 | 151 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
152 | { |
153 | bool found = false; | |
154 | struct netlink_tap *tmp; | |
155 | ||
156 | spin_lock(&netlink_tap_lock); | |
157 | ||
158 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
159 | if (nt == tmp) { | |
160 | list_del_rcu(&nt->list); | |
161 | found = true; | |
162 | goto out; | |
163 | } | |
164 | } | |
165 | ||
166 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
167 | out: | |
168 | spin_unlock(&netlink_tap_lock); | |
169 | ||
170 | if (found && nt->module) | |
171 | module_put(nt->module); | |
172 | ||
173 | return found ? 0 : -ENODEV; | |
174 | } | |
bcbde0d4 DB |
175 | |
176 | int netlink_remove_tap(struct netlink_tap *nt) | |
177 | { | |
178 | int ret; | |
179 | ||
180 | ret = __netlink_remove_tap(nt); | |
181 | synchronize_net(); | |
182 | ||
183 | return ret; | |
184 | } | |
185 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
186 | ||
5ffd5cdd DB |
187 | static bool netlink_filter_tap(const struct sk_buff *skb) |
188 | { | |
189 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
190 | |
191 | /* We take the more conservative approach and | |
192 | * whitelist socket protocols that may pass. | |
193 | */ | |
194 | switch (sk->sk_protocol) { | |
195 | case NETLINK_ROUTE: | |
196 | case NETLINK_USERSOCK: | |
197 | case NETLINK_SOCK_DIAG: | |
198 | case NETLINK_NFLOG: | |
199 | case NETLINK_XFRM: | |
200 | case NETLINK_FIB_LOOKUP: | |
201 | case NETLINK_NETFILTER: | |
202 | case NETLINK_GENERIC: | |
498044bb | 203 | return true; |
5ffd5cdd DB |
204 | } |
205 | ||
498044bb | 206 | return false; |
5ffd5cdd DB |
207 | } |
208 | ||
bcbde0d4 DB |
209 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
210 | struct net_device *dev) | |
211 | { | |
212 | struct sk_buff *nskb; | |
5ffd5cdd | 213 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
214 | int ret = -ENOMEM; |
215 | ||
216 | dev_hold(dev); | |
217 | nskb = skb_clone(skb, GFP_ATOMIC); | |
218 | if (nskb) { | |
219 | nskb->dev = dev; | |
5ffd5cdd | 220 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
221 | nskb->pkt_type = netlink_is_kernel(sk) ? |
222 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 223 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
224 | ret = dev_queue_xmit(nskb); |
225 | if (unlikely(ret > 0)) | |
226 | ret = net_xmit_errno(ret); | |
227 | } | |
228 | ||
229 | dev_put(dev); | |
230 | return ret; | |
231 | } | |
232 | ||
233 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
234 | { | |
235 | int ret; | |
236 | struct netlink_tap *tmp; | |
237 | ||
5ffd5cdd DB |
238 | if (!netlink_filter_tap(skb)) |
239 | return; | |
240 | ||
bcbde0d4 DB |
241 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
242 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
243 | if (unlikely(ret)) | |
244 | break; | |
245 | } | |
246 | } | |
247 | ||
248 | static void netlink_deliver_tap(struct sk_buff *skb) | |
249 | { | |
250 | rcu_read_lock(); | |
251 | ||
252 | if (unlikely(!list_empty(&netlink_tap_all))) | |
253 | __netlink_deliver_tap(skb); | |
254 | ||
255 | rcu_read_unlock(); | |
256 | } | |
257 | ||
73bfd370 DB |
258 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
259 | struct sk_buff *skb) | |
260 | { | |
261 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
262 | netlink_deliver_tap(skb); | |
263 | } | |
264 | ||
cd1df525 PM |
265 | static void netlink_overrun(struct sock *sk) |
266 | { | |
267 | struct netlink_sock *nlk = nlk_sk(sk); | |
268 | ||
269 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
270 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
271 | sk->sk_err = ENOBUFS; | |
272 | sk->sk_error_report(sk); | |
273 | } | |
274 | } | |
275 | atomic_inc(&sk->sk_drops); | |
276 | } | |
277 | ||
278 | static void netlink_rcv_wake(struct sock *sk) | |
279 | { | |
280 | struct netlink_sock *nlk = nlk_sk(sk); | |
281 | ||
282 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
283 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
284 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
285 | wake_up_interruptible(&nlk->wait); | |
286 | } | |
287 | ||
ccdfcc39 | 288 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
289 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
290 | { | |
291 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
292 | } | |
293 | ||
f9c22888 PM |
294 | static bool netlink_rx_is_mmaped(struct sock *sk) |
295 | { | |
296 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
297 | } | |
298 | ||
5fd96123 PM |
299 | static bool netlink_tx_is_mmaped(struct sock *sk) |
300 | { | |
301 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
302 | } | |
303 | ||
ccdfcc39 PM |
304 | static __pure struct page *pgvec_to_page(const void *addr) |
305 | { | |
306 | if (is_vmalloc_addr(addr)) | |
307 | return vmalloc_to_page(addr); | |
308 | else | |
309 | return virt_to_page(addr); | |
310 | } | |
311 | ||
312 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
313 | { | |
314 | unsigned int i; | |
315 | ||
316 | for (i = 0; i < len; i++) { | |
317 | if (pg_vec[i] != NULL) { | |
318 | if (is_vmalloc_addr(pg_vec[i])) | |
319 | vfree(pg_vec[i]); | |
320 | else | |
321 | free_pages((unsigned long)pg_vec[i], order); | |
322 | } | |
323 | } | |
324 | kfree(pg_vec); | |
325 | } | |
326 | ||
327 | static void *alloc_one_pg_vec_page(unsigned long order) | |
328 | { | |
329 | void *buffer; | |
330 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
331 | __GFP_NOWARN | __GFP_NORETRY; | |
332 | ||
333 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
334 | if (buffer != NULL) | |
335 | return buffer; | |
336 | ||
337 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
338 | if (buffer != NULL) | |
339 | return buffer; | |
340 | ||
341 | gfp_flags &= ~__GFP_NORETRY; | |
342 | return (void *)__get_free_pages(gfp_flags, order); | |
343 | } | |
344 | ||
345 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
346 | struct nl_mmap_req *req, unsigned int order) | |
347 | { | |
348 | unsigned int block_nr = req->nm_block_nr; | |
349 | unsigned int i; | |
8a849bb7 | 350 | void **pg_vec; |
ccdfcc39 PM |
351 | |
352 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
353 | if (pg_vec == NULL) | |
354 | return NULL; | |
355 | ||
356 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 357 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
358 | if (pg_vec[i] == NULL) |
359 | goto err1; | |
360 | } | |
361 | ||
362 | return pg_vec; | |
363 | err1: | |
364 | free_pg_vec(pg_vec, order, block_nr); | |
365 | return NULL; | |
366 | } | |
367 | ||
368 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
369 | bool closing, bool tx_ring) | |
370 | { | |
371 | struct netlink_sock *nlk = nlk_sk(sk); | |
372 | struct netlink_ring *ring; | |
373 | struct sk_buff_head *queue; | |
374 | void **pg_vec = NULL; | |
375 | unsigned int order = 0; | |
376 | int err; | |
377 | ||
378 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
379 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
380 | ||
381 | if (!closing) { | |
382 | if (atomic_read(&nlk->mapped)) | |
383 | return -EBUSY; | |
384 | if (atomic_read(&ring->pending)) | |
385 | return -EBUSY; | |
386 | } | |
387 | ||
388 | if (req->nm_block_nr) { | |
389 | if (ring->pg_vec != NULL) | |
390 | return -EBUSY; | |
391 | ||
392 | if ((int)req->nm_block_size <= 0) | |
393 | return -EINVAL; | |
74e83b23 | 394 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
395 | return -EINVAL; |
396 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
397 | return -EINVAL; | |
398 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
399 | return -EINVAL; | |
400 | ||
401 | ring->frames_per_block = req->nm_block_size / | |
402 | req->nm_frame_size; | |
403 | if (ring->frames_per_block == 0) | |
404 | return -EINVAL; | |
405 | if (ring->frames_per_block * req->nm_block_nr != | |
406 | req->nm_frame_nr) | |
407 | return -EINVAL; | |
408 | ||
409 | order = get_order(req->nm_block_size); | |
410 | pg_vec = alloc_pg_vec(nlk, req, order); | |
411 | if (pg_vec == NULL) | |
412 | return -ENOMEM; | |
413 | } else { | |
414 | if (req->nm_frame_nr) | |
415 | return -EINVAL; | |
416 | } | |
417 | ||
418 | err = -EBUSY; | |
419 | mutex_lock(&nlk->pg_vec_lock); | |
420 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
421 | err = 0; | |
422 | spin_lock_bh(&queue->lock); | |
423 | ||
424 | ring->frame_max = req->nm_frame_nr - 1; | |
425 | ring->head = 0; | |
426 | ring->frame_size = req->nm_frame_size; | |
427 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
428 | ||
429 | swap(ring->pg_vec_len, req->nm_block_nr); | |
430 | swap(ring->pg_vec_order, order); | |
431 | swap(ring->pg_vec, pg_vec); | |
432 | ||
433 | __skb_queue_purge(queue); | |
434 | spin_unlock_bh(&queue->lock); | |
435 | ||
436 | WARN_ON(atomic_read(&nlk->mapped)); | |
437 | } | |
438 | mutex_unlock(&nlk->pg_vec_lock); | |
439 | ||
440 | if (pg_vec) | |
441 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
442 | return err; | |
443 | } | |
444 | ||
445 | static void netlink_mm_open(struct vm_area_struct *vma) | |
446 | { | |
447 | struct file *file = vma->vm_file; | |
448 | struct socket *sock = file->private_data; | |
449 | struct sock *sk = sock->sk; | |
450 | ||
451 | if (sk) | |
452 | atomic_inc(&nlk_sk(sk)->mapped); | |
453 | } | |
454 | ||
455 | static void netlink_mm_close(struct vm_area_struct *vma) | |
456 | { | |
457 | struct file *file = vma->vm_file; | |
458 | struct socket *sock = file->private_data; | |
459 | struct sock *sk = sock->sk; | |
460 | ||
461 | if (sk) | |
462 | atomic_dec(&nlk_sk(sk)->mapped); | |
463 | } | |
464 | ||
465 | static const struct vm_operations_struct netlink_mmap_ops = { | |
466 | .open = netlink_mm_open, | |
467 | .close = netlink_mm_close, | |
468 | }; | |
469 | ||
470 | static int netlink_mmap(struct file *file, struct socket *sock, | |
471 | struct vm_area_struct *vma) | |
472 | { | |
473 | struct sock *sk = sock->sk; | |
474 | struct netlink_sock *nlk = nlk_sk(sk); | |
475 | struct netlink_ring *ring; | |
476 | unsigned long start, size, expected; | |
477 | unsigned int i; | |
478 | int err = -EINVAL; | |
479 | ||
480 | if (vma->vm_pgoff) | |
481 | return -EINVAL; | |
482 | ||
483 | mutex_lock(&nlk->pg_vec_lock); | |
484 | ||
485 | expected = 0; | |
486 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
487 | if (ring->pg_vec == NULL) | |
488 | continue; | |
489 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
490 | } | |
491 | ||
492 | if (expected == 0) | |
493 | goto out; | |
494 | ||
495 | size = vma->vm_end - vma->vm_start; | |
496 | if (size != expected) | |
497 | goto out; | |
498 | ||
499 | start = vma->vm_start; | |
500 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
501 | if (ring->pg_vec == NULL) | |
502 | continue; | |
503 | ||
504 | for (i = 0; i < ring->pg_vec_len; i++) { | |
505 | struct page *page; | |
506 | void *kaddr = ring->pg_vec[i]; | |
507 | unsigned int pg_num; | |
508 | ||
509 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
510 | page = pgvec_to_page(kaddr); | |
511 | err = vm_insert_page(vma, start, page); | |
512 | if (err < 0) | |
513 | goto out; | |
514 | start += PAGE_SIZE; | |
515 | kaddr += PAGE_SIZE; | |
516 | } | |
517 | } | |
518 | } | |
519 | ||
520 | atomic_inc(&nlk->mapped); | |
521 | vma->vm_ops = &netlink_mmap_ops; | |
522 | err = 0; | |
523 | out: | |
524 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 525 | return err; |
ccdfcc39 | 526 | } |
9652e931 | 527 | |
4682a035 | 528 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
529 | { |
530 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
531 | struct page *p_start, *p_end; | |
532 | ||
533 | /* First page is flushed through netlink_{get,set}_status */ | |
534 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 535 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
536 | while (p_start <= p_end) { |
537 | flush_dcache_page(p_start); | |
538 | p_start++; | |
539 | } | |
540 | #endif | |
541 | } | |
542 | ||
543 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
544 | { | |
545 | smp_rmb(); | |
546 | flush_dcache_page(pgvec_to_page(hdr)); | |
547 | return hdr->nm_status; | |
548 | } | |
549 | ||
550 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
551 | enum nl_mmap_status status) | |
552 | { | |
a18e6a18 | 553 | smp_mb(); |
9652e931 PM |
554 | hdr->nm_status = status; |
555 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
556 | } |
557 | ||
558 | static struct nl_mmap_hdr * | |
559 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
560 | { | |
561 | unsigned int pg_vec_pos, frame_off; | |
562 | ||
563 | pg_vec_pos = pos / ring->frames_per_block; | |
564 | frame_off = pos % ring->frames_per_block; | |
565 | ||
566 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
567 | } | |
568 | ||
569 | static struct nl_mmap_hdr * | |
570 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
571 | enum nl_mmap_status status) | |
572 | { | |
573 | struct nl_mmap_hdr *hdr; | |
574 | ||
575 | hdr = __netlink_lookup_frame(ring, pos); | |
576 | if (netlink_get_status(hdr) != status) | |
