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 | |
e041c683 | 117 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 118 | |
bcbde0d4 DB |
119 | static DEFINE_SPINLOCK(netlink_tap_lock); |
120 | static struct list_head netlink_tap_all __read_mostly; | |
121 | ||
b57ef81f | 122 | static inline u32 netlink_group_mask(u32 group) |
d629b836 PM |
123 | { |
124 | return group ? 1 << (group - 1) : 0; | |
125 | } | |
126 | ||
bcbde0d4 DB |
127 | int netlink_add_tap(struct netlink_tap *nt) |
128 | { | |
129 | if (unlikely(nt->dev->type != ARPHRD_NETLINK)) | |
130 | return -EINVAL; | |
131 | ||
132 | spin_lock(&netlink_tap_lock); | |
133 | list_add_rcu(&nt->list, &netlink_tap_all); | |
134 | spin_unlock(&netlink_tap_lock); | |
135 | ||
fcd4d35e | 136 | __module_get(nt->module); |
bcbde0d4 DB |
137 | |
138 | return 0; | |
139 | } | |
140 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
141 | ||
2173f8d9 | 142 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
143 | { |
144 | bool found = false; | |
145 | struct netlink_tap *tmp; | |
146 | ||
147 | spin_lock(&netlink_tap_lock); | |
148 | ||
149 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
150 | if (nt == tmp) { | |
151 | list_del_rcu(&nt->list); | |
152 | found = true; | |
153 | goto out; | |
154 | } | |
155 | } | |
156 | ||
157 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
158 | out: | |
159 | spin_unlock(&netlink_tap_lock); | |
160 | ||
161 | if (found && nt->module) | |
162 | module_put(nt->module); | |
163 | ||
164 | return found ? 0 : -ENODEV; | |
165 | } | |
bcbde0d4 DB |
166 | |
167 | int netlink_remove_tap(struct netlink_tap *nt) | |
168 | { | |
169 | int ret; | |
170 | ||
171 | ret = __netlink_remove_tap(nt); | |
172 | synchronize_net(); | |
173 | ||
174 | return ret; | |
175 | } | |
176 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
177 | ||
5ffd5cdd DB |
178 | static bool netlink_filter_tap(const struct sk_buff *skb) |
179 | { | |
180 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
181 | |
182 | /* We take the more conservative approach and | |
183 | * whitelist socket protocols that may pass. | |
184 | */ | |
185 | switch (sk->sk_protocol) { | |
186 | case NETLINK_ROUTE: | |
187 | case NETLINK_USERSOCK: | |
188 | case NETLINK_SOCK_DIAG: | |
189 | case NETLINK_NFLOG: | |
190 | case NETLINK_XFRM: | |
191 | case NETLINK_FIB_LOOKUP: | |
192 | case NETLINK_NETFILTER: | |
193 | case NETLINK_GENERIC: | |
498044bb | 194 | return true; |
5ffd5cdd DB |
195 | } |
196 | ||
498044bb | 197 | return false; |
5ffd5cdd DB |
198 | } |
199 | ||
bcbde0d4 DB |
200 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
201 | struct net_device *dev) | |
202 | { | |
203 | struct sk_buff *nskb; | |
5ffd5cdd | 204 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
205 | int ret = -ENOMEM; |
206 | ||
207 | dev_hold(dev); | |
208 | nskb = skb_clone(skb, GFP_ATOMIC); | |
209 | if (nskb) { | |
210 | nskb->dev = dev; | |
5ffd5cdd | 211 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
212 | nskb->pkt_type = netlink_is_kernel(sk) ? |
213 | PACKET_KERNEL : PACKET_USER; | |
4e48ed88 | 214 | skb_reset_network_header(nskb); |
bcbde0d4 DB |
215 | ret = dev_queue_xmit(nskb); |
216 | if (unlikely(ret > 0)) | |
217 | ret = net_xmit_errno(ret); | |
218 | } | |
219 | ||
220 | dev_put(dev); | |
221 | return ret; | |
222 | } | |
223 | ||
224 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
225 | { | |
226 | int ret; | |
227 | struct netlink_tap *tmp; | |
228 | ||
5ffd5cdd DB |
229 | if (!netlink_filter_tap(skb)) |
230 | return; | |
231 | ||
bcbde0d4 DB |
232 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
233 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
234 | if (unlikely(ret)) | |
235 | break; | |
236 | } | |
237 | } | |
238 | ||
239 | static void netlink_deliver_tap(struct sk_buff *skb) | |
240 | { | |
241 | rcu_read_lock(); | |
242 | ||
243 | if (unlikely(!list_empty(&netlink_tap_all))) | |
244 | __netlink_deliver_tap(skb); | |
245 | ||
246 | rcu_read_unlock(); | |
247 | } | |
248 | ||
73bfd370 DB |
249 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
250 | struct sk_buff *skb) | |
251 | { | |
252 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
253 | netlink_deliver_tap(skb); | |
254 | } | |
255 | ||
cd1df525 PM |
256 | static void netlink_overrun(struct sock *sk) |
257 | { | |
258 | struct netlink_sock *nlk = nlk_sk(sk); | |
259 | ||
260 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
261 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
262 | sk->sk_err = ENOBUFS; | |
263 | sk->sk_error_report(sk); | |
264 | } | |
265 | } | |
266 | atomic_inc(&sk->sk_drops); | |
267 | } | |
268 | ||
269 | static void netlink_rcv_wake(struct sock *sk) | |
270 | { | |
271 | struct netlink_sock *nlk = nlk_sk(sk); | |
272 | ||
273 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
274 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
275 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
276 | wake_up_interruptible(&nlk->wait); | |
277 | } | |
278 | ||
ccdfcc39 | 279 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
280 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
281 | { | |
282 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
283 | } | |
284 | ||
f9c22888 PM |
285 | static bool netlink_rx_is_mmaped(struct sock *sk) |
286 | { | |
287 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
288 | } | |
289 | ||
5fd96123 PM |
290 | static bool netlink_tx_is_mmaped(struct sock *sk) |
291 | { | |
292 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
293 | } | |
294 | ||
ccdfcc39 PM |
295 | static __pure struct page *pgvec_to_page(const void *addr) |
296 | { | |
297 | if (is_vmalloc_addr(addr)) | |
298 | return vmalloc_to_page(addr); | |
299 | else | |
300 | return virt_to_page(addr); | |
301 | } | |
302 | ||
303 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
304 | { | |
305 | unsigned int i; | |
306 | ||
307 | for (i = 0; i < len; i++) { | |
308 | if (pg_vec[i] != NULL) { | |
309 | if (is_vmalloc_addr(pg_vec[i])) | |
310 | vfree(pg_vec[i]); | |
311 | else | |
312 | free_pages((unsigned long)pg_vec[i], order); | |
313 | } | |
314 | } | |
315 | kfree(pg_vec); | |
316 | } | |
317 | ||
318 | static void *alloc_one_pg_vec_page(unsigned long order) | |
319 | { | |
320 | void *buffer; | |
321 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
322 | __GFP_NOWARN | __GFP_NORETRY; | |
323 | ||
324 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
325 | if (buffer != NULL) | |
326 | return buffer; | |
327 | ||
328 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
329 | if (buffer != NULL) | |
330 | return buffer; | |
331 | ||
332 | gfp_flags &= ~__GFP_NORETRY; | |
333 | return (void *)__get_free_pages(gfp_flags, order); | |
334 | } | |
335 | ||
336 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
337 | struct nl_mmap_req *req, unsigned int order) | |
338 | { | |
339 | unsigned int block_nr = req->nm_block_nr; | |
340 | unsigned int i; | |
8a849bb7 | 341 | void **pg_vec; |
ccdfcc39 PM |
342 | |
343 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
344 | if (pg_vec == NULL) | |
345 | return NULL; | |
346 | ||
347 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 348 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
349 | if (pg_vec[i] == NULL) |
350 | goto err1; | |
351 | } | |
352 | ||
353 | return pg_vec; | |
354 | err1: | |
355 | free_pg_vec(pg_vec, order, block_nr); | |
356 | return NULL; | |
357 | } | |
358 | ||
359 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
360 | bool closing, bool tx_ring) | |
361 | { | |
362 | struct netlink_sock *nlk = nlk_sk(sk); | |
363 | struct netlink_ring *ring; | |
364 | struct sk_buff_head *queue; | |
365 | void **pg_vec = NULL; | |
366 | unsigned int order = 0; | |
367 | int err; | |
368 | ||
369 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
370 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
371 | ||
372 | if (!closing) { | |
373 | if (atomic_read(&nlk->mapped)) | |
374 | return -EBUSY; | |
375 | if (atomic_read(&ring->pending)) | |
376 | return -EBUSY; | |
377 | } | |
378 | ||
379 | if (req->nm_block_nr) { | |
380 | if (ring->pg_vec != NULL) | |
381 | return -EBUSY; | |
382 | ||
383 | if ((int)req->nm_block_size <= 0) | |
384 | return -EINVAL; | |
74e83b23 | 385 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
386 | return -EINVAL; |
387 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
388 | return -EINVAL; | |
389 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
390 | return -EINVAL; | |
391 | ||
392 | ring->frames_per_block = req->nm_block_size / | |
393 | req->nm_frame_size; | |
394 | if (ring->frames_per_block == 0) | |
395 | return -EINVAL; | |
396 | if (ring->frames_per_block * req->nm_block_nr != | |
397 | req->nm_frame_nr) | |
398 | return -EINVAL; | |
399 | ||
400 | order = get_order(req->nm_block_size); | |
401 | pg_vec = alloc_pg_vec(nlk, req, order); | |
402 | if (pg_vec == NULL) | |
403 | return -ENOMEM; | |
404 | } else { | |
405 | if (req->nm_frame_nr) | |
406 | return -EINVAL; | |
407 | } | |
408 | ||
409 | err = -EBUSY; | |
410 | mutex_lock(&nlk->pg_vec_lock); | |
411 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
412 | err = 0; | |
413 | spin_lock_bh(&queue->lock); | |
414 | ||
415 | ring->frame_max = req->nm_frame_nr - 1; | |
416 | ring->head = 0; | |
417 | ring->frame_size = req->nm_frame_size; | |
418 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
419 | ||
420 | swap(ring->pg_vec_len, req->nm_block_nr); | |
421 | swap(ring->pg_vec_order, order); | |
422 | swap(ring->pg_vec, pg_vec); | |
423 | ||
424 | __skb_queue_purge(queue); | |
425 | spin_unlock_bh(&queue->lock); | |
426 | ||
427 | WARN_ON(atomic_read(&nlk->mapped)); | |
428 | } | |
429 | mutex_unlock(&nlk->pg_vec_lock); | |
430 | ||
431 | if (pg_vec) | |
432 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
433 | return err; | |
434 | } | |
435 | ||
436 | static void netlink_mm_open(struct vm_area_struct *vma) | |
437 | { | |
438 | struct file *file = vma->vm_file; | |
439 | struct socket *sock = file->private_data; | |
440 | struct sock *sk = sock->sk; | |
441 | ||
442 | if (sk) | |
443 | atomic_inc(&nlk_sk(sk)->mapped); | |
444 | } | |
445 | ||
446 | static void netlink_mm_close(struct vm_area_struct *vma) | |
447 | { | |
448 | struct file *file = vma->vm_file; | |
449 | struct socket *sock = file->private_data; | |
450 | struct sock *sk = sock->sk; | |
451 | ||
452 | if (sk) | |
453 | atomic_dec(&nlk_sk(sk)->mapped); | |
454 | } | |
455 | ||
456 | static const struct vm_operations_struct netlink_mmap_ops = { | |
457 | .open = netlink_mm_open, | |
458 | .close = netlink_mm_close, | |
459 | }; | |
460 | ||
461 | static int netlink_mmap(struct file *file, struct socket *sock, | |
462 | struct vm_area_struct *vma) | |
463 | { | |