577 | return NULL; | |
578 | ||
579 | return hdr; | |
580 | } | |
581 | ||
582 | static struct nl_mmap_hdr * | |
583 | netlink_current_frame(const struct netlink_ring *ring, | |
584 | enum nl_mmap_status status) | |
585 | { | |
586 | return netlink_lookup_frame(ring, ring->head, status); | |
587 | } | |
588 | ||
589 | static struct nl_mmap_hdr * | |
590 | netlink_previous_frame(const struct netlink_ring *ring, | |
591 | enum nl_mmap_status status) | |
592 | { | |
593 | unsigned int prev; | |
594 | ||
595 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
596 | return netlink_lookup_frame(ring, prev, status); | |
597 | } | |
598 | ||
599 | static void netlink_increment_head(struct netlink_ring *ring) | |
600 | { | |
601 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
602 | } | |
603 | ||
604 | static void netlink_forward_ring(struct netlink_ring *ring) | |
605 | { | |
606 | unsigned int head = ring->head, pos = head; | |
607 | const struct nl_mmap_hdr *hdr; | |
608 | ||
609 | do { | |
610 | hdr = __netlink_lookup_frame(ring, pos); | |
611 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
612 | break; | |
613 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
614 | break; | |
615 | netlink_increment_head(ring); | |
616 | } while (ring->head != head); | |
617 | } | |
618 | ||
cd1df525 PM |
619 | static bool netlink_dump_space(struct netlink_sock *nlk) |
620 | { | |
621 | struct netlink_ring *ring = &nlk->rx_ring; | |
622 | struct nl_mmap_hdr *hdr; | |
623 | unsigned int n; | |
624 | ||
625 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
626 | if (hdr == NULL) | |
627 | return false; | |
628 | ||
629 | n = ring->head + ring->frame_max / 2; | |
630 | if (n > ring->frame_max) | |
631 | n -= ring->frame_max; | |
632 | ||
633 | hdr = __netlink_lookup_frame(ring, n); | |
634 | ||
635 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
636 | } | |
637 | ||
9652e931 PM |
638 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
639 | poll_table *wait) | |
640 | { | |
641 | struct sock *sk = sock->sk; | |
642 | struct netlink_sock *nlk = nlk_sk(sk); | |
643 | unsigned int mask; | |
cd1df525 | 644 | int err; |
9652e931 | 645 | |
cd1df525 PM |
646 | if (nlk->rx_ring.pg_vec != NULL) { |
647 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
648 | * for dumps is performed here. A dump is allowed to continue | |
649 | * if at least half the ring is unused. | |
650 | */ | |
16b304f3 | 651 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
652 | err = netlink_dump(sk); |
653 | if (err < 0) { | |
ac30ef83 | 654 | sk->sk_err = -err; |
cd1df525 PM |
655 | sk->sk_error_report(sk); |
656 | break; | |
657 | } | |
658 | } | |
659 | netlink_rcv_wake(sk); | |
660 | } | |
5fd96123 | 661 | |
9652e931 PM |
662 | mask = datagram_poll(file, sock, wait); |
663 | ||
664 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
665 | if (nlk->rx_ring.pg_vec) { | |
666 | netlink_forward_ring(&nlk->rx_ring); | |
667 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
668 | mask |= POLLIN | POLLRDNORM; | |
669 | } | |
670 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
671 | ||
672 | spin_lock_bh(&sk->sk_write_queue.lock); | |
673 | if (nlk->tx_ring.pg_vec) { | |
674 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
675 | mask |= POLLOUT | POLLWRNORM; | |
676 | } | |
677 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
678 | ||
679 | return mask; | |
680 | } | |
681 | ||
682 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
683 | { | |
684 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
685 | } | |
686 | ||
687 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
688 | struct netlink_ring *ring, | |
689 | struct nl_mmap_hdr *hdr) | |
690 | { | |
691 | unsigned int size; | |
692 | void *data; | |
693 | ||
694 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
695 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
696 | ||
697 | skb->head = data; | |
698 | skb->data = data; | |
699 | skb_reset_tail_pointer(skb); | |
700 | skb->end = skb->tail + size; | |
701 | skb->len = 0; | |
702 | ||
703 | skb->destructor = netlink_skb_destructor; | |
704 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
705 | NETLINK_CB(skb).sk = sk; | |
706 | } | |
5fd96123 PM |
707 | |
708 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
709 | u32 dst_portid, u32 dst_group, | |
710 | struct sock_iocb *siocb) | |
711 | { | |
712 | struct netlink_sock *nlk = nlk_sk(sk); | |
713 | struct netlink_ring *ring; | |
714 | struct nl_mmap_hdr *hdr; | |
715 | struct sk_buff *skb; | |
716 | unsigned int maxlen; | |
5fd96123 PM |
717 | int err = 0, len = 0; |
718 | ||
5fd96123 PM |
719 | mutex_lock(&nlk->pg_vec_lock); |
720 | ||
721 | ring = &nlk->tx_ring; | |
722 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
723 | ||
724 | do { | |
4682a035 DM |
725 | unsigned int nm_len; |
726 | ||
5fd96123 PM |
727 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
728 | if (hdr == NULL) { | |
729 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
730 | atomic_read(&nlk->tx_ring.pending)) | |
731 | schedule(); | |
732 | continue; | |
733 | } | |
4682a035 DM |
734 | |
735 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
736 | if (nm_len > maxlen) { | |
5fd96123 PM |
737 | err = -EINVAL; |
738 | goto out; | |
739 | } | |
740 | ||
4682a035 | 741 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 742 | |
4682a035 DM |
743 | skb = alloc_skb(nm_len, GFP_KERNEL); |
744 | if (skb == NULL) { | |
745 | err = -ENOBUFS; | |
746 | goto out; | |
5fd96123 | 747 | } |
4682a035 DM |
748 | __skb_put(skb, nm_len); |
749 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
750 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
751 | |
752 | netlink_increment_head(ring); | |
753 | ||
754 | NETLINK_CB(skb).portid = nlk->portid; | |
755 | NETLINK_CB(skb).dst_group = dst_group; | |
756 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
757 | ||
758 | err = security_netlink_send(sk, skb); | |
759 | if (err) { | |
760 | kfree_skb(skb); | |
761 | goto out; | |
762 | } | |
763 | ||
764 | if (unlikely(dst_group)) { | |
765 | atomic_inc(&skb->users); | |
766 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
767 | GFP_KERNEL); | |
768 | } | |
769 | err = netlink_unicast(sk, skb, dst_portid, | |
770 | msg->msg_flags & MSG_DONTWAIT); | |
771 | if (err < 0) | |
772 | goto out; | |
773 | len += err; | |
774 | ||
775 | } while (hdr != NULL || | |
776 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
777 | atomic_read(&nlk->tx_ring.pending))); | |
778 | ||
779 | if (len > 0) | |
780 | err = len; | |
781 | out: | |
782 | mutex_unlock(&nlk->pg_vec_lock); | |
783 | return err; | |
784 | } | |
f9c22888 PM |
785 | |
786 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
787 | { | |
788 | struct nl_mmap_hdr *hdr; | |
789 | ||
790 | hdr = netlink_mmap_hdr(skb); | |
791 | hdr->nm_len = skb->len; | |
792 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
793 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
794 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
795 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 796 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
797 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
798 | ||
799 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
800 | kfree_skb(skb); | |
801 | } | |
802 | ||
803 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
804 | { | |
805 | struct netlink_sock *nlk = nlk_sk(sk); | |
806 | struct netlink_ring *ring = &nlk->rx_ring; | |
807 | struct nl_mmap_hdr *hdr; | |
808 | ||
809 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
810 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
811 | if (hdr == NULL) { | |
812 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
813 | kfree_skb(skb); | |
cd1df525 | 814 | netlink_overrun(sk); |
f9c22888 PM |
815 | return; |
816 | } | |
817 | netlink_increment_head(ring); | |
818 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
819 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
820 | ||
821 | hdr->nm_len = skb->len; | |
822 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
823 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
824 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
825 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
826 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
827 | } | |
828 | ||
ccdfcc39 | 829 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 830 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 831 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 832 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 833 | #define netlink_mmap sock_no_mmap |
9652e931 | 834 | #define netlink_poll datagram_poll |
5fd96123 | 835 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
836 | #endif /* CONFIG_NETLINK_MMAP */ |
837 | ||
cf0a018a PM |
838 | static void netlink_skb_destructor(struct sk_buff *skb) |
839 | { | |
9652e931 PM |
840 | #ifdef CONFIG_NETLINK_MMAP |
841 | struct nl_mmap_hdr *hdr; | |
842 | struct netlink_ring *ring; | |
843 | struct sock *sk; | |
844 | ||
845 | /* If a packet from the kernel to userspace was freed because of an | |
846 | * error without being delivered to userspace, the kernel must reset | |
847 | * the status. In the direction userspace to kernel, the status is | |
848 | * always reset here after the packet was processed and freed. | |
849 | */ | |
850 | if (netlink_skb_is_mmaped(skb)) { | |
851 | hdr = netlink_mmap_hdr(skb); | |
852 | sk = NETLINK_CB(skb).sk; | |
853 | ||
5fd96123 PM |
854 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
855 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
856 | ring = &nlk_sk(sk)->tx_ring; | |
857 | } else { | |
858 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
859 | hdr->nm_len = 0; | |
860 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
861 | } | |
862 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 863 | } |
9652e931 PM |
864 | |
865 | WARN_ON(atomic_read(&ring->pending) == 0); | |
866 | atomic_dec(&ring->pending); | |
867 | sock_put(sk); | |
868 | ||
5e71d9d7 | 869 | skb->head = NULL; |
9652e931 PM |
870 | } |
871 | #endif | |
c05cdb1b | 872 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
873 | if (!skb->cloned || |
874 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
875 | vfree(skb->head); | |
876 | ||
c05cdb1b PNA |
877 | skb->head = NULL; |
878 | } | |
9652e931 PM |
879 | if (skb->sk != NULL) |
880 | sock_rfree(skb); | |
cf0a018a PM |
881 | } |
882 | ||
883 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
884 | { | |
885 | WARN_ON(skb->sk != NULL); | |
886 | skb->sk = sk; | |
887 | skb->destructor = netlink_skb_destructor; | |
888 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
889 | sk_mem_charge(sk, skb->truesize); | |
890 | } | |
891 | ||
1da177e4 LT |
892 | static void netlink_sock_destruct(struct sock *sk) |
893 | { | |
3f660d66 HX |
894 | struct netlink_sock *nlk = nlk_sk(sk); |
895 | ||
16b304f3 PS |
896 | if (nlk->cb_running) { |
897 | if (nlk->cb.done) | |
898 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 899 | |
16b304f3 PS |
900 | module_put(nlk->cb.module); |
901 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
902 | } |
903 | ||
1da177e4 | 904 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
905 | #ifdef CONFIG_NETLINK_MMAP |
906 | if (1) { | |
907 | struct nl_mmap_req req; | |
908 | ||
909 | memset(&req, 0, sizeof(req)); | |
910 | if (nlk->rx_ring.pg_vec) | |
911 | netlink_set_ring(sk, &req, true, false); | |
912 | memset(&req, 0, sizeof(req)); | |
913 | if (nlk->tx_ring.pg_vec) | |
914 | netlink_set_ring(sk, &req, true, true); | |
915 | } | |
916 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
917 | |
918 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 919 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
920 | return; |
921 | } | |
547b792c IJ |
922 | |
923 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
924 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
925 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
926 | } |
927 | ||
6ac552fd PM |
928 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
929 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
930 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
931 | * this, _but_ remember, it adds useless work on UP machines. | |
932 | */ | |
933 | ||
d136f1bd | 934 | void netlink_table_grab(void) |
9a429c49 | 935 | __acquires(nl_table_lock) |
1da177e4 | 936 | { |
d136f1bd JB |
937 | might_sleep(); |
938 | ||
6abd219c | 939 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
940 | |
941 | if (atomic_read(&nl_table_users)) { | |
942 | DECLARE_WAITQUEUE(wait, current); | |
943 | ||
944 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 945 | for (;;) { |
1da177e4 LT |
946 | set_current_state(TASK_UNINTERRUPTIBLE); |
947 | if (atomic_read(&nl_table_users) == 0) | |
948 | break; | |
6abd219c | 949 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 950 | schedule(); |
6abd219c | 951 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
952 | } |
953 | ||
954 | __set_current_state(TASK_RUNNING); | |
955 | remove_wait_queue(&nl_table_wait, &wait); | |
956 | } | |
957 | } | |
958 | ||
d136f1bd | 959 | void netlink_table_ungrab(void) |
9a429c49 | 960 | __releases(nl_table_lock) |
1da177e4 | 961 | { |
6abd219c | 962 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
963 | wake_up(&nl_table_wait); |
964 | } | |
965 | ||
6ac552fd | 966 | static inline void |
1da177e4 LT |
967 | netlink_lock_table(void) |
968 | { | |
969 | /* read_lock() synchronizes us to netlink_table_grab */ | |
970 | ||
971 | read_lock(&nl_table_lock); | |
972 | atomic_inc(&nl_table_users); | |
973 | read_unlock(&nl_table_lock); | |
974 | } | |
975 | ||
6ac552fd | 976 | static inline void |
1da177e4 LT |
977 | netlink_unlock_table(void) |
978 | { | |
979 | if (atomic_dec_and_test(&nl_table_users)) | |
980 | wake_up(&nl_table_wait); | |
981 | } | |
982 | ||
e341694e | 983 | struct netlink_compare_arg |
1da177e4 | 984 | { |
e341694e TG |
985 | struct net *net; |
986 | u32 portid; | |
987 | }; | |
1da177e4 | 988 | |
e341694e | 989 | static bool netlink_compare(void *ptr, void *arg) |
1da177e4 | 990 | { |
e341694e TG |
991 | struct netlink_compare_arg *x = arg; |
992 | struct sock *sk = ptr; | |
1da177e4 | 993 | |
e341694e TG |
994 | return nlk_sk(sk)->portid == x->portid && |
995 | net_eq(sock_net(sk), x->net); | |
1da177e4 LT |
996 | } |
997 | ||
e341694e TG |
998 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
999 | struct net *net) | |
1da177e4 | 1000 | { |
e341694e TG |
1001 | struct netlink_compare_arg arg = { |
1002 | .net = net, | |
1003 | .portid = portid, | |
1004 | }; | |
1da177e4 | 1005 | |
8d24c0b4 | 1006 | return rhashtable_lookup_compare(&table->hash, &portid, |
e341694e | 1007 | &netlink_compare, &arg); |
1da177e4 LT |
1008 | } |
1009 | ||
e341694e | 1010 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1011 | { |
e341694e TG |
1012 | struct netlink_table *table = &nl_table[protocol]; |
1013 | struct sock *sk; | |
1da177e4 | 1014 | |
78fd1d0a | 1015 | read_lock(&nl_table_lock); |
e341694e TG |
1016 | rcu_read_lock(); |
1017 | sk = __netlink_lookup(table, portid, net); | |
1018 | if (sk) | |
1019 | sock_hold(sk); | |
1020 | rcu_read_unlock(); | |
78fd1d0a | 1021 | read_unlock(&nl_table_lock); |
1da177e4 | 1022 | |
e341694e | 1023 | return sk; |
1da177e4 LT |
1024 | } |
1025 | ||
90ddc4f0 | 1026 | static const struct proto_ops netlink_ops; |
1da177e4 | 1027 | |
4277a083 PM |
1028 | static void |
1029 | netlink_update_listeners(struct sock *sk) | |
1030 | { | |
1031 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1032 | unsigned long mask; |
1033 | unsigned int i; | |
6d772ac5 ED |
1034 | struct listeners *listeners; |
1035 | ||
1036 | listeners = nl_deref_protected(tbl->listeners); | |
1037 | if (!listeners) | |
1038 | return; | |
4277a083 | 1039 | |
b4ff4f04 | 1040 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1041 | mask = 0; |
b67bfe0d | 1042 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1043 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1044 | mask |= nlk_sk(sk)->groups[i]; | |
1045 | } | |
6d772ac5 | 1046 | listeners->masks[i] = mask; |
4277a083 PM |
1047 | } |
1048 | /* this function is only called with the netlink table "grabbed", which | |
1049 | * makes sure updates are visible before bind or setsockopt return. */ | |
1050 | } | |
1051 | ||
15e47304 | 1052 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1053 | { |
da12c90e | 1054 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1da177e4 | 1055 | int err = -EADDRINUSE; |
1da177e4 | 1056 | |
e341694e TG |
1057 | mutex_lock(&nl_sk_hash_lock); |
1058 | if (__netlink_lookup(table, portid, net)) | |
1da177e4 LT |
1059 | goto err; |
1060 | ||
1061 | err = -EBUSY; | |
15e47304 | 1062 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1063 | goto err; |
1064 | ||
1065 | err = -ENOMEM; | |
e341694e | 1066 | if (BITS_PER_LONG > 32 && unlikely(table->hash.nelems >= UINT_MAX)) |
1da177e4 LT |
1067 | goto err; |
1068 | ||
15e47304 | 1069 | nlk_sk(sk)->portid = portid; |
e341694e | 1070 | sock_hold(sk); |
6eba8224 | 1071 | rhashtable_insert(&table->hash, &nlk_sk(sk)->node); |
1da177e4 | 1072 | err = 0; |
1da177e4 | 1073 | err: |
e341694e | 1074 | mutex_unlock(&nl_sk_hash_lock); |
1da177e4 LT |
1075 | return err; |
1076 | } | |
1077 | ||
1078 | static void netlink_remove(struct sock *sk) | |
1079 | { | |
e341694e TG |
1080 | struct netlink_table *table; |
1081 | ||
1082 | mutex_lock(&nl_sk_hash_lock); | |
1083 | table = &nl_table[sk->sk_protocol]; | |
6eba8224 | 1084 | if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node)) { |
e341694e TG |
1085 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1086 | __sock_put(sk); | |
1087 | } | |
1088 | mutex_unlock(&nl_sk_hash_lock); | |
1089 | ||
1da177e4 | 1090 | netlink_table_grab(); |
b10dcb3b | 1091 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1092 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1093 | netlink_update_listeners(sk); |
1094 | } | |
1da177e4 LT |
1095 | netlink_table_ungrab(); |
1096 | } | |
1097 | ||
1098 | static struct proto netlink_proto = { | |
1099 | .name = "NETLINK", | |
1100 | .owner = THIS_MODULE, | |
1101 | .obj_size = sizeof(struct netlink_sock), | |
1102 | }; | |
1103 | ||
1b8d7ae4 EB |
1104 | static int __netlink_create(struct net *net, struct socket *sock, |
1105 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1106 | { |
1107 | struct sock *sk; | |
1108 | struct netlink_sock *nlk; | |
ab33a171 PM |
1109 | |
1110 | sock->ops = &netlink_ops; | |
1111 | ||
6257ff21 | 1112 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1113 | if (!sk) |
1114 | return -ENOMEM; | |
1115 | ||
1116 | sock_init_data(sock, sk); | |
1117 | ||
1118 | nlk = nlk_sk(sk); | |
658cb354 | 1119 | if (cb_mutex) { |
ffa4d721 | 1120 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1121 | } else { |
ffa4d721 PM |
1122 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1123 | mutex_init(nlk->cb_mutex); | |
1124 | } | |
ab33a171 | 1125 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1126 | #ifdef CONFIG_NETLINK_MMAP |
1127 | mutex_init(&nlk->pg_vec_lock); | |
1128 | #endif | |
ab33a171 PM |
1129 | |
1130 | sk->sk_destruct = netlink_sock_destruct; | |
1131 | sk->sk_protocol = protocol; | |
1132 | return 0; | |
1133 | } | |
1134 | ||
3f378b68 EP |
1135 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1136 | int kern) | |
ab33a171 PM |
1137 | { |
1138 | struct module *module = NULL; | |
af65bdfc | 1139 | struct mutex *cb_mutex; |
f7fa9b10 | 1140 | struct netlink_sock *nlk; |
023e2cfa JB |
1141 | int (*bind)(struct net *net, int group); |
1142 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1143 | int err = 0; |
1da177e4 LT |
1144 | |
1145 | sock->state = SS_UNCONNECTED; | |
1146 | ||
1147 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1148 | return -ESOCKTNOSUPPORT; | |
1149 | ||
6ac552fd | 1150 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1151 | return -EPROTONOSUPPORT; |
1152 | ||
77247bbb | 1153 | netlink_lock_table(); |
95a5afca | 1154 | #ifdef CONFIG_MODULES |
ab33a171 | 1155 | if (!nl_table[protocol].registered) { |
77247bbb | 1156 | netlink_unlock_table(); |
4fdb3bb7 | 1157 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1158 | netlink_lock_table(); |
4fdb3bb7 | 1159 | } |
ab33a171 PM |
1160 | #endif |
1161 | if (nl_table[protocol].registered && | |
1162 | try_module_get(nl_table[protocol].module)) | |
1163 | module = nl_table[protocol].module; | |
974c37e9 AD |
1164 | else |
1165 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1166 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1167 | bind = nl_table[protocol].bind; |
4f520900 | 1168 | unbind = nl_table[protocol].unbind; |
77247bbb | 1169 | netlink_unlock_table(); |
4fdb3bb7 | 1170 | |
974c37e9 AD |
1171 | if (err < 0) |
1172 | goto out; | |
1173 | ||
6ac552fd PM |
1174 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1175 | if (err < 0) | |
f7fa9b10 PM |
1176 | goto out_module; |
1177 | ||
6f756a8c | 1178 | local_bh_disable(); |
c1fd3b94 | 1179 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1180 | local_bh_enable(); |
1181 | ||
f7fa9b10 | 1182 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1183 | nlk->module = module; |
03292745 | 1184 | nlk->netlink_bind = bind; |
4f520900 | 1185 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1186 | out: |
1187 | return err; | |
1da177e4 | 1188 | |
ab33a171 PM |
1189 | out_module: |
1190 | module_put(module); | |
1191 | goto out; | |
1da177e4 LT |
1192 | } |
1193 | ||
1194 | static int netlink_release(struct socket *sock) | |
1195 | { | |
1196 | struct sock *sk = sock->sk; | |
1197 | struct netlink_sock *nlk; | |
1198 | ||
1199 | if (!sk) | |
1200 | return 0; | |
1201 | ||
1202 | netlink_remove(sk); | |
ac57b3a9 | 1203 | sock_orphan(sk); |
1da177e4 LT |
1204 | nlk = nlk_sk(sk); |
1205 | ||
3f660d66 HX |
1206 | /* |
1207 | * OK. Socket is unlinked, any packets that arrive now | |
1208 | * will be purged. | |
1209 | */ | |
1da177e4 | 1210 | |
1da177e4 LT |
1211 | sock->sk = NULL; |
1212 | wake_up_interruptible_all(&nlk->wait); | |
1213 | ||
1214 | skb_queue_purge(&sk->sk_write_queue); | |
1215 | ||
15e47304 | 1216 | if (nlk->portid) { |
1da177e4 | 1217 | struct netlink_notify n = { |
3b1e0a65 | 1218 | .net = sock_net(sk), |
1da177e4 | 1219 | .protocol = sk->sk_protocol, |
15e47304 | 1220 | .portid = nlk->portid, |
1da177e4 | 1221 | }; |
e041c683 AS |
1222 | atomic_notifier_call_chain(&netlink_chain, |
1223 | NETLINK_URELEASE, &n); | |
746fac4d | 1224 | } |
4fdb3bb7 | 1225 | |
5e7c001c | 1226 | module_put(nlk->module); |
4fdb3bb7 | 1227 | |
aed81560 | 1228 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1229 | netlink_table_grab(); |
869e58f8 DL |
1230 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1231 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1232 | struct listeners *old; |
1233 | ||
1234 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1235 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1236 | kfree_rcu(old, rcu); | |
869e58f8 | 1237 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1238 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1239 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1240 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1241 | nl_table[sk->sk_protocol].registered = 0; |
1242 | } | |
b10dcb3b | 1243 | netlink_table_ungrab(); |
658cb354 | 1244 | } |
77247bbb | 1245 | |
7d68536b JB |
1246 | if (nlk->netlink_unbind) { |
1247 | int i; | |
1248 | ||
1249 | for (i = 0; i < nlk->ngroups; i++) | |
1250 | if (test_bit(i, nlk->groups)) | |
023e2cfa | 1251 | nlk->netlink_unbind(sock_net(sk), i + 1); |
7d68536b | 1252 | } |
f7fa9b10 PM |
1253 | kfree(nlk->groups); |
1254 | nlk->groups = NULL; | |
1255 | ||
3755810c | 1256 | local_bh_disable(); |
c1fd3b94 | 1257 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1258 | local_bh_enable(); |
1da177e4 LT |
1259 | sock_put(sk); |
1260 | return 0; | |
1261 | } | |
1262 | ||
1263 | static int netlink_autobind(struct socket *sock) | |
1264 | { | |
1265 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1266 | struct net *net = sock_net(sk); |
da12c90e | 1267 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1268 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1269 | int err; |
1270 | static s32 rover = -4097; | |
1271 | ||
1272 | retry: | |
1273 | cond_resched(); | |
78fd1d0a | 1274 | netlink_table_grab(); |
e341694e TG |
1275 | rcu_read_lock(); |
1276 | if (__netlink_lookup(table, portid, net)) { | |
1277 | /* Bind collision, search negative portid values. */ | |
1278 | portid = rover--; | |
1279 | if (rover > -4097) | |
1280 | rover = -4097; | |
1281 | rcu_read_unlock(); | |
78fd1d0a | 1282 | netlink_table_ungrab(); |
e341694e | 1283 | goto retry; |
1da177e4 | 1284 | } |
e341694e | 1285 | rcu_read_unlock(); |
78fd1d0a | 1286 | netlink_table_ungrab(); |
1da177e4 | 1287 | |
15e47304 | 1288 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1289 | if (err == -EADDRINUSE) |
1290 | goto retry; | |
d470e3b4 DM |
1291 | |
1292 | /* If 2 threads race to autobind, that is fine. */ | |
1293 | if (err == -EBUSY) | |
1294 | err = 0; | |
1295 | ||
1296 | return err; | |
1da177e4 LT |
1297 | } |
1298 | ||
aa4cf945 EB |
1299 | /** |
1300 | * __netlink_ns_capable - General netlink message capability test | |
1301 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1302 | * @user_ns: The user namespace of the capability to use | |
1303 | * @cap: The capability to use | |
1304 | * | |
1305 | * Test to see if the opener of the socket we received the message | |
1306 | * from had when the netlink socket was created and the sender of the | |
1307 | * message has has the capability @cap in the user namespace @user_ns. | |
1308 | */ | |
1309 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1310 | struct user_namespace *user_ns, int cap) | |
1311 | { | |
2d7a85f4 EB |
1312 | return ((nsp->flags & NETLINK_SKB_DST) || |
1313 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1314 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1315 | } |