464 | struct sock *sk = sock->sk; | |
465 | struct netlink_sock *nlk = nlk_sk(sk); | |
466 | struct netlink_ring *ring; | |
467 | unsigned long start, size, expected; | |
468 | unsigned int i; | |
469 | int err = -EINVAL; | |
470 | ||
471 | if (vma->vm_pgoff) | |
472 | return -EINVAL; | |
473 | ||
474 | mutex_lock(&nlk->pg_vec_lock); | |
475 | ||
476 | expected = 0; | |
477 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
478 | if (ring->pg_vec == NULL) | |
479 | continue; | |
480 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
481 | } | |
482 | ||
483 | if (expected == 0) | |
484 | goto out; | |
485 | ||
486 | size = vma->vm_end - vma->vm_start; | |
487 | if (size != expected) | |
488 | goto out; | |
489 | ||
490 | start = vma->vm_start; | |
491 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
492 | if (ring->pg_vec == NULL) | |
493 | continue; | |
494 | ||
495 | for (i = 0; i < ring->pg_vec_len; i++) { | |
496 | struct page *page; | |
497 | void *kaddr = ring->pg_vec[i]; | |
498 | unsigned int pg_num; | |
499 | ||
500 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
501 | page = pgvec_to_page(kaddr); | |
502 | err = vm_insert_page(vma, start, page); | |
503 | if (err < 0) | |
504 | goto out; | |
505 | start += PAGE_SIZE; | |
506 | kaddr += PAGE_SIZE; | |
507 | } | |
508 | } | |
509 | } | |
510 | ||
511 | atomic_inc(&nlk->mapped); | |
512 | vma->vm_ops = &netlink_mmap_ops; | |
513 | err = 0; | |
514 | out: | |
515 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 516 | return err; |
ccdfcc39 | 517 | } |
9652e931 | 518 | |
4682a035 | 519 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr, unsigned int nm_len) |
9652e931 PM |
520 | { |
521 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
522 | struct page *p_start, *p_end; | |
523 | ||
524 | /* First page is flushed through netlink_{get,set}_status */ | |
525 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
4682a035 | 526 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + nm_len - 1); |
9652e931 PM |
527 | while (p_start <= p_end) { |
528 | flush_dcache_page(p_start); | |
529 | p_start++; | |
530 | } | |
531 | #endif | |
532 | } | |
533 | ||
534 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
535 | { | |
536 | smp_rmb(); | |
537 | flush_dcache_page(pgvec_to_page(hdr)); | |
538 | return hdr->nm_status; | |
539 | } | |
540 | ||
541 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
542 | enum nl_mmap_status status) | |
543 | { | |
a18e6a18 | 544 | smp_mb(); |
9652e931 PM |
545 | hdr->nm_status = status; |
546 | flush_dcache_page(pgvec_to_page(hdr)); | |
9652e931 PM |
547 | } |
548 | ||
549 | static struct nl_mmap_hdr * | |
550 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
551 | { | |
552 | unsigned int pg_vec_pos, frame_off; | |
553 | ||
554 | pg_vec_pos = pos / ring->frames_per_block; | |
555 | frame_off = pos % ring->frames_per_block; | |
556 | ||
557 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
558 | } | |
559 | ||
560 | static struct nl_mmap_hdr * | |
561 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
562 | enum nl_mmap_status status) | |
563 | { | |
564 | struct nl_mmap_hdr *hdr; | |
565 | ||
566 | hdr = __netlink_lookup_frame(ring, pos); | |
567 | if (netlink_get_status(hdr) != status) | |
568 | return NULL; | |
569 | ||
570 | return hdr; | |
571 | } | |
572 | ||
573 | static struct nl_mmap_hdr * | |
574 | netlink_current_frame(const struct netlink_ring *ring, | |
575 | enum nl_mmap_status status) | |
576 | { | |
577 | return netlink_lookup_frame(ring, ring->head, status); | |
578 | } | |
579 | ||
580 | static struct nl_mmap_hdr * | |
581 | netlink_previous_frame(const struct netlink_ring *ring, | |
582 | enum nl_mmap_status status) | |
583 | { | |
584 | unsigned int prev; | |
585 | ||
586 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
587 | return netlink_lookup_frame(ring, prev, status); | |
588 | } | |
589 | ||
590 | static void netlink_increment_head(struct netlink_ring *ring) | |
591 | { | |
592 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
593 | } | |
594 | ||
595 | static void netlink_forward_ring(struct netlink_ring *ring) | |
596 | { | |
597 | unsigned int head = ring->head, pos = head; | |
598 | const struct nl_mmap_hdr *hdr; | |
599 | ||
600 | do { | |
601 | hdr = __netlink_lookup_frame(ring, pos); | |
602 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
603 | break; | |
604 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
605 | break; | |
606 | netlink_increment_head(ring); | |
607 | } while (ring->head != head); | |
608 | } | |
609 | ||
cd1df525 PM |
610 | static bool netlink_dump_space(struct netlink_sock *nlk) |
611 | { | |
612 | struct netlink_ring *ring = &nlk->rx_ring; | |
613 | struct nl_mmap_hdr *hdr; | |
614 | unsigned int n; | |
615 | ||
616 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
617 | if (hdr == NULL) | |
618 | return false; | |
619 | ||
620 | n = ring->head + ring->frame_max / 2; | |
621 | if (n > ring->frame_max) | |
622 | n -= ring->frame_max; | |
623 | ||
624 | hdr = __netlink_lookup_frame(ring, n); | |
625 | ||
626 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
627 | } | |
628 | ||
9652e931 PM |
629 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
630 | poll_table *wait) | |
631 | { | |
632 | struct sock *sk = sock->sk; | |
633 | struct netlink_sock *nlk = nlk_sk(sk); | |
634 | unsigned int mask; | |
cd1df525 | 635 | int err; |
9652e931 | 636 | |
cd1df525 PM |
637 | if (nlk->rx_ring.pg_vec != NULL) { |
638 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
639 | * for dumps is performed here. A dump is allowed to continue | |
640 | * if at least half the ring is unused. | |
641 | */ | |
16b304f3 | 642 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
643 | err = netlink_dump(sk); |
644 | if (err < 0) { | |
ac30ef83 | 645 | sk->sk_err = -err; |
cd1df525 PM |
646 | sk->sk_error_report(sk); |
647 | break; | |
648 | } | |
649 | } | |
650 | netlink_rcv_wake(sk); | |
651 | } | |
5fd96123 | 652 | |
9652e931 PM |
653 | mask = datagram_poll(file, sock, wait); |
654 | ||
655 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
656 | if (nlk->rx_ring.pg_vec) { | |
657 | netlink_forward_ring(&nlk->rx_ring); | |
658 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
659 | mask |= POLLIN | POLLRDNORM; | |
660 | } | |
661 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
662 | ||
663 | spin_lock_bh(&sk->sk_write_queue.lock); | |
664 | if (nlk->tx_ring.pg_vec) { | |
665 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
666 | mask |= POLLOUT | POLLWRNORM; | |
667 | } | |
668 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
669 | ||
670 | return mask; | |
671 | } | |
672 | ||
673 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
674 | { | |
675 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
676 | } | |
677 | ||
678 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
679 | struct netlink_ring *ring, | |
680 | struct nl_mmap_hdr *hdr) | |
681 | { | |
682 | unsigned int size; | |
683 | void *data; | |
684 | ||
685 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
686 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
687 | ||
688 | skb->head = data; | |
689 | skb->data = data; | |
690 | skb_reset_tail_pointer(skb); | |
691 | skb->end = skb->tail + size; | |
692 | skb->len = 0; | |
693 | ||
694 | skb->destructor = netlink_skb_destructor; | |
695 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
696 | NETLINK_CB(skb).sk = sk; | |
697 | } | |
5fd96123 PM |
698 | |
699 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
700 | u32 dst_portid, u32 dst_group, | |
701 | struct sock_iocb *siocb) | |
702 | { | |
703 | struct netlink_sock *nlk = nlk_sk(sk); | |
704 | struct netlink_ring *ring; | |
705 | struct nl_mmap_hdr *hdr; | |
706 | struct sk_buff *skb; | |
707 | unsigned int maxlen; | |
5fd96123 PM |
708 | int err = 0, len = 0; |
709 | ||
5fd96123 PM |
710 | mutex_lock(&nlk->pg_vec_lock); |
711 | ||
712 | ring = &nlk->tx_ring; | |
713 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
714 | ||
715 | do { | |
4682a035 DM |
716 | unsigned int nm_len; |
717 | ||
5fd96123 PM |
718 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); |
719 | if (hdr == NULL) { | |
720 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
721 | atomic_read(&nlk->tx_ring.pending)) | |
722 | schedule(); | |
723 | continue; | |
724 | } | |
4682a035 DM |
725 | |
726 | nm_len = ACCESS_ONCE(hdr->nm_len); | |
727 | if (nm_len > maxlen) { | |
5fd96123 PM |
728 | err = -EINVAL; |
729 | goto out; | |
730 | } | |
731 | ||
4682a035 | 732 | netlink_frame_flush_dcache(hdr, nm_len); |
5fd96123 | 733 | |
4682a035 DM |
734 | skb = alloc_skb(nm_len, GFP_KERNEL); |
735 | if (skb == NULL) { | |
736 | err = -ENOBUFS; | |
737 | goto out; | |
5fd96123 | 738 | } |
4682a035 DM |
739 | __skb_put(skb, nm_len); |
740 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, nm_len); | |
741 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
5fd96123 PM |
742 | |
743 | netlink_increment_head(ring); | |
744 | ||
745 | NETLINK_CB(skb).portid = nlk->portid; | |
746 | NETLINK_CB(skb).dst_group = dst_group; | |
747 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
748 | ||
749 | err = security_netlink_send(sk, skb); | |
750 | if (err) { | |
751 | kfree_skb(skb); | |
752 | goto out; | |
753 | } | |
754 | ||
755 | if (unlikely(dst_group)) { | |
756 | atomic_inc(&skb->users); | |
757 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
758 | GFP_KERNEL); | |
759 | } | |
760 | err = netlink_unicast(sk, skb, dst_portid, | |
761 | msg->msg_flags & MSG_DONTWAIT); | |
762 | if (err < 0) | |
763 | goto out; | |
764 | len += err; | |
765 | ||
766 | } while (hdr != NULL || | |
767 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
768 | atomic_read(&nlk->tx_ring.pending))); | |
769 | ||
770 | if (len > 0) | |
771 | err = len; | |
772 | out: | |
773 | mutex_unlock(&nlk->pg_vec_lock); | |
774 | return err; | |
775 | } | |
f9c22888 PM |
776 | |
777 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
778 | { | |
779 | struct nl_mmap_hdr *hdr; | |
780 | ||
781 | hdr = netlink_mmap_hdr(skb); | |
782 | hdr->nm_len = skb->len; | |
783 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
784 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
785 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
786 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
4682a035 | 787 | netlink_frame_flush_dcache(hdr, hdr->nm_len); |
f9c22888 PM |
788 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); |
789 | ||
790 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
791 | kfree_skb(skb); | |
792 | } | |
793 | ||
794 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
795 | { | |
796 | struct netlink_sock *nlk = nlk_sk(sk); | |