1316 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1317 | ||
1318 | /** | |
1319 | * netlink_ns_capable - General netlink message capability test | |
1320 | * @skb: socket buffer holding a netlink command from userspace | |
1321 | * @user_ns: The user namespace of the capability to use | |
1322 | * @cap: The capability to use | |
1323 | * | |
1324 | * Test to see if the opener of the socket we received the message | |
1325 | * from had when the netlink socket was created and the sender of the | |
1326 | * message has has the capability @cap in the user namespace @user_ns. | |
1327 | */ | |
1328 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1329 | struct user_namespace *user_ns, int cap) | |
1330 | { | |
1331 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1332 | } | |
1333 | EXPORT_SYMBOL(netlink_ns_capable); | |
1334 | ||
1335 | /** | |
1336 | * netlink_capable - Netlink global message capability test | |
1337 | * @skb: socket buffer holding a netlink command from userspace | |
1338 | * @cap: The capability to use | |
1339 | * | |
1340 | * Test to see if the opener of the socket we received the message | |
1341 | * from had when the netlink socket was created and the sender of the | |
1342 | * message has has the capability @cap in all user namespaces. | |
1343 | */ | |
1344 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1345 | { | |
1346 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1347 | } | |
1348 | EXPORT_SYMBOL(netlink_capable); | |
1349 | ||
1350 | /** | |
1351 | * netlink_net_capable - Netlink network namespace message capability test | |
1352 | * @skb: socket buffer holding a netlink command from userspace | |
1353 | * @cap: The capability to use | |
1354 | * | |
1355 | * Test to see if the opener of the socket we received the message | |
1356 | * from had when the netlink socket was created and the sender of the | |
1357 | * message has has the capability @cap over the network namespace of | |
1358 | * the socket we received the message from. | |
1359 | */ | |
1360 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1361 | { | |
1362 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1363 | } | |
1364 | EXPORT_SYMBOL(netlink_net_capable); | |
1365 | ||
5187cd05 | 1366 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1367 | { |
9785e10a | 1368 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1369 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1370 | } |
1da177e4 | 1371 | |
f7fa9b10 PM |
1372 | static void |
1373 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1374 | { | |
1375 | struct netlink_sock *nlk = nlk_sk(sk); | |
1376 | ||
1377 | if (nlk->subscriptions && !subscriptions) | |
1378 | __sk_del_bind_node(sk); | |
1379 | else if (!nlk->subscriptions && subscriptions) | |
1380 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1381 | nlk->subscriptions = subscriptions; | |
1382 | } | |
1383 | ||
b4ff4f04 | 1384 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1385 | { |
1386 | struct netlink_sock *nlk = nlk_sk(sk); | |
1387 | unsigned int groups; | |
b4ff4f04 | 1388 | unsigned long *new_groups; |
513c2500 PM |
1389 | int err = 0; |
1390 | ||
b4ff4f04 JB |
1391 | netlink_table_grab(); |
1392 | ||
513c2500 | 1393 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1394 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1395 | err = -ENOENT; |
b4ff4f04 JB |
1396 | goto out_unlock; |
1397 | } | |
513c2500 | 1398 | |
b4ff4f04 JB |
1399 | if (nlk->ngroups >= groups) |
1400 | goto out_unlock; | |
513c2500 | 1401 | |
b4ff4f04 JB |
1402 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1403 | if (new_groups == NULL) { | |
1404 | err = -ENOMEM; | |
1405 | goto out_unlock; | |
1406 | } | |
6ac552fd | 1407 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1408 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1409 | ||
1410 | nlk->groups = new_groups; | |
513c2500 | 1411 | nlk->ngroups = groups; |
b4ff4f04 JB |
1412 | out_unlock: |
1413 | netlink_table_ungrab(); | |
1414 | return err; | |
513c2500 PM |
1415 | } |
1416 | ||
02c81ab9 | 1417 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1418 | struct sock *sk) |
4f520900 | 1419 | { |
023e2cfa | 1420 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1421 | int undo; |
1422 | ||
1423 | if (!nlk->netlink_unbind) | |
1424 | return; | |
1425 | ||
1426 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1427 | if (test_bit(undo, &groups)) |
023e2cfa | 1428 | nlk->netlink_unbind(sock_net(sk), undo); |
4f520900 RGB |
1429 | } |
1430 | ||
6ac552fd PM |
1431 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1432 | int addr_len) | |
1da177e4 LT |
1433 | { |
1434 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1435 | struct net *net = sock_net(sk); |
1da177e4 LT |
1436 | struct netlink_sock *nlk = nlk_sk(sk); |
1437 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1438 | int err; | |
4f520900 | 1439 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1440 | |
4e4b5376 HFS |
1441 | if (addr_len < sizeof(struct sockaddr_nl)) |
1442 | return -EINVAL; | |
1443 | ||
1da177e4 LT |
1444 | if (nladdr->nl_family != AF_NETLINK) |
1445 | return -EINVAL; | |
1446 | ||
1447 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1448 | if (groups) { |
5187cd05 | 1449 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1450 | return -EPERM; |
b4ff4f04 JB |
1451 | err = netlink_realloc_groups(sk); |
1452 | if (err) | |
1453 | return err; | |
513c2500 | 1454 | } |
1da177e4 | 1455 | |
4f520900 | 1456 | if (nlk->portid) |
15e47304 | 1457 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1458 | return -EINVAL; |
4f520900 RGB |
1459 | |
1460 | if (nlk->netlink_bind && groups) { | |
1461 | int group; | |
1462 | ||
1463 | for (group = 0; group < nlk->ngroups; group++) { | |
1464 | if (!test_bit(group, &groups)) | |
1465 | continue; | |
023e2cfa | 1466 | err = nlk->netlink_bind(net, group); |
4f520900 RGB |
1467 | if (!err) |
1468 | continue; | |
023e2cfa | 1469 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1470 | return err; |
1471 | } | |
1472 | } | |
1473 | ||
1474 | if (!nlk->portid) { | |
1da177e4 | 1475 | err = nladdr->nl_pid ? |
b4b51029 | 1476 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1477 | netlink_autobind(sock); |
4f520900 | 1478 | if (err) { |
023e2cfa | 1479 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1480 | return err; |
4f520900 | 1481 | } |
1da177e4 LT |
1482 | } |
1483 | ||
4f520900 | 1484 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1485 | return 0; |
1486 | ||
1487 | netlink_table_grab(); | |
f7fa9b10 | 1488 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1489 | hweight32(groups) - |
746fac4d | 1490 | hweight32(nlk->groups[0])); |
4f520900 | 1491 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1492 | netlink_update_listeners(sk); |
1da177e4 LT |
1493 | netlink_table_ungrab(); |
1494 | ||
1495 | return 0; | |
1496 | } | |
1497 | ||
1498 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1499 | int alen, int flags) | |
1500 | { | |
1501 | int err = 0; | |
1502 | struct sock *sk = sock->sk; | |
1503 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1504 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1505 | |
6503d961 CG |
1506 | if (alen < sizeof(addr->sa_family)) |
1507 | return -EINVAL; | |
1508 | ||
1da177e4 LT |
1509 | if (addr->sa_family == AF_UNSPEC) { |
1510 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1511 | nlk->dst_portid = 0; |
d629b836 | 1512 | nlk->dst_group = 0; |
1da177e4 LT |
1513 | return 0; |
1514 | } | |
1515 | if (addr->sa_family != AF_NETLINK) | |
1516 | return -EINVAL; | |
1517 | ||
46833a86 | 1518 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1519 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1520 | return -EPERM; |
1521 | ||
15e47304 | 1522 | if (!nlk->portid) |
1da177e4 LT |
1523 | err = netlink_autobind(sock); |
1524 | ||
1525 | if (err == 0) { | |
1526 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1527 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1528 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1529 | } |
1530 | ||
1531 | return err; | |
1532 | } | |
1533 | ||
6ac552fd PM |
1534 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1535 | int *addr_len, int peer) | |
1da177e4 LT |
1536 | { |
1537 | struct sock *sk = sock->sk; | |
1538 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1539 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1540 | |
1da177e4 LT |
1541 | nladdr->nl_family = AF_NETLINK; |
1542 | nladdr->nl_pad = 0; | |
1543 | *addr_len = sizeof(*nladdr); | |
1544 | ||
1545 | if (peer) { | |
15e47304 | 1546 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1547 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1548 | } else { |
15e47304 | 1549 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1550 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1551 | } |
1552 | return 0; | |
1553 | } | |
1554 | ||
15e47304 | 1555 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1556 | { |
1da177e4 LT |
1557 | struct sock *sock; |
1558 | struct netlink_sock *nlk; | |
1559 | ||
15e47304 | 1560 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1561 | if (!sock) |
1562 | return ERR_PTR(-ECONNREFUSED); | |
1563 | ||
1564 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1565 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1566 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1567 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1568 | sock_put(sock); |
1569 | return ERR_PTR(-ECONNREFUSED); | |
1570 | } | |
1571 | return sock; | |
1572 | } | |
1573 | ||
1574 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1575 | { | |
496ad9aa | 1576 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1577 | struct sock *sock; |
1578 | ||
1579 | if (!S_ISSOCK(inode->i_mode)) | |
1580 | return ERR_PTR(-ENOTSOCK); | |
1581 | ||
1582 | sock = SOCKET_I(inode)->sk; | |
1583 | if (sock->sk_family != AF_NETLINK) | |
1584 | return ERR_PTR(-EINVAL); | |
1585 | ||
1586 | sock_hold(sock); | |
1587 | return sock; | |
1588 | } | |
1589 | ||
3a36515f PN |
1590 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1591 | int broadcast) | |
c05cdb1b PNA |
1592 | { |
1593 | struct sk_buff *skb; | |
1594 | void *data; | |
1595 | ||
3a36515f | 1596 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1597 | return alloc_skb(size, GFP_KERNEL); |
1598 | ||
3a36515f PN |
1599 | size = SKB_DATA_ALIGN(size) + |
1600 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1601 | |
1602 | data = vmalloc(size); | |
1603 | if (data == NULL) | |
3a36515f | 1604 | return NULL; |
c05cdb1b | 1605 | |
3a36515f PN |
1606 | skb = build_skb(data, size); |
1607 | if (skb == NULL) | |
1608 | vfree(data); | |
1609 | else { | |
1610 | skb->head_frag = 0; | |
1611 | skb->destructor = netlink_skb_destructor; | |
1612 | } | |
c05cdb1b PNA |
1613 | |
1614 | return skb; | |
c05cdb1b PNA |
1615 | } |
1616 | ||
1da177e4 LT |
1617 | /* |
1618 | * Attach a skb to a netlink socket. | |
1619 | * The caller must hold a reference to the destination socket. On error, the | |
1620 | * reference is dropped. The skb is not send to the destination, just all | |
1621 | * all error checks are performed and memory in the queue is reserved. | |
1622 | * Return values: | |
1623 | * < 0: error. skb freed, reference to sock dropped. | |
1624 | * 0: continue | |
1625 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1626 | */ | |
9457afee | 1627 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1628 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1629 | { |
1630 | struct netlink_sock *nlk; | |
1631 | ||
1632 | nlk = nlk_sk(sk); | |
1633 | ||
5fd96123 PM |
1634 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1635 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1636 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1637 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1638 | if (!*timeo) { |
aed81560 | 1639 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1640 | netlink_overrun(sk); |
1641 | sock_put(sk); | |
1642 | kfree_skb(skb); | |
1643 | return -EAGAIN; | |
1644 | } | |
1645 | ||
1646 | __set_current_state(TASK_INTERRUPTIBLE); | |
1647 | add_wait_queue(&nlk->wait, &wait); | |
1648 | ||
1649 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1650 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1651 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1652 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1653 | |
1654 | __set_current_state(TASK_RUNNING); | |
1655 | remove_wait_queue(&nlk->wait, &wait); | |
1656 | sock_put(sk); | |
1657 | ||
1658 | if (signal_pending(current)) { | |
1659 | kfree_skb(skb); | |
c3d8d1e3 | 1660 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1661 | } |
1662 | return 1; | |
1663 | } | |
cf0a018a | 1664 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1665 | return 0; |
1666 | } | |
1667 | ||
4a7e7c2a | 1668 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1669 | { |
1da177e4 LT |
1670 | int len = skb->len; |
1671 | ||
bcbde0d4 DB |
1672 | netlink_deliver_tap(skb); |
1673 | ||
f9c22888 PM |
1674 | #ifdef CONFIG_NETLINK_MMAP |
1675 | if (netlink_skb_is_mmaped(skb)) | |
1676 | netlink_queue_mmaped_skb(sk, skb); | |
1677 | else if (netlink_rx_is_mmaped(sk)) | |
1678 | netlink_ring_set_copied(sk, skb); | |
1679 | else | |
1680 | #endif /* CONFIG_NETLINK_MMAP */ | |
1681 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1682 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1683 | return len; |
1684 | } | |
1685 | ||
1686 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1687 | { | |
1688 | int len = __netlink_sendskb(sk, skb); | |
1689 | ||
1da177e4 LT |
1690 | sock_put(sk); |
1691 | return len; | |
1692 | } | |
1693 | ||
1694 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1695 | { | |
1696 | kfree_skb(skb); | |
1697 | sock_put(sk); | |
1698 | } | |
1699 | ||
b57ef81f | 1700 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1701 | { |
1702 | int delta; | |
1703 | ||
1298ca46 | 1704 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1705 | if (netlink_skb_is_mmaped(skb)) |
1706 | return skb; | |
1da177e4 | 1707 | |
4305b541 | 1708 | delta = skb->end - skb->tail; |
c05cdb1b | 1709 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1710 | return skb; |
1711 | ||
1712 | if (skb_shared(skb)) { | |
1713 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1714 | if (!nskb) | |
1715 | return skb; | |
8460c00f | 1716 | consume_skb(skb); |
1da177e4 LT |
1717 | skb = nskb; |
1718 | } | |
1719 | ||
1720 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1721 | skb->truesize -= delta; | |
1722 | ||
1723 | return skb; | |
1724 | } | |
1725 | ||
3fbc2905 EB |
1726 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1727 | struct sock *ssk) | |
cd40b7d3 DL |
1728 | { |
1729 | int ret; | |
1730 | struct netlink_sock *nlk = nlk_sk(sk); | |
1731 | ||
1732 | ret = -ECONNREFUSED; | |
1733 | if (nlk->netlink_rcv != NULL) { | |
1734 | ret = skb->len; | |
cf0a018a | 1735 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1736 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1737 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1738 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1739 | consume_skb(skb); |
1740 | } else { | |
1741 | kfree_skb(skb); | |
cd40b7d3 | 1742 | } |
cd40b7d3 DL |
1743 | sock_put(sk); |
1744 | return ret; | |
1745 | } | |
1746 | ||
1747 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1748 | u32 portid, int nonblock) |
1da177e4 LT |
1749 | { |
1750 | struct sock *sk; | |
1751 | int err; | |
1752 | long timeo; | |
1753 | ||
1754 | skb = netlink_trim(skb, gfp_any()); | |
1755 | ||
1756 | timeo = sock_sndtimeo(ssk, nonblock); | |
1757 | retry: | |
15e47304 | 1758 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1759 | if (IS_ERR(sk)) { |
1760 | kfree_skb(skb); | |
1761 | return PTR_ERR(sk); | |
1762 | } | |
cd40b7d3 | 1763 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1764 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1765 | |
b1153f29 | 1766 | if (sk_filter(sk, skb)) { |
84874607 | 1767 | err = skb->len; |
b1153f29 SH |
1768 | kfree_skb(skb); |
1769 | sock_put(sk); | |
1770 | return err; | |
1771 | } | |
1772 | ||
9457afee | 1773 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1774 | if (err == 1) |
1775 | goto retry; | |
1776 | if (err) | |
1777 | return err; | |
1778 | ||
7ee015e0 | 1779 | return netlink_sendskb(sk, skb); |
1da177e4 | 1780 | } |
6ac552fd | 1781 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1782 | |
f9c22888 PM |
1783 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1784 | u32 dst_portid, gfp_t gfp_mask) | |
1785 | { | |
1786 | #ifdef CONFIG_NETLINK_MMAP | |
1787 | struct sock *sk = NULL; | |
1788 | struct sk_buff *skb; | |
1789 | struct netlink_ring *ring; | |
1790 | struct nl_mmap_hdr *hdr; | |
1791 | unsigned int maxlen; | |
1792 | ||
1793 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1794 | if (IS_ERR(sk)) | |
1795 | goto out; | |
1796 | ||
1797 | ring = &nlk_sk(sk)->rx_ring; | |
1798 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1799 | if (ring->pg_vec == NULL) | |
1800 | goto out_put; | |
1801 | ||
aae9f0e2 TG |
1802 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1803 | goto out_put; | |
1804 | ||
f9c22888 PM |
1805 | skb = alloc_skb_head(gfp_mask); |
1806 | if (skb == NULL) | |
1807 | goto err1; | |
1808 | ||
1809 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1810 | /* check again under lock */ | |
1811 | if (ring->pg_vec == NULL) | |
1812 | goto out_free; | |
1813 | ||
aae9f0e2 | 1814 | /* check again under lock */ |
f9c22888 PM |
1815 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1816 | if (maxlen < size) | |
1817 | goto out_free; | |
1818 | ||
1819 | netlink_forward_ring(ring); | |
1820 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1821 | if (hdr == NULL) | |
1822 | goto err2; | |
1823 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1824 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1825 | atomic_inc(&ring->pending); | |
1826 | netlink_increment_head(ring); | |
1827 | ||
1828 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1829 | return skb; | |
1830 | ||
1831 | err2: | |
1832 | kfree_skb(skb); | |
1833 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1834 | netlink_overrun(sk); |
f9c22888 PM |
1835 | err1: |
1836 | sock_put(sk); | |
1837 | return NULL; | |
1838 | ||
1839 | out_free: | |
1840 | kfree_skb(skb); | |
1841 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1842 | out_put: | |
1843 | sock_put(sk); | |
1844 | out: | |
1845 | #endif | |
1846 | return alloc_skb(size, gfp_mask); | |
1847 | } | |
1848 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1849 | ||
4277a083 PM |
1850 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1851 | { | |
1852 | int res = 0; | |
5c398dc8 | 1853 | struct listeners *listeners; |
4277a083 | 1854 | |
aed81560 | 1855 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1856 | |
1857 | rcu_read_lock(); | |
1858 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1859 | ||
6d772ac5 | 1860 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1861 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1862 | |
1863 | rcu_read_unlock(); | |
1864 | ||
4277a083 PM |
1865 | return res; |
1866 | } | |
1867 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1868 | ||
b57ef81f | 1869 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1870 | { |
1871 | struct netlink_sock *nlk = nlk_sk(sk); | |
1872 | ||
1873 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1874 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1875 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1876 | __netlink_sendskb(sk, skb); |
2c645800 | 1877 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1878 | } |
1879 | return -1; | |
1880 | } | |
1881 | ||
1882 | struct netlink_broadcast_data { | |
1883 | struct sock *exclude_sk; | |
b4b51029 | 1884 | struct net *net; |
15e47304 | 1885 | u32 portid; |
1da177e4 LT |
1886 | u32 group; |
1887 | int failure; | |
ff491a73 | 1888 | int delivery_failure; |
1da177e4 LT |
1889 | int congested; |
1890 | int delivered; | |
7d877f3b | 1891 | gfp_t allocation; |
1da177e4 | 1892 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1893 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1894 | void *tx_data; | |
1da177e4 LT |
1895 | }; |
1896 | ||
46c9521f RR |
1897 | static void do_one_broadcast(struct sock *sk, |
1898 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1899 | { |
1900 | struct netlink_sock *nlk = nlk_sk(sk); | |
1901 | int val; | |
1902 | ||
1903 | if (p->exclude_sk == sk) | |
46c9521f | 1904 | return; |
1da177e4 | 1905 | |
15e47304 | 1906 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1907 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1908 | return; |
1da177e4 | 1909 | |
878628fb | 1910 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1911 | return; |
b4b51029 | 1912 | |
1da177e4 LT |
1913 | if (p->failure) { |
1914 | netlink_overrun(sk); | |
46c9521f | 1915 | return; |
1da177e4 LT |
1916 | } |
1917 | ||
1918 | sock_hold(sk); | |
1919 | if (p->skb2 == NULL) { | |
68acc024 | 1920 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1921 | p->skb2 = skb_clone(p->skb, p->allocation); |
1922 | } else { | |
68acc024 TC |
1923 | p->skb2 = skb_get(p->skb); |
1924 | /* | |
1925 | * skb ownership may have been set when | |
1926 | * delivered to a previous socket. | |
1927 | */ | |
1928 | skb_orphan(p->skb2); | |
1da177e4 LT |
1929 | } |
1930 | } | |
1931 | if (p->skb2 == NULL) { | |
1932 | netlink_overrun(sk); | |
1933 | /* Clone failed. Notify ALL listeners. */ | |
1934 | p->failure = 1; | |
be0c22a4 PNA |
1935 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1936 | p->delivery_failure = 1; | |
910a7e90 EB |
1937 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1938 | kfree_skb(p->skb2); | |
1939 | p->skb2 = NULL; | |
b1153f29 SH |
1940 | } else if (sk_filter(sk, p->skb2)) { |
1941 | kfree_skb(p->skb2); | |
1942 | p->skb2 = NULL; | |
1da177e4 LT |
1943 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1944 | netlink_overrun(sk); | |
be0c22a4 PNA |
1945 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1946 | p->delivery_failure = 1; | |
1da177e4 LT |
1947 | } else { |
1948 | p->congested |= val; | |
1949 | p->delivered = 1; | |
1950 | p->skb2 = NULL; | |
1951 | } | |
1952 | sock_put(sk); | |
1da177e4 LT |
1953 | } |
1954 | ||
15e47304 | 1955 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1956 | u32 group, gfp_t allocation, |
1957 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1958 | void *filter_data) | |
1da177e4 | 1959 | { |
3b1e0a65 | 1960 | struct net *net = sock_net(ssk); |
1da177e4 | 1961 | struct netlink_broadcast_data info; |
1da177e4 LT |
1962 | struct sock *sk; |
1963 | ||
1964 | skb = netlink_trim(skb, allocation); | |
1965 | ||
1966 | info.exclude_sk = ssk; | |
b4b51029 | 1967 | info.net = net; |
15e47304 | 1968 | info.portid = portid; |
1da177e4 LT |
1969 | info.group = group; |
1970 | info.failure = 0; | |
ff491a73 | 1971 | info.delivery_failure = 0; |
1da177e4 LT |
1972 | info.congested = 0; |
1973 | info.delivered = 0; | |
1974 | info.allocation = allocation; | |
1975 | info.skb = skb; | |
1976 | info.skb2 = NULL; | |
910a7e90 EB |
1977 | info.tx_filter = filter; |
1978 | info.tx_data = filter_data; | |
1da177e4 LT |
1979 | |
1980 | /* While we sleep in clone, do not allow to change socket list */ | |
1981 | ||
1982 | netlink_lock_table(); | |
1983 | ||
b67bfe0d | 1984 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1985 | do_one_broadcast(sk, &info); |
1986 | ||
70d4bf6d | 1987 | consume_skb(skb); |
aa1c6a6f | 1988 | |
1da177e4 LT |
1989 | netlink_unlock_table(); |
1990 | ||
70d4bf6d NH |
1991 | if (info.delivery_failure) { |
1992 | kfree_skb(info.skb2); | |
ff491a73 | 1993 | return -ENOBUFS; |
658cb354 ED |
1994 | } |
1995 | consume_skb(info.skb2); | |
ff491a73 | 1996 | |
1da177e4 LT |
1997 | if (info.delivered) { |
1998 | if (info.congested && (allocation & __GFP_WAIT)) | |
1999 | yield(); | |
2000 | return 0; | |
2001 | } | |
1da177e4 LT |
2002 | return -ESRCH; |
2003 | } | |
910a7e90 EB |
2004 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2005 | ||
15e47304 | 2006 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2007 | u32 group, gfp_t allocation) |
2008 | { | |
15e47304 | 2009 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2010 | NULL, NULL); |
2011 | } | |
6ac552fd | 2012 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2013 | |
2014 | struct netlink_set_err_data { | |
2015 | struct sock *exclude_sk; | |
15e47304 | 2016 | u32 portid; |
1da177e4 LT |
2017 | u32 group; |
2018 | int code; | |
2019 | }; | |
2020 | ||
b57ef81f | 2021 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2022 | { |
2023 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2024 | int ret = 0; |
1da177e4 LT |
2025 | |
2026 | if (sk == p->exclude_sk) | |
2027 | goto out; | |
2028 | ||
09ad9bc7 | 2029 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2030 | goto out; |
2031 | ||
15e47304 | 2032 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2033 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2034 | goto out; |
2035 | ||
1a50307b PNA |
2036 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2037 | ret = 1; | |
2038 | goto out; | |
2039 | } | |
2040 | ||
1da177e4 LT |
2041 | sk->sk_err = p->code; |
2042 | sk->sk_error_report(sk); | |
2043 | out: | |
1a50307b | 2044 | return ret; |
1da177e4 LT |
2045 | } |
2046 | ||
4843b93c PNA |
2047 | /** |
2048 | * netlink_set_err - report error to broadcast listeners | |
2049 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2050 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2051 | * @group: the broadcast group that will notice the error |
4843b93c | 2052 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2053 | * |
2054 | * This function returns the number of broadcast listeners that have set the | |
2055 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2056 | */ |
15e47304 | 2057 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2058 | { |
2059 | struct netlink_set_err_data info; | |