797 | struct netlink_ring *ring = &nlk->rx_ring; | |
798 | struct nl_mmap_hdr *hdr; | |
799 | ||
800 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
801 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
802 | if (hdr == NULL) { | |
803 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
804 | kfree_skb(skb); | |
cd1df525 | 805 | netlink_overrun(sk); |
f9c22888 PM |
806 | return; |
807 | } | |
808 | netlink_increment_head(ring); | |
809 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
810 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
811 | ||
812 | hdr->nm_len = skb->len; | |
813 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
814 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
815 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
816 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
817 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
818 | } | |
819 | ||
ccdfcc39 | 820 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 821 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 822 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 823 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 824 | #define netlink_mmap sock_no_mmap |
9652e931 | 825 | #define netlink_poll datagram_poll |
5fd96123 | 826 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
827 | #endif /* CONFIG_NETLINK_MMAP */ |
828 | ||
cf0a018a PM |
829 | static void netlink_skb_destructor(struct sk_buff *skb) |
830 | { | |
9652e931 PM |
831 | #ifdef CONFIG_NETLINK_MMAP |
832 | struct nl_mmap_hdr *hdr; | |
833 | struct netlink_ring *ring; | |
834 | struct sock *sk; | |
835 | ||
836 | /* If a packet from the kernel to userspace was freed because of an | |
837 | * error without being delivered to userspace, the kernel must reset | |
838 | * the status. In the direction userspace to kernel, the status is | |
839 | * always reset here after the packet was processed and freed. | |
840 | */ | |
841 | if (netlink_skb_is_mmaped(skb)) { | |
842 | hdr = netlink_mmap_hdr(skb); | |
843 | sk = NETLINK_CB(skb).sk; | |
844 | ||
5fd96123 PM |
845 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
846 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
847 | ring = &nlk_sk(sk)->tx_ring; | |
848 | } else { | |
849 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
850 | hdr->nm_len = 0; | |
851 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
852 | } | |
853 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 854 | } |
9652e931 PM |
855 | |
856 | WARN_ON(atomic_read(&ring->pending) == 0); | |
857 | atomic_dec(&ring->pending); | |
858 | sock_put(sk); | |
859 | ||
5e71d9d7 | 860 | skb->head = NULL; |
9652e931 PM |
861 | } |
862 | #endif | |
c05cdb1b | 863 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
864 | if (!skb->cloned || |
865 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
866 | vfree(skb->head); | |
867 | ||
c05cdb1b PNA |
868 | skb->head = NULL; |
869 | } | |
9652e931 PM |
870 | if (skb->sk != NULL) |
871 | sock_rfree(skb); | |
cf0a018a PM |
872 | } |
873 | ||
874 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
875 | { | |
876 | WARN_ON(skb->sk != NULL); | |
877 | skb->sk = sk; | |
878 | skb->destructor = netlink_skb_destructor; | |
879 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
880 | sk_mem_charge(sk, skb->truesize); | |
881 | } | |
882 | ||
1da177e4 LT |
883 | static void netlink_sock_destruct(struct sock *sk) |
884 | { | |
3f660d66 HX |
885 | struct netlink_sock *nlk = nlk_sk(sk); |
886 | ||
16b304f3 PS |
887 | if (nlk->cb_running) { |
888 | if (nlk->cb.done) | |
889 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 890 | |
16b304f3 PS |
891 | module_put(nlk->cb.module); |
892 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
893 | } |
894 | ||
1da177e4 | 895 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
896 | #ifdef CONFIG_NETLINK_MMAP |
897 | if (1) { | |
898 | struct nl_mmap_req req; | |
899 | ||
900 | memset(&req, 0, sizeof(req)); | |
901 | if (nlk->rx_ring.pg_vec) | |
902 | netlink_set_ring(sk, &req, true, false); | |
903 | memset(&req, 0, sizeof(req)); | |
904 | if (nlk->tx_ring.pg_vec) | |
905 | netlink_set_ring(sk, &req, true, true); | |
906 | } | |
907 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
908 | |
909 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 910 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
911 | return; |
912 | } | |
547b792c IJ |
913 | |
914 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
915 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
916 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
917 | } |
918 | ||
6ac552fd PM |
919 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
920 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
921 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
922 | * this, _but_ remember, it adds useless work on UP machines. | |
923 | */ | |
924 | ||
d136f1bd | 925 | void netlink_table_grab(void) |
9a429c49 | 926 | __acquires(nl_table_lock) |
1da177e4 | 927 | { |
d136f1bd JB |
928 | might_sleep(); |
929 | ||
6abd219c | 930 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
931 | |
932 | if (atomic_read(&nl_table_users)) { | |
933 | DECLARE_WAITQUEUE(wait, current); | |
934 | ||
935 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 936 | for (;;) { |
1da177e4 LT |
937 | set_current_state(TASK_UNINTERRUPTIBLE); |
938 | if (atomic_read(&nl_table_users) == 0) | |
939 | break; | |
6abd219c | 940 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 941 | schedule(); |
6abd219c | 942 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
943 | } |
944 | ||
945 | __set_current_state(TASK_RUNNING); | |
946 | remove_wait_queue(&nl_table_wait, &wait); | |
947 | } | |
948 | } | |
949 | ||
d136f1bd | 950 | void netlink_table_ungrab(void) |
9a429c49 | 951 | __releases(nl_table_lock) |
1da177e4 | 952 | { |
6abd219c | 953 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
954 | wake_up(&nl_table_wait); |
955 | } | |
956 | ||
6ac552fd | 957 | static inline void |
1da177e4 LT |
958 | netlink_lock_table(void) |
959 | { | |
960 | /* read_lock() synchronizes us to netlink_table_grab */ | |
961 | ||
962 | read_lock(&nl_table_lock); | |
963 | atomic_inc(&nl_table_users); | |
964 | read_unlock(&nl_table_lock); | |
965 | } | |
966 | ||
6ac552fd | 967 | static inline void |
1da177e4 LT |
968 | netlink_unlock_table(void) |
969 | { | |
970 | if (atomic_dec_and_test(&nl_table_users)) | |
971 | wake_up(&nl_table_wait); | |
972 | } | |
973 | ||
e341694e | 974 | struct netlink_compare_arg |
1da177e4 | 975 | { |
e341694e TG |
976 | struct net *net; |
977 | u32 portid; | |
978 | }; | |
1da177e4 | 979 | |
e341694e | 980 | static bool netlink_compare(void *ptr, void *arg) |
1da177e4 | 981 | { |
e341694e TG |
982 | struct netlink_compare_arg *x = arg; |
983 | struct sock *sk = ptr; | |
1da177e4 | 984 | |
e341694e TG |
985 | return nlk_sk(sk)->portid == x->portid && |
986 | net_eq(sock_net(sk), x->net); | |
1da177e4 LT |
987 | } |
988 | ||
e341694e TG |
989 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
990 | struct net *net) | |
1da177e4 | 991 | { |
e341694e TG |
992 | struct netlink_compare_arg arg = { |
993 | .net = net, | |
994 | .portid = portid, | |
995 | }; | |
1da177e4 | 996 | |
8d24c0b4 | 997 | return rhashtable_lookup_compare(&table->hash, &portid, |
e341694e | 998 | &netlink_compare, &arg); |
1da177e4 LT |
999 | } |
1000 | ||
e341694e | 1001 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1002 | { |
e341694e TG |
1003 | struct netlink_table *table = &nl_table[protocol]; |
1004 | struct sock *sk; | |
1da177e4 | 1005 | |
78fd1d0a | 1006 | read_lock(&nl_table_lock); |
e341694e TG |
1007 | rcu_read_lock(); |
1008 | sk = __netlink_lookup(table, portid, net); | |
1009 | if (sk) | |
1010 | sock_hold(sk); | |
1011 | rcu_read_unlock(); | |
78fd1d0a | 1012 | read_unlock(&nl_table_lock); |
1da177e4 | 1013 | |
e341694e | 1014 | return sk; |
1da177e4 LT |
1015 | } |
1016 | ||
90ddc4f0 | 1017 | static const struct proto_ops netlink_ops; |
1da177e4 | 1018 | |
4277a083 PM |
1019 | static void |
1020 | netlink_update_listeners(struct sock *sk) | |
1021 | { | |
1022 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1023 | unsigned long mask; |
1024 | unsigned int i; | |
6d772ac5 ED |
1025 | struct listeners *listeners; |
1026 | ||
1027 | listeners = nl_deref_protected(tbl->listeners); | |
1028 | if (!listeners) | |
1029 | return; | |
4277a083 | 1030 | |
b4ff4f04 | 1031 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1032 | mask = 0; |
b67bfe0d | 1033 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1034 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1035 | mask |= nlk_sk(sk)->groups[i]; | |
1036 | } | |
6d772ac5 | 1037 | listeners->masks[i] = mask; |
4277a083 PM |
1038 | } |
1039 | /* this function is only called with the netlink table "grabbed", which | |
1040 | * makes sure updates are visible before bind or setsockopt return. */ | |
1041 | } | |
1042 | ||
15e47304 | 1043 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1044 | { |
da12c90e | 1045 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1da177e4 | 1046 | int err = -EADDRINUSE; |
1da177e4 | 1047 | |
e341694e TG |
1048 | mutex_lock(&nl_sk_hash_lock); |
1049 | if (__netlink_lookup(table, portid, net)) | |
1da177e4 LT |
1050 | goto err; |
1051 | ||
1052 | err = -EBUSY; | |
15e47304 | 1053 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1054 | goto err; |
1055 | ||
1056 | err = -ENOMEM; | |
97defe1e TG |
1057 | if (BITS_PER_LONG > 32 && |
1058 | unlikely(atomic_read(&table->hash.nelems) >= UINT_MAX)) | |
1da177e4 LT |
1059 | goto err; |
1060 | ||
15e47304 | 1061 | nlk_sk(sk)->portid = portid; |
e341694e | 1062 | sock_hold(sk); |
6eba8224 | 1063 | rhashtable_insert(&table->hash, &nlk_sk(sk)->node); |
1da177e4 | 1064 | err = 0; |
1da177e4 | 1065 | err: |
e341694e | 1066 | mutex_unlock(&nl_sk_hash_lock); |
1da177e4 LT |
1067 | return err; |
1068 | } | |
1069 | ||
1070 | static void netlink_remove(struct sock *sk) | |
1071 | { | |
e341694e TG |
1072 | struct netlink_table *table; |
1073 | ||
1074 | mutex_lock(&nl_sk_hash_lock); | |
1075 | table = &nl_table[sk->sk_protocol]; | |
6eba8224 | 1076 | if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node)) { |
e341694e TG |
1077 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
1078 | __sock_put(sk); | |
1079 | } | |
1080 | mutex_unlock(&nl_sk_hash_lock); | |
1081 | ||
1da177e4 | 1082 | netlink_table_grab(); |
b10dcb3b | 1083 | if (nlk_sk(sk)->subscriptions) { |
1da177e4 | 1084 | __sk_del_bind_node(sk); |
b10dcb3b JB |
1085 | netlink_update_listeners(sk); |
1086 | } | |
1da177e4 LT |
1087 | netlink_table_ungrab(); |
1088 | } | |
1089 | ||
1090 | static struct proto netlink_proto = { | |
1091 | .name = "NETLINK", | |
1092 | .owner = THIS_MODULE, | |