1da177e4 | 2060 | struct sock *sk; |
1a50307b | 2061 | int ret = 0; |
1da177e4 LT |
2062 | |
2063 | info.exclude_sk = ssk; | |
15e47304 | 2064 | info.portid = portid; |
1da177e4 | 2065 | info.group = group; |
4843b93c PNA |
2066 | /* sk->sk_err wants a positive error value */ |
2067 | info.code = -code; | |
1da177e4 LT |
2068 | |
2069 | read_lock(&nl_table_lock); | |
2070 | ||
b67bfe0d | 2071 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2072 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2073 | |
2074 | read_unlock(&nl_table_lock); | |
1a50307b | 2075 | return ret; |
1da177e4 | 2076 | } |
dd5b6ce6 | 2077 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2078 | |
84659eb5 JB |
2079 | /* must be called with netlink table grabbed */ |
2080 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2081 | unsigned int group, | |
2082 | int is_new) | |
2083 | { | |
2084 | int old, new = !!is_new, subscriptions; | |
2085 | ||
2086 | old = test_bit(group - 1, nlk->groups); | |
2087 | subscriptions = nlk->subscriptions - old + new; | |
2088 | if (new) | |
2089 | __set_bit(group - 1, nlk->groups); | |
2090 | else | |
2091 | __clear_bit(group - 1, nlk->groups); | |
2092 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2093 | netlink_update_listeners(&nlk->sk); | |
2094 | } | |
2095 | ||
9a4595bc | 2096 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2097 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2098 | { |
2099 | struct sock *sk = sock->sk; | |
2100 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2101 | unsigned int val = 0; |
2102 | int err; | |
9a4595bc PM |
2103 | |
2104 | if (level != SOL_NETLINK) | |
2105 | return -ENOPROTOOPT; | |
2106 | ||
ccdfcc39 PM |
2107 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2108 | optlen >= sizeof(int) && | |
eb496534 | 2109 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2110 | return -EFAULT; |
2111 | ||
2112 | switch (optname) { | |
2113 | case NETLINK_PKTINFO: | |
2114 | if (val) | |
2115 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2116 | else | |
2117 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2118 | err = 0; | |
2119 | break; | |
2120 | case NETLINK_ADD_MEMBERSHIP: | |
2121 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2122 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2123 | return -EPERM; |
b4ff4f04 JB |
2124 | err = netlink_realloc_groups(sk); |
2125 | if (err) | |
2126 | return err; | |
9a4595bc PM |
2127 | if (!val || val - 1 >= nlk->ngroups) |
2128 | return -EINVAL; | |
7774d5e0 | 2129 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2130 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2131 | if (err) |
2132 | return err; | |
2133 | } | |
9a4595bc | 2134 | netlink_table_grab(); |
84659eb5 JB |
2135 | netlink_update_socket_mc(nlk, val, |
2136 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2137 | netlink_table_ungrab(); |
7774d5e0 | 2138 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2139 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2140 | |
9a4595bc PM |
2141 | err = 0; |
2142 | break; | |
2143 | } | |
be0c22a4 PNA |
2144 | case NETLINK_BROADCAST_ERROR: |
2145 | if (val) | |
2146 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2147 | else | |
2148 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2149 | err = 0; | |
2150 | break; | |
38938bfe PNA |
2151 | case NETLINK_NO_ENOBUFS: |
2152 | if (val) { | |
2153 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2154 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2155 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2156 | } else { |
38938bfe | 2157 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2158 | } |
38938bfe PNA |
2159 | err = 0; |
2160 | break; | |
ccdfcc39 PM |
2161 | #ifdef CONFIG_NETLINK_MMAP |
2162 | case NETLINK_RX_RING: | |
2163 | case NETLINK_TX_RING: { | |
2164 | struct nl_mmap_req req; | |
2165 | ||
2166 | /* Rings might consume more memory than queue limits, require | |
2167 | * CAP_NET_ADMIN. | |
2168 | */ | |
2169 | if (!capable(CAP_NET_ADMIN)) | |
2170 | return -EPERM; | |
2171 | if (optlen < sizeof(req)) | |
2172 | return -EINVAL; | |
2173 | if (copy_from_user(&req, optval, sizeof(req))) | |
2174 | return -EFAULT; | |
2175 | err = netlink_set_ring(sk, &req, false, | |
2176 | optname == NETLINK_TX_RING); | |
2177 | break; | |
2178 | } | |
2179 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2180 | default: |
2181 | err = -ENOPROTOOPT; | |
2182 | } | |
2183 | return err; | |
2184 | } | |
2185 | ||
2186 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2187 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2188 | { |
2189 | struct sock *sk = sock->sk; | |
2190 | struct netlink_sock *nlk = nlk_sk(sk); | |
2191 | int len, val, err; | |
2192 | ||
2193 | if (level != SOL_NETLINK) | |
2194 | return -ENOPROTOOPT; | |
2195 | ||
2196 | if (get_user(len, optlen)) | |
2197 | return -EFAULT; | |
2198 | if (len < 0) | |
2199 | return -EINVAL; | |
2200 | ||
2201 | switch (optname) { | |
2202 | case NETLINK_PKTINFO: | |
2203 | if (len < sizeof(int)) | |
2204 | return -EINVAL; | |
2205 | len = sizeof(int); | |
2206 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2207 | if (put_user(len, optlen) || |
2208 | put_user(val, optval)) | |
2209 | return -EFAULT; | |
9a4595bc PM |
2210 | err = 0; |
2211 | break; | |
be0c22a4 PNA |
2212 | case NETLINK_BROADCAST_ERROR: |
2213 | if (len < sizeof(int)) | |
2214 | return -EINVAL; | |
2215 | len = sizeof(int); | |
2216 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2217 | if (put_user(len, optlen) || | |
2218 | put_user(val, optval)) | |
2219 | return -EFAULT; | |
2220 | err = 0; | |
2221 | break; | |
38938bfe PNA |
2222 | case NETLINK_NO_ENOBUFS: |
2223 | if (len < sizeof(int)) | |
2224 | return -EINVAL; | |
2225 | len = sizeof(int); | |
2226 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2227 | if (put_user(len, optlen) || | |
2228 | put_user(val, optval)) | |
2229 | return -EFAULT; | |
2230 | err = 0; | |
2231 | break; | |
9a4595bc PM |
2232 | default: |
2233 | err = -ENOPROTOOPT; | |
2234 | } | |
2235 | return err; | |
2236 | } | |
2237 | ||
2238 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2239 | { | |
2240 | struct nl_pktinfo info; | |
2241 | ||
2242 | info.group = NETLINK_CB(skb).dst_group; | |
2243 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2244 | } | |
2245 | ||
1da177e4 LT |
2246 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2247 | struct msghdr *msg, size_t len) | |
2248 | { | |
2249 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2250 | struct sock *sk = sock->sk; | |
2251 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2252 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2253 | u32 dst_portid; |
d629b836 | 2254 | u32 dst_group; |
1da177e4 LT |
2255 | struct sk_buff *skb; |
2256 | int err; | |
2257 | struct scm_cookie scm; | |
2d7a85f4 | 2258 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2259 | |
2260 | if (msg->msg_flags&MSG_OOB) | |
2261 | return -EOPNOTSUPP; | |
2262 | ||
16e57262 | 2263 | if (NULL == siocb->scm) |
1da177e4 | 2264 | siocb->scm = &scm; |
16e57262 | 2265 | |
e0e3cea4 | 2266 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2267 | if (err < 0) |
2268 | return err; | |
2269 | ||
2270 | if (msg->msg_namelen) { | |
b47030c7 | 2271 | err = -EINVAL; |
1da177e4 | 2272 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2273 | goto out; |
15e47304 | 2274 | dst_portid = addr->nl_pid; |
d629b836 | 2275 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2276 | err = -EPERM; |
15e47304 | 2277 | if ((dst_group || dst_portid) && |
5187cd05 | 2278 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2279 | goto out; |
2d7a85f4 | 2280 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2281 | } else { |
15e47304 | 2282 | dst_portid = nlk->dst_portid; |
d629b836 | 2283 | dst_group = nlk->dst_group; |
1da177e4 LT |
2284 | } |
2285 | ||
15e47304 | 2286 | if (!nlk->portid) { |
1da177e4 LT |
2287 | err = netlink_autobind(sock); |
2288 | if (err) | |
2289 | goto out; | |
2290 | } | |
2291 | ||
5fd96123 | 2292 | if (netlink_tx_is_mmaped(sk) && |
c0371da6 | 2293 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 PM |
2294 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
2295 | siocb); | |
2296 | goto out; | |
2297 | } | |
2298 | ||
1da177e4 LT |
2299 | err = -EMSGSIZE; |
2300 | if (len > sk->sk_sndbuf - 32) | |
2301 | goto out; | |
2302 | err = -ENOBUFS; | |
3a36515f | 2303 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2304 | if (skb == NULL) |
1da177e4 LT |
2305 | goto out; |
2306 | ||
15e47304 | 2307 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2308 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2309 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2310 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2311 | |
1da177e4 | 2312 | err = -EFAULT; |
6ce8e9ce | 2313 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2314 | kfree_skb(skb); |
2315 | goto out; | |
2316 | } | |
2317 | ||
2318 | err = security_netlink_send(sk, skb); | |
2319 | if (err) { | |
2320 | kfree_skb(skb); | |
2321 | goto out; | |
2322 | } | |
2323 | ||
d629b836 | 2324 | if (dst_group) { |
1da177e4 | 2325 | atomic_inc(&skb->users); |
15e47304 | 2326 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2327 | } |
15e47304 | 2328 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2329 | |
2330 | out: | |
b47030c7 | 2331 | scm_destroy(siocb->scm); |
1da177e4 LT |
2332 | return err; |
2333 | } | |
2334 | ||
2335 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2336 | struct msghdr *msg, size_t len, | |
2337 | int flags) | |
2338 | { | |
2339 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2340 | struct scm_cookie scm; | |
2341 | struct sock *sk = sock->sk; | |
2342 | struct netlink_sock *nlk = nlk_sk(sk); | |
2343 | int noblock = flags&MSG_DONTWAIT; | |
2344 | size_t copied; | |
68d6ac6d | 2345 | struct sk_buff *skb, *data_skb; |
b44d211e | 2346 | int err, ret; |
1da177e4 LT |
2347 | |
2348 | if (flags&MSG_OOB) | |
2349 | return -EOPNOTSUPP; | |
2350 | ||
2351 | copied = 0; | |
2352 | ||
6ac552fd PM |
2353 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2354 | if (skb == NULL) | |
1da177e4 LT |
2355 | goto out; |
2356 | ||
68d6ac6d JB |
2357 | data_skb = skb; |
2358 | ||
1dacc76d JB |
2359 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2360 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2361 | /* |
68d6ac6d JB |
2362 | * If this skb has a frag_list, then here that means that we |
2363 | * will have to use the frag_list skb's data for compat tasks | |
2364 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2365 | * |
68d6ac6d JB |
2366 | * If we need to send the compat skb, assign it to the |
2367 | * 'data_skb' variable so that it will be used below for data | |
2368 | * copying. We keep 'skb' for everything else, including | |
2369 | * freeing both later. | |
1dacc76d | 2370 | */ |
68d6ac6d JB |
2371 | if (flags & MSG_CMSG_COMPAT) |
2372 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2373 | } |
2374 | #endif | |
2375 | ||
9063e21f ED |
2376 | /* Record the max length of recvmsg() calls for future allocations */ |
2377 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2378 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2379 | 16384); | |
2380 | ||
68d6ac6d | 2381 | copied = data_skb->len; |
1da177e4 LT |
2382 | if (len < copied) { |
2383 | msg->msg_flags |= MSG_TRUNC; | |
2384 | copied = len; | |
2385 | } | |
2386 | ||
68d6ac6d | 2387 | skb_reset_transport_header(data_skb); |
51f3d02b | 2388 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2389 | |
2390 | if (msg->msg_name) { | |
342dfc30 | 2391 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2392 | addr->nl_family = AF_NETLINK; |
2393 | addr->nl_pad = 0; | |
15e47304 | 2394 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2395 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2396 | msg->msg_namelen = sizeof(*addr); |
2397 | } | |
2398 | ||
cc9a06cd PM |
2399 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2400 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2401 | ||
1da177e4 LT |
2402 | if (NULL == siocb->scm) { |
2403 | memset(&scm, 0, sizeof(scm)); | |
2404 | siocb->scm = &scm; | |
2405 | } | |
2406 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2407 | if (flags & MSG_TRUNC) |
68d6ac6d | 2408 | copied = data_skb->len; |
daa3766e | 2409 | |
1da177e4 LT |
2410 | skb_free_datagram(sk, skb); |
2411 | ||
16b304f3 PS |
2412 | if (nlk->cb_running && |
2413 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2414 | ret = netlink_dump(sk); |
2415 | if (ret) { | |
ac30ef83 | 2416 | sk->sk_err = -ret; |
b44d211e AV |
2417 | sk->sk_error_report(sk); |
2418 | } | |
2419 | } | |
1da177e4 LT |
2420 | |
2421 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2422 | out: |
2423 | netlink_rcv_wake(sk); | |
2424 | return err ? : copied; | |
2425 | } | |
2426 | ||
676d2369 | 2427 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2428 | { |
cd40b7d3 | 2429 | BUG(); |
1da177e4 LT |
2430 | } |