1093 | .obj_size = sizeof(struct netlink_sock), | |
1094 | }; | |
1095 | ||
1b8d7ae4 EB |
1096 | static int __netlink_create(struct net *net, struct socket *sock, |
1097 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1098 | { |
1099 | struct sock *sk; | |
1100 | struct netlink_sock *nlk; | |
ab33a171 PM |
1101 | |
1102 | sock->ops = &netlink_ops; | |
1103 | ||
6257ff21 | 1104 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1105 | if (!sk) |
1106 | return -ENOMEM; | |
1107 | ||
1108 | sock_init_data(sock, sk); | |
1109 | ||
1110 | nlk = nlk_sk(sk); | |
658cb354 | 1111 | if (cb_mutex) { |
ffa4d721 | 1112 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1113 | } else { |
ffa4d721 PM |
1114 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1115 | mutex_init(nlk->cb_mutex); | |
1116 | } | |
ab33a171 | 1117 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1118 | #ifdef CONFIG_NETLINK_MMAP |
1119 | mutex_init(&nlk->pg_vec_lock); | |
1120 | #endif | |
ab33a171 PM |
1121 | |
1122 | sk->sk_destruct = netlink_sock_destruct; | |
1123 | sk->sk_protocol = protocol; | |
1124 | return 0; | |
1125 | } | |
1126 | ||
3f378b68 EP |
1127 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1128 | int kern) | |
ab33a171 PM |
1129 | { |
1130 | struct module *module = NULL; | |
af65bdfc | 1131 | struct mutex *cb_mutex; |
f7fa9b10 | 1132 | struct netlink_sock *nlk; |
023e2cfa JB |
1133 | int (*bind)(struct net *net, int group); |
1134 | void (*unbind)(struct net *net, int group); | |
ab33a171 | 1135 | int err = 0; |
1da177e4 LT |
1136 | |
1137 | sock->state = SS_UNCONNECTED; | |
1138 | ||
1139 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1140 | return -ESOCKTNOSUPPORT; | |
1141 | ||
6ac552fd | 1142 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1143 | return -EPROTONOSUPPORT; |
1144 | ||
77247bbb | 1145 | netlink_lock_table(); |
95a5afca | 1146 | #ifdef CONFIG_MODULES |
ab33a171 | 1147 | if (!nl_table[protocol].registered) { |
77247bbb | 1148 | netlink_unlock_table(); |
4fdb3bb7 | 1149 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1150 | netlink_lock_table(); |
4fdb3bb7 | 1151 | } |
ab33a171 PM |
1152 | #endif |
1153 | if (nl_table[protocol].registered && | |
1154 | try_module_get(nl_table[protocol].module)) | |
1155 | module = nl_table[protocol].module; | |
974c37e9 AD |
1156 | else |
1157 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1158 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1159 | bind = nl_table[protocol].bind; |
4f520900 | 1160 | unbind = nl_table[protocol].unbind; |
77247bbb | 1161 | netlink_unlock_table(); |
4fdb3bb7 | 1162 | |
974c37e9 AD |
1163 | if (err < 0) |
1164 | goto out; | |
1165 | ||
6ac552fd PM |
1166 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1167 | if (err < 0) | |
f7fa9b10 PM |
1168 | goto out_module; |
1169 | ||
6f756a8c | 1170 | local_bh_disable(); |
c1fd3b94 | 1171 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1172 | local_bh_enable(); |
1173 | ||
f7fa9b10 | 1174 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1175 | nlk->module = module; |
03292745 | 1176 | nlk->netlink_bind = bind; |
4f520900 | 1177 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1178 | out: |
1179 | return err; | |
1da177e4 | 1180 | |
ab33a171 PM |
1181 | out_module: |
1182 | module_put(module); | |
1183 | goto out; | |
1da177e4 LT |
1184 | } |
1185 | ||
1186 | static int netlink_release(struct socket *sock) | |
1187 | { | |
1188 | struct sock *sk = sock->sk; | |
1189 | struct netlink_sock *nlk; | |
1190 | ||
1191 | if (!sk) | |
1192 | return 0; | |
1193 | ||
1194 | netlink_remove(sk); | |
ac57b3a9 | 1195 | sock_orphan(sk); |
1da177e4 LT |
1196 | nlk = nlk_sk(sk); |
1197 | ||
3f660d66 HX |
1198 | /* |
1199 | * OK. Socket is unlinked, any packets that arrive now | |
1200 | * will be purged. | |
1201 | */ | |
1da177e4 | 1202 | |
1da177e4 LT |
1203 | sock->sk = NULL; |
1204 | wake_up_interruptible_all(&nlk->wait); | |
1205 | ||
1206 | skb_queue_purge(&sk->sk_write_queue); | |
1207 | ||
15e47304 | 1208 | if (nlk->portid) { |
1da177e4 | 1209 | struct netlink_notify n = { |
3b1e0a65 | 1210 | .net = sock_net(sk), |
1da177e4 | 1211 | .protocol = sk->sk_protocol, |
15e47304 | 1212 | .portid = nlk->portid, |
1da177e4 | 1213 | }; |
e041c683 AS |
1214 | atomic_notifier_call_chain(&netlink_chain, |
1215 | NETLINK_URELEASE, &n); | |
746fac4d | 1216 | } |
4fdb3bb7 | 1217 | |
5e7c001c | 1218 | module_put(nlk->module); |
4fdb3bb7 | 1219 | |
aed81560 | 1220 | if (netlink_is_kernel(sk)) { |
b10dcb3b | 1221 | netlink_table_grab(); |
869e58f8 DL |
1222 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1223 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1224 | struct listeners *old; |
1225 | ||
1226 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1227 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1228 | kfree_rcu(old, rcu); | |
869e58f8 | 1229 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1230 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1231 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1232 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1233 | nl_table[sk->sk_protocol].registered = 0; |
1234 | } | |
b10dcb3b | 1235 | netlink_table_ungrab(); |
658cb354 | 1236 | } |
77247bbb | 1237 | |
7d68536b JB |
1238 | if (nlk->netlink_unbind) { |
1239 | int i; | |
1240 | ||
1241 | for (i = 0; i < nlk->ngroups; i++) | |
1242 | if (test_bit(i, nlk->groups)) | |
023e2cfa | 1243 | nlk->netlink_unbind(sock_net(sk), i + 1); |
7d68536b | 1244 | } |
f7fa9b10 PM |
1245 | kfree(nlk->groups); |
1246 | nlk->groups = NULL; | |
1247 | ||
3755810c | 1248 | local_bh_disable(); |
c1fd3b94 | 1249 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1250 | local_bh_enable(); |
1da177e4 LT |
1251 | sock_put(sk); |
1252 | return 0; | |
1253 | } | |
1254 | ||
1255 | static int netlink_autobind(struct socket *sock) | |
1256 | { | |
1257 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1258 | struct net *net = sock_net(sk); |
da12c90e | 1259 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1260 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1261 | int err; |
1262 | static s32 rover = -4097; | |
1263 | ||
1264 | retry: | |
1265 | cond_resched(); | |
78fd1d0a | 1266 | netlink_table_grab(); |
e341694e TG |
1267 | rcu_read_lock(); |
1268 | if (__netlink_lookup(table, portid, net)) { | |
1269 | /* Bind collision, search negative portid values. */ | |
1270 | portid = rover--; | |
1271 | if (rover > -4097) | |
1272 | rover = -4097; | |
1273 | rcu_read_unlock(); | |
78fd1d0a | 1274 | netlink_table_ungrab(); |
e341694e | 1275 | goto retry; |
1da177e4 | 1276 | } |
e341694e | 1277 | rcu_read_unlock(); |
78fd1d0a | 1278 | netlink_table_ungrab(); |
1da177e4 | 1279 | |
15e47304 | 1280 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1281 | if (err == -EADDRINUSE) |
1282 | goto retry; | |
d470e3b4 DM |
1283 | |
1284 | /* If 2 threads race to autobind, that is fine. */ | |
1285 | if (err == -EBUSY) | |
1286 | err = 0; | |
1287 | ||
1288 | return err; | |
1da177e4 LT |
1289 | } |
1290 | ||
aa4cf945 EB |
1291 | /** |
1292 | * __netlink_ns_capable - General netlink message capability test | |
1293 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1294 | * @user_ns: The user namespace of the capability to use | |
1295 | * @cap: The capability to use | |
1296 | * | |
1297 | * Test to see if the opener of the socket we received the message | |
1298 | * from had when the netlink socket was created and the sender of the | |
1299 | * message has has the capability @cap in the user namespace @user_ns. | |
1300 | */ | |
1301 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1302 | struct user_namespace *user_ns, int cap) | |
1303 | { | |
2d7a85f4 EB |
1304 | return ((nsp->flags & NETLINK_SKB_DST) || |
1305 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1306 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1307 | } |
1308 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1309 | ||
1310 | /** | |
1311 | * netlink_ns_capable - General netlink message capability test | |
1312 | * @skb: socket buffer holding a netlink command from userspace | |
1313 | * @user_ns: The user namespace of the capability to use | |
1314 | * @cap: The capability to use | |
1315 | * | |
1316 | * Test to see if the opener of the socket we received the message | |
1317 | * from had when the netlink socket was created and the sender of the | |
1318 | * message has has the capability @cap in the user namespace @user_ns. | |
1319 | */ | |
1320 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1321 | struct user_namespace *user_ns, int cap) | |
1322 | { | |
1323 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1324 | } | |
1325 | EXPORT_SYMBOL(netlink_ns_capable); | |
1326 | ||
1327 | /** | |
1328 | * netlink_capable - Netlink global message capability test | |
1329 | * @skb: socket buffer holding a netlink command from userspace | |
1330 | * @cap: The capability to use | |
1331 | * | |
1332 | * Test to see if the opener of the socket we received the message | |
1333 | * from had when the netlink socket was created and the sender of the | |
1334 | * message has has the capability @cap in all user namespaces. | |
1335 | */ | |
1336 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1337 | { | |
1338 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1339 | } | |
1340 | EXPORT_SYMBOL(netlink_capable); | |
1341 | ||
1342 | /** | |
1343 | * netlink_net_capable - Netlink network namespace message capability test | |
1344 | * @skb: socket buffer holding a netlink command from userspace | |
1345 | * @cap: The capability to use | |
1346 | * | |
1347 | * Test to see if the opener of the socket we received the message | |
1348 | * from had when the netlink socket was created and the sender of the | |
1349 | * message has has the capability @cap over the network namespace of | |
1350 | * the socket we received the message from. | |
1351 | */ | |
1352 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1353 | { | |
1354 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1355 | } | |
1356 | EXPORT_SYMBOL(netlink_net_capable); | |
1357 | ||
5187cd05 | 1358 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1359 | { |
9785e10a | 1360 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1361 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1362 | } |
1da177e4 | 1363 | |
f7fa9b10 PM |
1364 | static void |
1365 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1366 | { | |
1367 | struct netlink_sock *nlk = nlk_sk(sk); | |
1368 | ||
1369 | if (nlk->subscriptions && !subscriptions) | |
1370 | __sk_del_bind_node(sk); | |
1371 | else if (!nlk->subscriptions && subscriptions) | |
1372 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1373 | nlk->subscriptions = subscriptions; | |
1374 | } | |
1375 | ||
b4ff4f04 | 1376 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1377 | { |
1378 | struct netlink_sock *nlk = nlk_sk(sk); | |
1379 | unsigned int groups; | |
b4ff4f04 | 1380 | unsigned long *new_groups; |
513c2500 PM |
1381 | int err = 0; |
1382 | ||
b4ff4f04 JB |
1383 | netlink_table_grab(); |
1384 | ||
513c2500 | 1385 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1386 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1387 | err = -ENOENT; |
b4ff4f04 JB |
1388 | goto out_unlock; |
1389 | } | |
513c2500 | 1390 | |
b4ff4f04 JB |
1391 | if (nlk->ngroups >= groups) |
1392 | goto out_unlock; | |
513c2500 | 1393 | |
b4ff4f04 JB |
1394 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1395 | if (new_groups == NULL) { | |
1396 | err = -ENOMEM; | |
1397 | goto out_unlock; | |
1398 | } | |
6ac552fd | 1399 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1400 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1401 | ||
1402 | nlk->groups = new_groups; | |
513c2500 | 1403 | nlk->ngroups = groups; |
b4ff4f04 JB |
1404 | out_unlock: |
1405 | netlink_table_ungrab(); | |
1406 | return err; | |
513c2500 PM |
1407 | } |
1408 | ||
02c81ab9 | 1409 | static void netlink_undo_bind(int group, long unsigned int groups, |
023e2cfa | 1410 | struct sock *sk) |
4f520900 | 1411 | { |
023e2cfa | 1412 | struct netlink_sock *nlk = nlk_sk(sk); |
4f520900 RGB |
1413 | int undo; |
1414 | ||
1415 | if (!nlk->netlink_unbind) | |
1416 | return; | |
1417 | ||
1418 | for (undo = 0; undo < group; undo++) | |
6251edd9 | 1419 | if (test_bit(undo, &groups)) |
023e2cfa | 1420 | nlk->netlink_unbind(sock_net(sk), undo); |
4f520900 RGB |
1421 | } |
1422 | ||
6ac552fd PM |
1423 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1424 | int addr_len) | |
1da177e4 LT |
1425 | { |
1426 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1427 | struct net *net = sock_net(sk); |
1da177e4 LT |
1428 | struct netlink_sock *nlk = nlk_sk(sk); |
1429 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1430 | int err; | |
4f520900 | 1431 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1432 | |
4e4b5376 HFS |
1433 | if (addr_len < sizeof(struct sockaddr_nl)) |
1434 | return -EINVAL; | |
1435 | ||
1da177e4 LT |
1436 | if (nladdr->nl_family != AF_NETLINK) |
1437 | return -EINVAL; | |
1438 | ||
1439 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1440 | if (groups) { |
5187cd05 | 1441 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1442 | return -EPERM; |
b4ff4f04 JB |
1443 | err = netlink_realloc_groups(sk); |
1444 | if (err) | |
1445 | return err; | |
513c2500 | 1446 | } |
1da177e4 | 1447 | |
4f520900 | 1448 | if (nlk->portid) |
15e47304 | 1449 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1450 | return -EINVAL; |
4f520900 RGB |
1451 | |
1452 | if (nlk->netlink_bind && groups) { | |
1453 | int group; | |
1454 | ||
1455 | for (group = 0; group < nlk->ngroups; group++) { | |
1456 | if (!test_bit(group, &groups)) | |
1457 | continue; | |
023e2cfa | 1458 | err = nlk->netlink_bind(net, group); |
4f520900 RGB |
1459 | if (!err) |
1460 | continue; | |
023e2cfa | 1461 | netlink_undo_bind(group, groups, sk); |
4f520900 RGB |
1462 | return err; |
1463 | } | |
1464 | } | |
1465 | ||
1466 | if (!nlk->portid) { | |
1da177e4 | 1467 | err = nladdr->nl_pid ? |
b4b51029 | 1468 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1469 | netlink_autobind(sock); |
4f520900 | 1470 | if (err) { |
023e2cfa | 1471 | netlink_undo_bind(nlk->ngroups, groups, sk); |
1da177e4 | 1472 | return err; |
4f520900 | 1473 | } |
1da177e4 LT |
1474 | } |
1475 | ||
4f520900 | 1476 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1477 | return 0; |
1478 | ||
1479 | netlink_table_grab(); | |
f7fa9b10 | 1480 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1481 | hweight32(groups) - |
746fac4d | 1482 | hweight32(nlk->groups[0])); |
4f520900 | 1483 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1484 | netlink_update_listeners(sk); |
1da177e4 LT |
1485 | netlink_table_ungrab(); |
1486 | ||
1487 | return 0; | |
1488 | } | |
1489 | ||
1490 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1491 | int alen, int flags) | |
1492 | { | |
1493 | int err = 0; | |
1494 | struct sock *sk = sock->sk; | |
1495 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1496 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1497 | |
6503d961 CG |
1498 | if (alen < sizeof(addr->sa_family)) |
1499 | return -EINVAL; | |
1500 | ||
1da177e4 LT |
1501 | if (addr->sa_family == AF_UNSPEC) { |
1502 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1503 | nlk->dst_portid = 0; |
d629b836 | 1504 | nlk->dst_group = 0; |
1da177e4 LT |
1505 | return 0; |
1506 | } | |
1507 | if (addr->sa_family != AF_NETLINK) | |
1508 | return -EINVAL; | |
1509 | ||
46833a86 | 1510 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1511 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1512 | return -EPERM; |
1513 | ||
15e47304 | 1514 | if (!nlk->portid) |
1da177e4 LT |
1515 | err = netlink_autobind(sock); |
1516 | ||
1517 | if (err == 0) { | |
1518 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1519 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1520 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1521 | } |
1522 | ||
1523 | return err; | |
1524 | } | |
1525 | ||
6ac552fd PM |
1526 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1527 | int *addr_len, int peer) | |
1da177e4 LT |
1528 | { |
1529 | struct sock *sk = sock->sk; | |
1530 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1531 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1532 | |
1da177e4 LT |
1533 | nladdr->nl_family = AF_NETLINK; |
1534 | nladdr->nl_pad = 0; | |
1535 | *addr_len = sizeof(*nladdr); | |
1536 | ||
1537 | if (peer) { | |
15e47304 | 1538 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1539 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1540 | } else { |
15e47304 | 1541 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1542 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1543 | } |
1544 | return 0; | |
1545 | } | |
1546 | ||
15e47304 | 1547 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1548 | { |
1da177e4 LT |
1549 | struct sock *sock; |
1550 | struct netlink_sock *nlk; | |
1551 | ||
15e47304 | 1552 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1553 | if (!sock) |
1554 | return ERR_PTR(-ECONNREFUSED); | |
1555 | ||
1556 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1557 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1558 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1559 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1560 | sock_put(sock); |
1561 | return ERR_PTR(-ECONNREFUSED); | |
1562 | } | |
1563 | return sock; | |
1564 | } | |
1565 | ||
1566 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1567 | { | |
496ad9aa | 1568 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1569 | struct sock *sock; |
1570 | ||
1571 | if (!S_ISSOCK(inode->i_mode)) | |
1572 | return ERR_PTR(-ENOTSOCK); | |
1573 | ||
1574 | sock = SOCKET_I(inode)->sk; | |
1575 | if (sock->sk_family != AF_NETLINK) | |
1576 | return ERR_PTR(-EINVAL); | |
1577 | ||
1578 | sock_hold(sock); | |
1579 | return sock; | |
1580 | } | |
1581 | ||
3a36515f PN |
1582 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1583 | int broadcast) | |
c05cdb1b PNA |
1584 | { |
1585 | struct sk_buff *skb; | |
1586 | void *data; | |
1587 | ||
3a36515f | 1588 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1589 | return alloc_skb(size, GFP_KERNEL); |
1590 | ||
3a36515f PN |
1591 | size = SKB_DATA_ALIGN(size) + |
1592 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1593 | |
1594 | data = vmalloc(size); | |
1595 | if (data == NULL) | |
3a36515f | 1596 | return NULL; |
c05cdb1b | 1597 | |
3a36515f PN |
1598 | skb = build_skb(data, size); |
1599 | if (skb == NULL) | |
1600 | vfree(data); | |
1601 | else { | |
1602 | skb->head_frag = 0; | |
1603 | skb->destructor = netlink_skb_destructor; | |
1604 | } | |
c05cdb1b PNA |
1605 | |
1606 | return skb; | |
c05cdb1b PNA |
1607 | } |
1608 | ||
1da177e4 LT |
1609 | /* |
1610 | * Attach a skb to a netlink socket. | |
1611 | * The caller must hold a reference to the destination socket. On error, the | |
1612 | * reference is dropped. The skb is not send to the destination, just all | |
1613 | * all error checks are performed and memory in the queue is reserved. | |
1614 | * Return values: | |
1615 | * < 0: error. skb freed, reference to sock dropped. | |
1616 | * 0: continue | |
1617 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1618 | */ | |
9457afee | 1619 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1620 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1621 | { |
1622 | struct netlink_sock *nlk; | |
1623 | ||
1624 | nlk = nlk_sk(sk); | |
1625 | ||
5fd96123 PM |
1626 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1627 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1628 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1629 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1630 | if (!*timeo) { |
aed81560 | 1631 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1632 | netlink_overrun(sk); |
1633 | sock_put(sk); | |
1634 | kfree_skb(skb); | |
1635 | return -EAGAIN; | |
1636 | } | |
1637 | ||
1638 | __set_current_state(TASK_INTERRUPTIBLE); | |
1639 | add_wait_queue(&nlk->wait, &wait); | |
1640 | ||
1641 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1642 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1643 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1644 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1645 | |
1646 | __set_current_state(TASK_RUNNING); | |
1647 | remove_wait_queue(&nlk->wait, &wait); | |
1648 | sock_put(sk); | |
1649 | ||
1650 | if (signal_pending(current)) { | |
1651 | kfree_skb(skb); | |
c3d8d1e3 | 1652 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1653 | } |
1654 | return 1; | |
1655 | } | |
cf0a018a | 1656 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1657 | return 0; |
1658 | } | |
1659 | ||
4a7e7c2a | 1660 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1661 | { |
1da177e4 LT |
1662 | int len = skb->len; |
1663 | ||
bcbde0d4 DB |
1664 | netlink_deliver_tap(skb); |
1665 | ||
f9c22888 PM |
1666 | #ifdef CONFIG_NETLINK_MMAP |
1667 | if (netlink_skb_is_mmaped(skb)) | |
1668 | netlink_queue_mmaped_skb(sk, skb); | |
1669 | else if (netlink_rx_is_mmaped(sk)) | |
1670 | netlink_ring_set_copied(sk, skb); | |
1671 | else | |
1672 | #endif /* CONFIG_NETLINK_MMAP */ | |
1673 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1674 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1675 | return len; |
1676 | } | |
1677 | ||
1678 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1679 | { | |
1680 | int len = __netlink_sendskb(sk, skb); | |
1681 | ||
1da177e4 LT |