2431 | ||
2432 | /* | |
746fac4d | 2433 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2434 | * complete set of kernel non-blocking support for message |
2435 | * queueing. | |
2436 | */ | |
2437 | ||
2438 | struct sock * | |
9f00d977 PNA |
2439 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2440 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2441 | { |
2442 | struct socket *sock; | |
2443 | struct sock *sk; | |
77247bbb | 2444 | struct netlink_sock *nlk; |
5c398dc8 | 2445 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2446 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2447 | unsigned int groups; | |
1da177e4 | 2448 | |
fab2caf6 | 2449 | BUG_ON(!nl_table); |
1da177e4 | 2450 | |
6ac552fd | 2451 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2452 | return NULL; |
2453 | ||
2454 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2455 | return NULL; | |
2456 | ||
23fe1866 PE |
2457 | /* |
2458 | * We have to just have a reference on the net from sk, but don't | |
2459 | * get_net it. Besides, we cannot get and then put the net here. | |
2460 | * So we create one inside init_net and the move it to net. | |
2461 | */ | |
2462 | ||
2463 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2464 | goto out_sock_release_nosk; | |
2465 | ||
2466 | sk = sock->sk; | |
edf02087 | 2467 | sk_change_net(sk, net); |
4fdb3bb7 | 2468 | |
a31f2d17 | 2469 | if (!cfg || cfg->groups < 32) |
4277a083 | 2470 | groups = 32; |
a31f2d17 PNA |
2471 | else |
2472 | groups = cfg->groups; | |
4277a083 | 2473 | |
5c398dc8 | 2474 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2475 | if (!listeners) |
2476 | goto out_sock_release; | |
2477 | ||
1da177e4 | 2478 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2479 | if (cfg && cfg->input) |
2480 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2481 | |
b4b51029 | 2482 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2483 | goto out_sock_release; |
4fdb3bb7 | 2484 | |
77247bbb PM |
2485 | nlk = nlk_sk(sk); |
2486 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2487 | |
4fdb3bb7 | 2488 | netlink_table_grab(); |
b4b51029 EB |
2489 | if (!nl_table[unit].registered) { |
2490 | nl_table[unit].groups = groups; | |
5c398dc8 | 2491 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2492 | nl_table[unit].cb_mutex = cb_mutex; |
2493 | nl_table[unit].module = module; | |
9785e10a PNA |
2494 | if (cfg) { |
2495 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2496 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2497 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2498 | if (cfg->compare) |
2499 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2500 | } |
b4b51029 | 2501 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2502 | } else { |
2503 | kfree(listeners); | |
869e58f8 | 2504 | nl_table[unit].registered++; |
b4b51029 | 2505 | } |
4fdb3bb7 | 2506 | netlink_table_ungrab(); |
77247bbb PM |
2507 | return sk; |
2508 | ||
4fdb3bb7 | 2509 | out_sock_release: |
4277a083 | 2510 | kfree(listeners); |
9dfbec1f | 2511 | netlink_kernel_release(sk); |
23fe1866 PE |
2512 | return NULL; |
2513 | ||
2514 | out_sock_release_nosk: | |
4fdb3bb7 | 2515 | sock_release(sock); |
77247bbb | 2516 | return NULL; |
1da177e4 | 2517 | } |
9f00d977 | 2518 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2519 | |
2520 | void | |
2521 | netlink_kernel_release(struct sock *sk) | |
2522 | { | |
edf02087 | 2523 | sk_release_kernel(sk); |
b7c6ba6e DL |
2524 | } |
2525 | EXPORT_SYMBOL(netlink_kernel_release); | |
2526 | ||
d136f1bd | 2527 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2528 | { |
5c398dc8 | 2529 | struct listeners *new, *old; |
b4ff4f04 | 2530 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2531 | |
2532 | if (groups < 32) | |
2533 | groups = 32; | |
2534 | ||
b4ff4f04 | 2535 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2536 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2537 | if (!new) | |
d136f1bd | 2538 | return -ENOMEM; |
6d772ac5 | 2539 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2540 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2541 | rcu_assign_pointer(tbl->listeners, new); | |
2542 | ||
37b6b935 | 2543 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2544 | } |
2545 | tbl->groups = groups; | |
2546 | ||
d136f1bd JB |
2547 | return 0; |
2548 | } | |
2549 | ||
2550 | /** | |
2551 | * netlink_change_ngroups - change number of multicast groups | |
2552 | * | |
2553 | * This changes the number of multicast groups that are available | |
2554 | * on a certain netlink family. Note that it is not possible to | |
2555 | * change the number of groups to below 32. Also note that it does | |
2556 | * not implicitly call netlink_clear_multicast_users() when the | |
2557 | * number of groups is reduced. | |
2558 | * | |
2559 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2560 | * @groups: The new number of groups. | |
2561 | */ | |
2562 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2563 | { | |
2564 | int err; | |
2565 | ||
2566 | netlink_table_grab(); | |
2567 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2568 | netlink_table_ungrab(); |
d136f1bd | 2569 | |
b4ff4f04 JB |
2570 | return err; |
2571 | } | |
b4ff4f04 | 2572 | |
b8273570 JB |
2573 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2574 | { | |
2575 | struct sock *sk; | |
b8273570 JB |
2576 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2577 | ||
b67bfe0d | 2578 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2579 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2580 | } | |
2581 | ||
a46621a3 | 2582 | struct nlmsghdr * |
15e47304 | 2583 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2584 | { |
2585 | struct nlmsghdr *nlh; | |
573ce260 | 2586 | int size = nlmsg_msg_size(len); |
a46621a3 | 2587 | |
23b45672 | 2588 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2589 | nlh->nlmsg_type = type; |
2590 | nlh->nlmsg_len = size; | |
2591 | nlh->nlmsg_flags = flags; | |
15e47304 | 2592 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2593 | nlh->nlmsg_seq = seq; |
2594 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2595 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2596 | return nlh; |
2597 | } | |
2598 | EXPORT_SYMBOL(__nlmsg_put); | |
2599 | ||
1da177e4 LT |
2600 | /* |
2601 | * It looks a bit ugly. | |
2602 | * It would be better to create kernel thread. | |
2603 | */ | |
2604 | ||
2605 | static int netlink_dump(struct sock *sk) | |
2606 | { | |
2607 | struct netlink_sock *nlk = nlk_sk(sk); | |
2608 | struct netlink_callback *cb; | |
c7ac8679 | 2609 | struct sk_buff *skb = NULL; |
1da177e4 | 2610 | struct nlmsghdr *nlh; |
bf8b79e4 | 2611 | int len, err = -ENOBUFS; |
c7ac8679 | 2612 | int alloc_size; |
1da177e4 | 2613 | |
af65bdfc | 2614 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2615 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2616 | err = -EINVAL; |
2617 | goto errout_skb; | |
1da177e4 LT |
2618 | } |
2619 | ||
16b304f3 | 2620 | cb = &nlk->cb; |
c7ac8679 GR |
2621 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2622 | ||
f9c22888 PM |
2623 | if (!netlink_rx_is_mmaped(sk) && |
2624 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2625 | goto errout_skb; | |
9063e21f ED |
2626 | |
2627 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2628 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2629 | * to reduce number of system calls on dump operations, if user | |
2630 | * ever provided a big enough buffer. | |
2631 | */ | |
2632 | if (alloc_size < nlk->max_recvmsg_len) { | |
2633 | skb = netlink_alloc_skb(sk, | |
2634 | nlk->max_recvmsg_len, | |
2635 | nlk->portid, | |
2636 | GFP_KERNEL | | |
2637 | __GFP_NOWARN | | |
2638 | __GFP_NORETRY); | |
2639 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2640 | if (skb) | |
2641 | skb_reserve(skb, skb_tailroom(skb) - | |
2642 | nlk->max_recvmsg_len); | |
2643 | } | |
2644 | if (!skb) | |
2645 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2646 | GFP_KERNEL); | |
c7ac8679 | 2647 | if (!skb) |
c63d6ea3 | 2648 | goto errout_skb; |
f9c22888 | 2649 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2650 | |
1da177e4 LT |
2651 | len = cb->dump(skb, cb); |
2652 | ||
2653 | if (len > 0) { | |
af65bdfc | 2654 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2655 | |
2656 | if (sk_filter(sk, skb)) | |
2657 | kfree_skb(skb); | |
4a7e7c2a ED |
2658 | else |
2659 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2660 | return 0; |
2661 | } | |
2662 | ||
bf8b79e4 TG |
2663 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2664 | if (!nlh) | |
2665 | goto errout_skb; | |
2666 | ||
670dc283 JB |
2667 | nl_dump_check_consistent(cb, nlh); |
2668 | ||
bf8b79e4 TG |
2669 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2670 | ||
b1153f29 SH |
2671 | if (sk_filter(sk, skb)) |
2672 | kfree_skb(skb); | |
4a7e7c2a ED |
2673 | else |
2674 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2675 | |
a8f74b22 TG |
2676 | if (cb->done) |
2677 | cb->done(cb); | |
1da177e4 | 2678 | |
16b304f3 PS |
2679 | nlk->cb_running = false; |
2680 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2681 | module_put(cb->module); |
16b304f3 | 2682 | consume_skb(cb->skb); |
1da177e4 | 2683 | return 0; |
1797754e | 2684 | |
bf8b79e4 | 2685 | errout_skb: |
af65bdfc | 2686 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2687 | kfree_skb(skb); |
bf8b79e4 | 2688 | return err; |
1da177e4 LT |
2689 | } |
2690 | ||
6dc878a8 G |
2691 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2692 | const struct nlmsghdr *nlh, | |
2693 | struct netlink_dump_control *control) | |
1da177e4 LT |
2694 | { |
2695 | struct netlink_callback *cb; | |
2696 | struct sock *sk; | |
2697 | struct netlink_sock *nlk; | |
b44d211e | 2698 | int ret; |
1da177e4 | 2699 | |
f9c22888 PM |
2700 | /* Memory mapped dump requests need to be copied to avoid looping |
2701 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2702 | * a reference to the skb. | |
2703 | */ | |
2704 | if (netlink_skb_is_mmaped(skb)) { | |
2705 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2706 | if (skb == NULL) |
f9c22888 | 2707 | return -ENOBUFS; |
f9c22888 PM |
2708 | } else |
2709 | atomic_inc(&skb->users); | |
2710 | ||
15e47304 | 2711 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2712 | if (sk == NULL) { |
16b304f3 PS |
2713 | ret = -ECONNREFUSED; |
2714 | goto error_free; | |
1da177e4 | 2715 | } |
6dc878a8 | 2716 | |
16b304f3 | 2717 | nlk = nlk_sk(sk); |
af65bdfc | 2718 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2719 | /* A dump is in progress... */ |
16b304f3 | 2720 | if (nlk->cb_running) { |
6dc878a8 | 2721 | ret = -EBUSY; |
16b304f3 | 2722 | goto error_unlock; |
1da177e4 | 2723 | } |
6dc878a8 | 2724 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2725 | if (!try_module_get(control->module)) { |
6dc878a8 | 2726 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2727 | goto error_unlock; |
6dc878a8 G |
2728 | } |
2729 | ||
16b304f3 PS |
2730 | cb = &nlk->cb; |
2731 | memset(cb, 0, sizeof(*cb)); | |
2732 | cb->dump = control->dump; | |
2733 | cb->done = control->done; | |
2734 | cb->nlh = nlh; | |
2735 | cb->data = control->data; | |
2736 | cb->module = control->module; | |
2737 | cb->min_dump_alloc = control->min_dump_alloc; | |
2738 | cb->skb = skb; | |
2739 | ||
2740 | nlk->cb_running = true; | |
2741 | ||
af65bdfc | 2742 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2743 | |
b44d211e | 2744 | ret = netlink_dump(sk); |
1da177e4 | 2745 | sock_put(sk); |
5c58298c | 2746 | |
b44d211e AV |
2747 | if (ret) |
2748 | return ret; | |
2749 | ||
5c58298c DL |
2750 | /* We successfully started a dump, by returning -EINTR we |
2751 | * signal not to send ACK even if it was requested. | |
2752 | */ | |
2753 | return -EINTR; | |
16b304f3 PS |
2754 | |
2755 | error_unlock: | |
2756 | sock_put(sk); | |
2757 | mutex_unlock(nlk->cb_mutex); | |
2758 | error_free: | |
2759 | kfree_skb(skb); | |
2760 | return ret; | |
1da177e4 | 2761 | } |
6dc878a8 | 2762 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2763 | |
2764 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2765 | { | |
2766 | struct sk_buff *skb; | |
2767 | struct nlmsghdr *rep; | |
2768 | struct nlmsgerr *errmsg; | |
339bf98f | 2769 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2770 | |
339bf98f TG |
2771 | /* error messages get the original request appened */ |
2772 | if (err) | |
2773 | payload += nlmsg_len(nlh); | |
1da177e4 | 2774 | |
f9c22888 PM |
2775 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2776 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2777 | if (!skb) { |
2778 | struct sock *sk; | |
2779 | ||
3b1e0a65 | 2780 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2781 | in_skb->sk->sk_protocol, |
15e47304 | 2782 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2783 | if (sk) { |
2784 | sk->sk_err = ENOBUFS; | |
2785 | sk->sk_error_report(sk); | |
2786 | sock_put(sk); | |
2787 | } | |
2788 | return; | |
2789 | } | |
2790 | ||