1682 | sock_put(sk); |
1683 | return len; | |
1684 | } | |
1685 | ||
1686 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1687 | { | |
1688 | kfree_skb(skb); | |
1689 | sock_put(sk); | |
1690 | } | |
1691 | ||
b57ef81f | 1692 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1693 | { |
1694 | int delta; | |
1695 | ||
1298ca46 | 1696 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1697 | if (netlink_skb_is_mmaped(skb)) |
1698 | return skb; | |
1da177e4 | 1699 | |
4305b541 | 1700 | delta = skb->end - skb->tail; |
c05cdb1b | 1701 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1702 | return skb; |
1703 | ||
1704 | if (skb_shared(skb)) { | |
1705 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1706 | if (!nskb) | |
1707 | return skb; | |
8460c00f | 1708 | consume_skb(skb); |
1da177e4 LT |
1709 | skb = nskb; |
1710 | } | |
1711 | ||
1712 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1713 | skb->truesize -= delta; | |
1714 | ||
1715 | return skb; | |
1716 | } | |
1717 | ||
3fbc2905 EB |
1718 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1719 | struct sock *ssk) | |
cd40b7d3 DL |
1720 | { |
1721 | int ret; | |
1722 | struct netlink_sock *nlk = nlk_sk(sk); | |
1723 | ||
1724 | ret = -ECONNREFUSED; | |
1725 | if (nlk->netlink_rcv != NULL) { | |
1726 | ret = skb->len; | |
cf0a018a | 1727 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1728 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1729 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1730 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1731 | consume_skb(skb); |
1732 | } else { | |
1733 | kfree_skb(skb); | |
cd40b7d3 | 1734 | } |
cd40b7d3 DL |
1735 | sock_put(sk); |
1736 | return ret; | |
1737 | } | |
1738 | ||
1739 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1740 | u32 portid, int nonblock) |
1da177e4 LT |
1741 | { |
1742 | struct sock *sk; | |
1743 | int err; | |
1744 | long timeo; | |
1745 | ||
1746 | skb = netlink_trim(skb, gfp_any()); | |
1747 | ||
1748 | timeo = sock_sndtimeo(ssk, nonblock); | |
1749 | retry: | |
15e47304 | 1750 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1751 | if (IS_ERR(sk)) { |
1752 | kfree_skb(skb); | |
1753 | return PTR_ERR(sk); | |
1754 | } | |
cd40b7d3 | 1755 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1756 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1757 | |
b1153f29 | 1758 | if (sk_filter(sk, skb)) { |
84874607 | 1759 | err = skb->len; |
b1153f29 SH |
1760 | kfree_skb(skb); |
1761 | sock_put(sk); | |
1762 | return err; | |
1763 | } | |
1764 | ||
9457afee | 1765 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1766 | if (err == 1) |
1767 | goto retry; | |
1768 | if (err) | |
1769 | return err; | |
1770 | ||
7ee015e0 | 1771 | return netlink_sendskb(sk, skb); |
1da177e4 | 1772 | } |
6ac552fd | 1773 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1774 | |
f9c22888 PM |
1775 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1776 | u32 dst_portid, gfp_t gfp_mask) | |
1777 | { | |
1778 | #ifdef CONFIG_NETLINK_MMAP | |
1779 | struct sock *sk = NULL; | |
1780 | struct sk_buff *skb; | |
1781 | struct netlink_ring *ring; | |
1782 | struct nl_mmap_hdr *hdr; | |
1783 | unsigned int maxlen; | |
1784 | ||
1785 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1786 | if (IS_ERR(sk)) | |
1787 | goto out; | |
1788 | ||
1789 | ring = &nlk_sk(sk)->rx_ring; | |
1790 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1791 | if (ring->pg_vec == NULL) | |
1792 | goto out_put; | |
1793 | ||
aae9f0e2 TG |
1794 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1795 | goto out_put; | |
1796 | ||
f9c22888 PM |
1797 | skb = alloc_skb_head(gfp_mask); |
1798 | if (skb == NULL) | |
1799 | goto err1; | |
1800 | ||
1801 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1802 | /* check again under lock */ | |
1803 | if (ring->pg_vec == NULL) | |
1804 | goto out_free; | |
1805 | ||
aae9f0e2 | 1806 | /* check again under lock */ |
f9c22888 PM |
1807 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1808 | if (maxlen < size) | |
1809 | goto out_free; | |
1810 | ||
1811 | netlink_forward_ring(ring); | |
1812 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1813 | if (hdr == NULL) | |
1814 | goto err2; | |
1815 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1816 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1817 | atomic_inc(&ring->pending); | |
1818 | netlink_increment_head(ring); | |
1819 | ||
1820 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1821 | return skb; | |
1822 | ||
1823 | err2: | |
1824 | kfree_skb(skb); | |
1825 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1826 | netlink_overrun(sk); |
f9c22888 PM |
1827 | err1: |
1828 | sock_put(sk); | |
1829 | return NULL; | |
1830 | ||
1831 | out_free: | |
1832 | kfree_skb(skb); | |
1833 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1834 | out_put: | |
1835 | sock_put(sk); | |
1836 | out: | |
1837 | #endif | |
1838 | return alloc_skb(size, gfp_mask); | |
1839 | } | |
1840 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1841 | ||
4277a083 PM |
1842 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1843 | { | |
1844 | int res = 0; | |
5c398dc8 | 1845 | struct listeners *listeners; |
4277a083 | 1846 | |
aed81560 | 1847 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1848 | |
1849 | rcu_read_lock(); | |
1850 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1851 | ||
6d772ac5 | 1852 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1853 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1854 | |
1855 | rcu_read_unlock(); | |
1856 | ||
4277a083 PM |
1857 | return res; |
1858 | } | |
1859 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1860 | ||
b57ef81f | 1861 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1862 | { |
1863 | struct netlink_sock *nlk = nlk_sk(sk); | |
1864 | ||
1865 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1866 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1867 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1868 | __netlink_sendskb(sk, skb); |
2c645800 | 1869 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1870 | } |
1871 | return -1; | |
1872 | } | |
1873 | ||
1874 | struct netlink_broadcast_data { | |
1875 | struct sock *exclude_sk; | |
b4b51029 | 1876 | struct net *net; |
15e47304 | 1877 | u32 portid; |
1da177e4 LT |
1878 | u32 group; |
1879 | int failure; | |
ff491a73 | 1880 | int delivery_failure; |
1da177e4 LT |
1881 | int congested; |
1882 | int delivered; | |
7d877f3b | 1883 | gfp_t allocation; |
1da177e4 | 1884 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1885 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1886 | void *tx_data; | |
1da177e4 LT |
1887 | }; |
1888 | ||
46c9521f RR |
1889 | static void do_one_broadcast(struct sock *sk, |
1890 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1891 | { |
1892 | struct netlink_sock *nlk = nlk_sk(sk); | |
1893 | int val; | |
1894 | ||
1895 | if (p->exclude_sk == sk) | |
46c9521f | 1896 | return; |
1da177e4 | 1897 | |
15e47304 | 1898 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1899 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1900 | return; |
1da177e4 | 1901 | |
878628fb | 1902 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1903 | return; |
b4b51029 | 1904 | |
1da177e4 LT |
1905 | if (p->failure) { |
1906 | netlink_overrun(sk); | |
46c9521f | 1907 | return; |
1da177e4 LT |
1908 | } |
1909 | ||
1910 | sock_hold(sk); | |
1911 | if (p->skb2 == NULL) { | |
68acc024 | 1912 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1913 | p->skb2 = skb_clone(p->skb, p->allocation); |
1914 | } else { | |
68acc024 TC |
1915 | p->skb2 = skb_get(p->skb); |
1916 | /* | |
1917 | * skb ownership may have been set when | |
1918 | * delivered to a previous socket. | |
1919 | */ | |
1920 | skb_orphan(p->skb2); | |
1da177e4 LT |
1921 | } |
1922 | } | |
1923 | if (p->skb2 == NULL) { | |
1924 | netlink_overrun(sk); | |
1925 | /* Clone failed. Notify ALL listeners. */ | |
1926 | p->failure = 1; | |
be0c22a4 PNA |
1927 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1928 | p->delivery_failure = 1; | |
910a7e90 EB |
1929 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1930 | kfree_skb(p->skb2); | |
1931 | p->skb2 = NULL; | |
b1153f29 SH |
1932 | } else if (sk_filter(sk, p->skb2)) { |
1933 | kfree_skb(p->skb2); | |
1934 | p->skb2 = NULL; | |
1da177e4 LT |
1935 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1936 | netlink_overrun(sk); | |
be0c22a4 PNA |
1937 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1938 | p->delivery_failure = 1; | |
1da177e4 LT |
1939 | } else { |
1940 | p->congested |= val; | |
1941 | p->delivered = 1; | |
1942 | p->skb2 = NULL; | |
1943 | } | |
1944 | sock_put(sk); | |
1da177e4 LT |
1945 | } |
1946 | ||
15e47304 | 1947 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1948 | u32 group, gfp_t allocation, |
1949 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1950 | void *filter_data) | |
1da177e4 | 1951 | { |
3b1e0a65 | 1952 | struct net *net = sock_net(ssk); |
1da177e4 | 1953 | struct netlink_broadcast_data info; |
1da177e4 LT |
1954 | struct sock *sk; |
1955 | ||
1956 | skb = netlink_trim(skb, allocation); | |
1957 | ||
1958 | info.exclude_sk = ssk; | |
b4b51029 | 1959 | info.net = net; |
15e47304 | 1960 | info.portid = portid; |
1da177e4 LT |
1961 | info.group = group; |
1962 | info.failure = 0; | |
ff491a73 | 1963 | info.delivery_failure = 0; |
1da177e4 LT |
1964 | info.congested = 0; |
1965 | info.delivered = 0; | |
1966 | info.allocation = allocation; | |
1967 | info.skb = skb; | |
1968 | info.skb2 = NULL; | |
910a7e90 EB |
1969 | info.tx_filter = filter; |
1970 | info.tx_data = filter_data; | |
1da177e4 LT |
1971 | |
1972 | /* While we sleep in clone, do not allow to change socket list */ | |
1973 | ||
1974 | netlink_lock_table(); | |
1975 | ||
b67bfe0d | 1976 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1977 | do_one_broadcast(sk, &info); |
1978 | ||
70d4bf6d | 1979 | consume_skb(skb); |
aa1c6a6f | 1980 | |
1da177e4 LT |
1981 | netlink_unlock_table(); |
1982 | ||
70d4bf6d NH |
1983 | if (info.delivery_failure) { |
1984 | kfree_skb(info.skb2); | |
ff491a73 | 1985 | return -ENOBUFS; |
658cb354 ED |
1986 | } |
1987 | consume_skb(info.skb2); | |
ff491a73 | 1988 | |
1da177e4 LT |
1989 | if (info.delivered) { |
1990 | if (info.congested && (allocation & __GFP_WAIT)) | |
1991 | yield(); | |
1992 | return 0; | |
1993 | } | |
1da177e4 LT |
1994 | return -ESRCH; |
1995 | } | |
910a7e90 EB |
1996 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
1997 | ||
15e47304 | 1998 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1999 | u32 group, gfp_t allocation) |
2000 | { | |
15e47304 | 2001 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2002 | NULL, NULL); |
2003 | } | |
6ac552fd | 2004 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2005 | |
2006 | struct netlink_set_err_data { | |
2007 | struct sock *exclude_sk; | |
15e47304 | 2008 | u32 portid; |
1da177e4 LT |
2009 | u32 group; |
2010 | int code; | |
2011 | }; | |
2012 | ||
b57ef81f | 2013 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2014 | { |
2015 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2016 | int ret = 0; |
1da177e4 LT |
2017 | |
2018 | if (sk == p->exclude_sk) | |
2019 | goto out; | |
2020 | ||
09ad9bc7 | 2021 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2022 | goto out; |
2023 | ||
15e47304 | 2024 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2025 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2026 | goto out; |
2027 | ||
1a50307b PNA |
2028 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2029 | ret = 1; | |
2030 | goto out; | |
2031 | } | |
2032 | ||
1da177e4 LT |
2033 | sk->sk_err = p->code; |
2034 | sk->sk_error_report(sk); | |
2035 | out: | |
1a50307b | 2036 | return ret; |
1da177e4 LT |
2037 | } |
2038 | ||
4843b93c PNA |
2039 | /** |
2040 | * netlink_set_err - report error to broadcast listeners | |
2041 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2042 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2043 | * @group: the broadcast group that will notice the error |
4843b93c | 2044 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2045 | * |
2046 | * This function returns the number of broadcast listeners that have set the | |
2047 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2048 | */ |
15e47304 | 2049 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2050 | { |
2051 | struct netlink_set_err_data info; | |
1da177e4 | 2052 | struct sock *sk; |
1a50307b | 2053 | int ret = 0; |
1da177e4 LT |
2054 | |
2055 | info.exclude_sk = ssk; | |
15e47304 | 2056 | info.portid = portid; |
1da177e4 | 2057 | info.group = group; |
4843b93c PNA |
2058 | /* sk->sk_err wants a positive error value */ |
2059 | info.code = -code; | |
1da177e4 LT |
2060 | |
2061 | read_lock(&nl_table_lock); | |
2062 | ||
b67bfe0d | 2063 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2064 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2065 | |
2066 | read_unlock(&nl_table_lock); | |
1a50307b | 2067 | return ret; |
1da177e4 | 2068 | } |
dd5b6ce6 | 2069 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2070 | |
84659eb5 JB |
2071 | /* must be called with netlink table grabbed */ |
2072 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2073 | unsigned int group, | |
2074 | int is_new) | |
2075 | { | |
2076 | int old, new = !!is_new, subscriptions; | |
2077 | ||
2078 | old = test_bit(group - 1, nlk->groups); | |
2079 | subscriptions = nlk->subscriptions - old + new; | |
2080 | if (new) | |
2081 | __set_bit(group - 1, nlk->groups); | |
2082 | else | |
2083 | __clear_bit(group - 1, nlk->groups); | |
2084 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2085 | netlink_update_listeners(&nlk->sk); | |
2086 | } | |
2087 | ||
9a4595bc | 2088 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2089 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2090 | { |
2091 | struct sock *sk = sock->sk; | |
2092 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2093 | unsigned int val = 0; |
2094 | int err; | |
9a4595bc PM |
2095 | |
2096 | if (level != SOL_NETLINK) | |
2097 | return -ENOPROTOOPT; | |
2098 | ||
ccdfcc39 PM |
2099 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2100 | optlen >= sizeof(int) && | |
eb496534 | 2101 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2102 | return -EFAULT; |
2103 | ||
2104 | switch (optname) { | |
2105 | case NETLINK_PKTINFO: | |
2106 | if (val) | |
2107 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2108 | else | |
2109 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2110 | err = 0; | |
2111 | break; | |
2112 | case NETLINK_ADD_MEMBERSHIP: | |
2113 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2114 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2115 | return -EPERM; |
b4ff4f04 JB |
2116 | err = netlink_realloc_groups(sk); |
2117 | if (err) | |
2118 | return err; | |
9a4595bc PM |
2119 | if (!val || val - 1 >= nlk->ngroups) |
2120 | return -EINVAL; | |
7774d5e0 | 2121 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
023e2cfa | 2122 | err = nlk->netlink_bind(sock_net(sk), val); |
4f520900 RGB |
2123 | if (err) |
2124 | return err; | |
2125 | } | |
9a4595bc | 2126 | netlink_table_grab(); |
84659eb5 JB |
2127 | netlink_update_socket_mc(nlk, val, |
2128 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2129 | netlink_table_ungrab(); |
7774d5e0 | 2130 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
023e2cfa | 2131 | nlk->netlink_unbind(sock_net(sk), val); |
03292745 | 2132 | |
9a4595bc PM |
2133 | err = 0; |
2134 | break; | |
2135 | } | |
be0c22a4 PNA |
2136 | case NETLINK_BROADCAST_ERROR: |
2137 | if (val) | |
2138 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2139 | else | |
2140 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2141 | err = 0; | |
2142 | break; | |
38938bfe PNA |
2143 | case NETLINK_NO_ENOBUFS: |
2144 | if (val) { | |
2145 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2146 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2147 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2148 | } else { |
38938bfe | 2149 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2150 | } |
38938bfe PNA |
2151 | err = 0; |
2152 | break; | |
ccdfcc39 PM |
2153 | #ifdef CONFIG_NETLINK_MMAP |
2154 | case NETLINK_RX_RING: | |
2155 | case NETLINK_TX_RING: { | |
2156 | struct nl_mmap_req req; | |
2157 | ||
2158 | /* Rings might consume more memory than queue limits, require | |
2159 | * CAP_NET_ADMIN. | |
2160 | */ | |
2161 | if (!capable(CAP_NET_ADMIN)) | |
2162 | return -EPERM; | |
2163 | if (optlen < sizeof(req)) | |
2164 | return -EINVAL; | |
2165 | if (copy_from_user(&req, optval, sizeof(req))) | |
2166 | return -EFAULT; | |
2167 | err = netlink_set_ring(sk, &req, false, | |
2168 | optname == NETLINK_TX_RING); | |
2169 | break; | |
2170 | } | |
2171 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2172 | default: |
2173 | err = -ENOPROTOOPT; | |
2174 | } | |
2175 | return err; | |
2176 | } | |
2177 | ||
2178 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2179 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2180 | { |
2181 | struct sock *sk = sock->sk; | |
2182 | struct netlink_sock *nlk = nlk_sk(sk); | |
2183 | int len, val, err; | |
2184 | ||
2185 | if (level != SOL_NETLINK) | |
2186 | return -ENOPROTOOPT; | |
2187 | ||
2188 | if (get_user(len, optlen)) | |
2189 | return -EFAULT; | |
2190 | if (len < 0) | |
2191 | return -EINVAL; | |
2192 | ||
2193 | switch (optname) { | |
2194 | case NETLINK_PKTINFO: | |
2195 | if (len < sizeof(int)) | |
2196 | return -EINVAL; | |
2197 | len = sizeof(int); | |
2198 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2199 | if (put_user(len, optlen) || |
2200 | put_user(val, optval)) | |
2201 | return -EFAULT; | |
9a4595bc PM |
2202 | err = 0; |
2203 | break; | |
be0c22a4 PNA |
2204 | case NETLINK_BROADCAST_ERROR: |
2205 | if (len < sizeof(int)) | |
2206 | return -EINVAL; | |
2207 | len = sizeof(int); | |
2208 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2209 | if (put_user(len, optlen) || | |
2210 | put_user(val, optval)) | |
2211 | return -EFAULT; | |
2212 | err = 0; | |
2213 | break; | |
38938bfe PNA |
2214 | case NETLINK_NO_ENOBUFS: |
2215 | if (len < sizeof(int)) | |
2216 | return -EINVAL; | |
2217 | len = sizeof(int); | |
2218 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2219 | if (put_user(len, optlen) || | |
2220 | put_user(val, optval)) | |
2221 | return -EFAULT; | |
2222 | err = 0; | |
2223 | break; | |
9a4595bc PM |
2224 | default: |
2225 | err = -ENOPROTOOPT; | |
2226 | } | |
2227 | return err; | |
2228 | } | |
2229 | ||
2230 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2231 | { | |
2232 | struct nl_pktinfo info; | |
2233 | ||
2234 | info.group = NETLINK_CB(skb).dst_group; | |
2235 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2236 | } | |
2237 | ||
1da177e4 LT |
2238 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2239 | struct msghdr *msg, size_t len) | |
2240 | { | |
2241 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2242 | struct sock *sk = sock->sk; | |
2243 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2244 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2245 | u32 dst_portid; |
d629b836 | 2246 | u32 dst_group; |
1da177e4 LT |
2247 | struct sk_buff *skb; |
2248 | int err; | |
2249 | struct scm_cookie scm; | |
2d7a85f4 | 2250 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2251 | |
2252 | if (msg->msg_flags&MSG_OOB) | |
2253 | return -EOPNOTSUPP; | |
2254 | ||
16e57262 | 2255 | if (NULL == siocb->scm) |
1da177e4 | 2256 | siocb->scm = &scm; |
16e57262 | 2257 | |
e0e3cea4 | 2258 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2259 | if (err < 0) |
2260 | return err; | |
2261 | ||
2262 | if (msg->msg_namelen) { | |
b47030c7 | 2263 | err = -EINVAL; |
1da177e4 | 2264 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2265 | goto out; |
15e47304 | 2266 | dst_portid = addr->nl_pid; |
d629b836 | 2267 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2268 | err = -EPERM; |
15e47304 | 2269 | if ((dst_group || dst_portid) && |
5187cd05 | 2270 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2271 | goto out; |
2d7a85f4 | 2272 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2273 | } else { |
15e47304 | 2274 | dst_portid = nlk->dst_portid; |
d629b836 | 2275 | dst_group = nlk->dst_group; |
1da177e4 LT |
2276 | } |
2277 | ||
15e47304 | 2278 | if (!nlk->portid) { |
1da177e4 LT |
2279 | err = netlink_autobind(sock); |
2280 | if (err) | |
2281 | goto out; | |
2282 | } | |
2283 | ||
5fd96123 | 2284 | if (netlink_tx_is_mmaped(sk) && |
c0371da6 | 2285 | msg->msg_iter.iov->iov_base == NULL) { |
5fd96123 PM |
2286 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, |
2287 | siocb); | |
2288 | goto out; | |
2289 | } | |
2290 | ||
1da177e4 LT |
2291 | err = -EMSGSIZE; |
2292 | if (len > sk->sk_sndbuf - 32) | |
2293 | goto out; | |
2294 | err = -ENOBUFS; | |
3a36515f | 2295 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2296 | if (skb == NULL) |
1da177e4 LT |
2297 | goto out; |
2298 | ||
15e47304 | 2299 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2300 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2301 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2302 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2303 | |
1da177e4 | 2304 | err = -EFAULT; |
6ce8e9ce | 2305 | if (memcpy_from_msg(skb_put(skb, len), msg, len)) { |
1da177e4 LT |
2306 | kfree_skb(skb); |
2307 | goto out; | |
2308 | } | |
2309 | ||
2310 | err = security_netlink_send(sk, skb); | |