15e47304 | 2791 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2792 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2793 | errmsg = nlmsg_data(rep); |
1da177e4 | 2794 | errmsg->error = err; |
bf8b79e4 | 2795 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2796 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2797 | } |
6ac552fd | 2798 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2799 | |
cd40b7d3 | 2800 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2801 | struct nlmsghdr *)) |
82ace47a | 2802 | { |
82ace47a TG |
2803 | struct nlmsghdr *nlh; |
2804 | int err; | |
2805 | ||
2806 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2807 | int msglen; |
2808 | ||
b529ccf2 | 2809 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2810 | err = 0; |
82ace47a | 2811 | |
ad8e4b75 | 2812 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2813 | return 0; |
2814 | ||
d35b6856 TG |
2815 | /* Only requests are handled by the kernel */ |
2816 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2817 | goto ack; |
45e7ae7f TG |
2818 | |
2819 | /* Skip control messages */ | |
2820 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2821 | goto ack; |
d35b6856 | 2822 | |
1d00a4eb | 2823 | err = cb(skb, nlh); |
5c58298c DL |
2824 | if (err == -EINTR) |
2825 | goto skip; | |
2826 | ||
2827 | ack: | |
d35b6856 | 2828 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2829 | netlink_ack(skb, nlh, err); |
82ace47a | 2830 | |
5c58298c | 2831 | skip: |
6ac552fd | 2832 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2833 | if (msglen > skb->len) |
2834 | msglen = skb->len; | |
2835 | skb_pull(skb, msglen); | |
82ace47a TG |
2836 | } |
2837 | ||
2838 | return 0; | |
2839 | } | |
6ac552fd | 2840 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2841 | |
d387f6ad TG |
2842 | /** |
2843 | * nlmsg_notify - send a notification netlink message | |
2844 | * @sk: netlink socket to use | |
2845 | * @skb: notification message | |
15e47304 | 2846 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2847 | * @group: destination multicast group or 0 |
2848 | * @report: 1 to report back, 0 to disable | |
2849 | * @flags: allocation flags | |
2850 | */ | |
15e47304 | 2851 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2852 | unsigned int group, int report, gfp_t flags) |
2853 | { | |
2854 | int err = 0; | |
2855 | ||
2856 | if (group) { | |
15e47304 | 2857 | int exclude_portid = 0; |
d387f6ad TG |
2858 | |
2859 | if (report) { | |
2860 | atomic_inc(&skb->users); | |
15e47304 | 2861 | exclude_portid = portid; |
d387f6ad TG |
2862 | } |
2863 | ||
1ce85fe4 PNA |
2864 | /* errors reported via destination sk->sk_err, but propagate |
2865 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2866 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2867 | } |
2868 | ||
1ce85fe4 PNA |
2869 | if (report) { |
2870 | int err2; | |
2871 | ||
15e47304 | 2872 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2873 | if (!err || err == -ESRCH) |
2874 | err = err2; | |
2875 | } | |
d387f6ad TG |
2876 | |
2877 | return err; | |
2878 | } | |
6ac552fd | 2879 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2880 | |
1da177e4 LT |
2881 | #ifdef CONFIG_PROC_FS |
2882 | struct nl_seq_iter { | |
e372c414 | 2883 | struct seq_net_private p; |
1da177e4 LT |
2884 | int link; |
2885 | int hash_idx; | |
2886 | }; | |
2887 | ||
2888 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2889 | { | |
2890 | struct nl_seq_iter *iter = seq->private; | |
2891 | int i, j; | |
e341694e | 2892 | struct netlink_sock *nlk; |
1da177e4 | 2893 | struct sock *s; |
1da177e4 LT |
2894 | loff_t off = 0; |
2895 | ||
6ac552fd | 2896 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e | 2897 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2898 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
e341694e TG |
2899 | |
2900 | for (j = 0; j < tbl->size; j++) { | |
88d6ed15 TG |
2901 | struct rhash_head *node; |
2902 | ||
2903 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { | |
e341694e | 2904 | s = (struct sock *)nlk; |
1da177e4 | 2905 | |
1218854a | 2906 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2907 | continue; |
1da177e4 LT |
2908 | if (off == pos) { |
2909 | iter->link = i; | |
2910 | iter->hash_idx = j; | |
2911 | return s; | |
2912 | } | |
2913 | ++off; | |
2914 | } | |
2915 | } | |
2916 | } | |
2917 | return NULL; | |
2918 | } | |
2919 | ||
2920 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
78fd1d0a | 2921 | __acquires(nl_table_lock) __acquires(RCU) |
1da177e4 | 2922 | { |
78fd1d0a | 2923 | read_lock(&nl_table_lock); |
e341694e | 2924 | rcu_read_lock(); |
1da177e4 LT |
2925 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
2926 | } | |
2927 | ||
2928 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2929 | { | |
78fd1d0a | 2930 | struct rhashtable *ht; |
88d6ed15 TG |
2931 | const struct bucket_table *tbl; |
2932 | struct rhash_head *node; | |
e341694e | 2933 | struct netlink_sock *nlk; |
1da177e4 | 2934 | struct nl_seq_iter *iter; |
da12c90e | 2935 | struct net *net; |
1da177e4 LT |
2936 | int i, j; |
2937 | ||
2938 | ++*pos; | |
2939 | ||
2940 | if (v == SEQ_START_TOKEN) | |
2941 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2942 | |
da12c90e | 2943 | net = seq_file_net(seq); |
b4b51029 | 2944 | iter = seq->private; |
e341694e TG |
2945 | nlk = v; |
2946 | ||
78fd1d0a TG |
2947 | i = iter->link; |
2948 | ht = &nl_table[i].hash; | |
88d6ed15 TG |
2949 | tbl = rht_dereference_rcu(ht->tbl, ht); |
2950 | rht_for_each_entry_rcu_continue(nlk, node, nlk->node.next, tbl, iter->hash_idx, node) | |
e341694e TG |
2951 | if (net_eq(sock_net((struct sock *)nlk), net)) |
2952 | return nlk; | |
1da177e4 | 2953 | |
1da177e4 LT |
2954 | j = iter->hash_idx + 1; |
2955 | ||
2956 | do { | |
da12c90e | 2957 | |
e341694e | 2958 | for (; j < tbl->size; j++) { |
88d6ed15 | 2959 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { |
e341694e TG |
2960 | if (net_eq(sock_net((struct sock *)nlk), net)) { |
2961 | iter->link = i; | |
2962 | iter->hash_idx = j; | |
2963 | return nlk; | |
2964 | } | |
1da177e4 LT |
2965 | } |
2966 | } | |
2967 | ||
2968 | j = 0; | |
2969 | } while (++i < MAX_LINKS); | |
2970 | ||
2971 | return NULL; | |
2972 | } | |
2973 | ||
2974 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
78fd1d0a | 2975 | __releases(RCU) __releases(nl_table_lock) |
1da177e4 | 2976 | { |
e341694e | 2977 | rcu_read_unlock(); |
78fd1d0a | 2978 | read_unlock(&nl_table_lock); |
1da177e4 LT |
2979 | } |
2980 | ||
2981 | ||
2982 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2983 | { | |
658cb354 | 2984 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2985 | seq_puts(seq, |
2986 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2987 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2988 | } else { |
1da177e4 LT |
2989 | struct sock *s = v; |
2990 | struct netlink_sock *nlk = nlk_sk(s); | |
2991 | ||
16b304f3 | 2992 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2993 | s, |
2994 | s->sk_protocol, | |
15e47304 | 2995 | nlk->portid, |
513c2500 | 2996 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2997 | sk_rmem_alloc_get(s), |
2998 | sk_wmem_alloc_get(s), | |
16b304f3 | 2999 | nlk->cb_running, |
38938bfe | 3000 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
3001 | atomic_read(&s->sk_drops), |
3002 | sock_i_ino(s) | |
1da177e4 LT |
3003 | ); |
3004 | ||
3005 | } | |
3006 | return 0; | |
3007 | } | |
3008 | ||
56b3d975 | 3009 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3010 | .start = netlink_seq_start, |
3011 | .next = netlink_seq_next, | |
3012 | .stop = netlink_seq_stop, | |
3013 | .show = netlink_seq_show, | |
3014 | }; | |
3015 | ||
3016 | ||
3017 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3018 | { | |
e372c414 DL |
3019 | return seq_open_net(inode, file, &netlink_seq_ops, |
3020 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3021 | } |
3022 | ||
da7071d7 | 3023 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3024 | .owner = THIS_MODULE, |
3025 | .open = netlink_seq_open, | |
3026 | .read = seq_read, | |
3027 | .llseek = seq_lseek, | |
e372c414 | 3028 | .release = seq_release_net, |
1da177e4 LT |
3029 | }; |
3030 | ||
3031 | #endif | |
3032 | ||
3033 | int netlink_register_notifier(struct notifier_block *nb) | |
3034 | { | |
e041c683 | 3035 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3036 | } |
6ac552fd | 3037 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3038 | |
3039 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3040 | { | |
e041c683 | 3041 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3042 | } |
6ac552fd | 3043 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3044 | |
90ddc4f0 | 3045 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3046 | .family = PF_NETLINK, |
3047 | .owner = THIS_MODULE, | |
3048 | .release = netlink_release, | |
3049 | .bind = netlink_bind, | |
3050 | .connect = netlink_connect, | |
3051 | .socketpair = sock_no_socketpair, | |
3052 | .accept = sock_no_accept, | |
3053 | .getname = netlink_getname, | |
9652e931 | 3054 | .poll = netlink_poll, |
1da177e4 LT |
3055 | .ioctl = sock_no_ioctl, |
3056 | .listen = sock_no_listen, | |
3057 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3058 | .setsockopt = netlink_setsockopt, |
3059 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3060 | .sendmsg = netlink_sendmsg, |
3061 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3062 | .mmap = netlink_mmap, |
1da177e4 LT |
3063 | .sendpage = sock_no_sendpage, |
3064 | }; | |
3065 | ||
ec1b4cf7 | 3066 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3067 | .family = PF_NETLINK, |
3068 | .create = netlink_create, | |
3069 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3070 | }; | |
3071 | ||
4665079c | 3072 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3073 | { |
3074 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3075 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3076 | return -ENOMEM; |
3077 | #endif | |
3078 | return 0; | |
3079 | } | |
3080 | ||
4665079c | 3081 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3082 | { |
3083 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3084 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3085 | #endif |
3086 | } | |
3087 | ||
b963ea89 DM |
3088 | static void __init netlink_add_usersock_entry(void) |
3089 | { | |
5c398dc8 | 3090 | struct listeners *listeners; |
b963ea89 DM |
3091 | int groups = 32; |
3092 | ||
5c398dc8 | 3093 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3094 | if (!listeners) |
5c398dc8 | 3095 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3096 | |
3097 | netlink_table_grab(); | |
3098 | ||
3099 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3100 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3101 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3102 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3103 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3104 | |
3105 | netlink_table_ungrab(); | |
3106 | } | |
3107 | ||
022cbae6 | 3108 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3109 | .init = netlink_net_init, |
3110 | .exit = netlink_net_exit, | |
3111 | }; | |
3112 | ||
1da177e4 LT |
3113 | static int __init netlink_proto_init(void) |
3114 | { | |
1da177e4 | 3115 | int i; |
1da177e4 | 3116 | int err = proto_register(&netlink_proto, 0); |
e341694e TG |
3117 | struct rhashtable_params ht_params = { |
3118 | .head_offset = offsetof(struct netlink_sock, node), | |
3119 | .key_offset = offsetof(struct netlink_sock, portid), | |
3120 | .key_len = sizeof(u32), /* portid */ | |
7f19fc5e | 3121 | .hashfn = jhash, |
e341694e TG |
3122 | .max_shift = 16, /* 64K */ |
3123 | .grow_decision = rht_grow_above_75, | |
3124 | .shrink_decision = rht_shrink_below_30, | |
97127566 | 3125 | #ifdef CONFIG_PROVE_LOCKING |
e341694e | 3126 | .mutex_is_held = lockdep_nl_sk_hash_is_held, |
97127566 | 3127 | #endif |
e341694e | 3128 | }; |
1da177e4 LT |
3129 | |
3130 | if (err != 0) | |
3131 | goto out; | |
3132 | ||
fab25745 | 3133 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3134 | |
0da974f4 | 3135 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3136 | if (!nl_table) |
3137 | goto panic; | |
1da177e4 | 3138 | |
1da177e4 | 3139 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e TG |
3140 | if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) { |
3141 | while (--i > 0) | |
3142 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3143 | kfree(nl_table); |
fab2caf6 | 3144 | goto panic; |
1da177e4 | 3145 | } |
1da177e4 LT |
3146 | } |
3147 | ||
bcbde0d4 DB |
3148 | INIT_LIST_HEAD(&netlink_tap_all); |
3149 | ||
b963ea89 DM |
3150 | netlink_add_usersock_entry(); |
3151 | ||
1da177e4 | 3152 | sock_register(&netlink_family_ops); |
b4b51029 | 3153 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3154 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3155 | rtnetlink_init(); |
3156 | out: | |
3157 | return err; | |
fab2caf6 AM |
3158 | panic: |
3159 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3160 | } |
3161 | ||
1da177e4 | 3162 | core_initcall(netlink_proto_init); |