2311 | if (err) { | |
2312 | kfree_skb(skb); | |
2313 | goto out; | |
2314 | } | |
2315 | ||
d629b836 | 2316 | if (dst_group) { |
1da177e4 | 2317 | atomic_inc(&skb->users); |
15e47304 | 2318 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2319 | } |
15e47304 | 2320 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2321 | |
2322 | out: | |
b47030c7 | 2323 | scm_destroy(siocb->scm); |
1da177e4 LT |
2324 | return err; |
2325 | } | |
2326 | ||
2327 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2328 | struct msghdr *msg, size_t len, | |
2329 | int flags) | |
2330 | { | |
2331 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2332 | struct scm_cookie scm; | |
2333 | struct sock *sk = sock->sk; | |
2334 | struct netlink_sock *nlk = nlk_sk(sk); | |
2335 | int noblock = flags&MSG_DONTWAIT; | |
2336 | size_t copied; | |
68d6ac6d | 2337 | struct sk_buff *skb, *data_skb; |
b44d211e | 2338 | int err, ret; |
1da177e4 LT |
2339 | |
2340 | if (flags&MSG_OOB) | |
2341 | return -EOPNOTSUPP; | |
2342 | ||
2343 | copied = 0; | |
2344 | ||
6ac552fd PM |
2345 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2346 | if (skb == NULL) | |
1da177e4 LT |
2347 | goto out; |
2348 | ||
68d6ac6d JB |
2349 | data_skb = skb; |
2350 | ||
1dacc76d JB |
2351 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2352 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2353 | /* |
68d6ac6d JB |
2354 | * If this skb has a frag_list, then here that means that we |
2355 | * will have to use the frag_list skb's data for compat tasks | |
2356 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2357 | * |
68d6ac6d JB |
2358 | * If we need to send the compat skb, assign it to the |
2359 | * 'data_skb' variable so that it will be used below for data | |
2360 | * copying. We keep 'skb' for everything else, including | |
2361 | * freeing both later. | |
1dacc76d | 2362 | */ |
68d6ac6d JB |
2363 | if (flags & MSG_CMSG_COMPAT) |
2364 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2365 | } |
2366 | #endif | |
2367 | ||
9063e21f ED |
2368 | /* Record the max length of recvmsg() calls for future allocations */ |
2369 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2370 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2371 | 16384); | |
2372 | ||
68d6ac6d | 2373 | copied = data_skb->len; |
1da177e4 LT |
2374 | if (len < copied) { |
2375 | msg->msg_flags |= MSG_TRUNC; | |
2376 | copied = len; | |
2377 | } | |
2378 | ||
68d6ac6d | 2379 | skb_reset_transport_header(data_skb); |
51f3d02b | 2380 | err = skb_copy_datagram_msg(data_skb, 0, msg, copied); |
1da177e4 LT |
2381 | |
2382 | if (msg->msg_name) { | |
342dfc30 | 2383 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2384 | addr->nl_family = AF_NETLINK; |
2385 | addr->nl_pad = 0; | |
15e47304 | 2386 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2387 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2388 | msg->msg_namelen = sizeof(*addr); |
2389 | } | |
2390 | ||
cc9a06cd PM |
2391 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2392 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2393 | ||
1da177e4 LT |
2394 | if (NULL == siocb->scm) { |
2395 | memset(&scm, 0, sizeof(scm)); | |
2396 | siocb->scm = &scm; | |
2397 | } | |
2398 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2399 | if (flags & MSG_TRUNC) |
68d6ac6d | 2400 | copied = data_skb->len; |
daa3766e | 2401 | |
1da177e4 LT |
2402 | skb_free_datagram(sk, skb); |
2403 | ||
16b304f3 PS |
2404 | if (nlk->cb_running && |
2405 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2406 | ret = netlink_dump(sk); |
2407 | if (ret) { | |
ac30ef83 | 2408 | sk->sk_err = -ret; |
b44d211e AV |
2409 | sk->sk_error_report(sk); |
2410 | } | |
2411 | } | |
1da177e4 LT |
2412 | |
2413 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2414 | out: |
2415 | netlink_rcv_wake(sk); | |
2416 | return err ? : copied; | |
2417 | } | |
2418 | ||
676d2369 | 2419 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2420 | { |
cd40b7d3 | 2421 | BUG(); |
1da177e4 LT |
2422 | } |
2423 | ||
2424 | /* | |
746fac4d | 2425 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2426 | * complete set of kernel non-blocking support for message |
2427 | * queueing. | |
2428 | */ | |
2429 | ||
2430 | struct sock * | |
9f00d977 PNA |
2431 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2432 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2433 | { |
2434 | struct socket *sock; | |
2435 | struct sock *sk; | |
77247bbb | 2436 | struct netlink_sock *nlk; |
5c398dc8 | 2437 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2438 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2439 | unsigned int groups; | |
1da177e4 | 2440 | |
fab2caf6 | 2441 | BUG_ON(!nl_table); |
1da177e4 | 2442 | |
6ac552fd | 2443 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2444 | return NULL; |
2445 | ||
2446 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2447 | return NULL; | |
2448 | ||
23fe1866 PE |
2449 | /* |
2450 | * We have to just have a reference on the net from sk, but don't | |
2451 | * get_net it. Besides, we cannot get and then put the net here. | |
2452 | * So we create one inside init_net and the move it to net. | |
2453 | */ | |
2454 | ||
2455 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2456 | goto out_sock_release_nosk; | |
2457 | ||
2458 | sk = sock->sk; | |
edf02087 | 2459 | sk_change_net(sk, net); |
4fdb3bb7 | 2460 | |
a31f2d17 | 2461 | if (!cfg || cfg->groups < 32) |
4277a083 | 2462 | groups = 32; |
a31f2d17 PNA |
2463 | else |
2464 | groups = cfg->groups; | |
4277a083 | 2465 | |
5c398dc8 | 2466 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2467 | if (!listeners) |
2468 | goto out_sock_release; | |
2469 | ||
1da177e4 | 2470 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2471 | if (cfg && cfg->input) |
2472 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2473 | |
b4b51029 | 2474 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2475 | goto out_sock_release; |
4fdb3bb7 | 2476 | |
77247bbb PM |
2477 | nlk = nlk_sk(sk); |
2478 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2479 | |
4fdb3bb7 | 2480 | netlink_table_grab(); |
b4b51029 EB |
2481 | if (!nl_table[unit].registered) { |
2482 | nl_table[unit].groups = groups; | |
5c398dc8 | 2483 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2484 | nl_table[unit].cb_mutex = cb_mutex; |
2485 | nl_table[unit].module = module; | |
9785e10a PNA |
2486 | if (cfg) { |
2487 | nl_table[unit].bind = cfg->bind; | |
6251edd9 | 2488 | nl_table[unit].unbind = cfg->unbind; |
9785e10a | 2489 | nl_table[unit].flags = cfg->flags; |
da12c90e G |
2490 | if (cfg->compare) |
2491 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2492 | } |
b4b51029 | 2493 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2494 | } else { |
2495 | kfree(listeners); | |
869e58f8 | 2496 | nl_table[unit].registered++; |
b4b51029 | 2497 | } |
4fdb3bb7 | 2498 | netlink_table_ungrab(); |
77247bbb PM |
2499 | return sk; |
2500 | ||
4fdb3bb7 | 2501 | out_sock_release: |
4277a083 | 2502 | kfree(listeners); |
9dfbec1f | 2503 | netlink_kernel_release(sk); |
23fe1866 PE |
2504 | return NULL; |
2505 | ||
2506 | out_sock_release_nosk: | |
4fdb3bb7 | 2507 | sock_release(sock); |
77247bbb | 2508 | return NULL; |
1da177e4 | 2509 | } |
9f00d977 | 2510 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2511 | |
2512 | void | |
2513 | netlink_kernel_release(struct sock *sk) | |
2514 | { | |
edf02087 | 2515 | sk_release_kernel(sk); |
b7c6ba6e DL |
2516 | } |
2517 | EXPORT_SYMBOL(netlink_kernel_release); | |
2518 | ||
d136f1bd | 2519 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2520 | { |
5c398dc8 | 2521 | struct listeners *new, *old; |
b4ff4f04 | 2522 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2523 | |
2524 | if (groups < 32) | |
2525 | groups = 32; | |
2526 | ||
b4ff4f04 | 2527 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2528 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2529 | if (!new) | |
d136f1bd | 2530 | return -ENOMEM; |
6d772ac5 | 2531 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2532 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2533 | rcu_assign_pointer(tbl->listeners, new); | |
2534 | ||
37b6b935 | 2535 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2536 | } |
2537 | tbl->groups = groups; | |
2538 | ||
d136f1bd JB |
2539 | return 0; |
2540 | } | |
2541 | ||
2542 | /** | |
2543 | * netlink_change_ngroups - change number of multicast groups | |
2544 | * | |
2545 | * This changes the number of multicast groups that are available | |
2546 | * on a certain netlink family. Note that it is not possible to | |
2547 | * change the number of groups to below 32. Also note that it does | |
2548 | * not implicitly call netlink_clear_multicast_users() when the | |
2549 | * number of groups is reduced. | |
2550 | * | |
2551 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2552 | * @groups: The new number of groups. | |
2553 | */ | |
2554 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2555 | { | |
2556 | int err; | |
2557 | ||
2558 | netlink_table_grab(); | |
2559 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2560 | netlink_table_ungrab(); |
d136f1bd | 2561 | |
b4ff4f04 JB |
2562 | return err; |
2563 | } | |
b4ff4f04 | 2564 | |
b8273570 JB |
2565 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2566 | { | |
2567 | struct sock *sk; | |
b8273570 JB |
2568 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2569 | ||
b67bfe0d | 2570 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2571 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2572 | } | |
2573 | ||
a46621a3 | 2574 | struct nlmsghdr * |
15e47304 | 2575 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2576 | { |
2577 | struct nlmsghdr *nlh; | |
573ce260 | 2578 | int size = nlmsg_msg_size(len); |
a46621a3 | 2579 | |
23b45672 | 2580 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2581 | nlh->nlmsg_type = type; |
2582 | nlh->nlmsg_len = size; | |
2583 | nlh->nlmsg_flags = flags; | |
15e47304 | 2584 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2585 | nlh->nlmsg_seq = seq; |
2586 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2587 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2588 | return nlh; |
2589 | } | |
2590 | EXPORT_SYMBOL(__nlmsg_put); | |
2591 | ||
1da177e4 LT |
2592 | /* |
2593 | * It looks a bit ugly. | |
2594 | * It would be better to create kernel thread. | |
2595 | */ | |
2596 | ||
2597 | static int netlink_dump(struct sock *sk) | |
2598 | { | |
2599 | struct netlink_sock *nlk = nlk_sk(sk); | |
2600 | struct netlink_callback *cb; | |
c7ac8679 | 2601 | struct sk_buff *skb = NULL; |
1da177e4 | 2602 | struct nlmsghdr *nlh; |
bf8b79e4 | 2603 | int len, err = -ENOBUFS; |
c7ac8679 | 2604 | int alloc_size; |
1da177e4 | 2605 | |
af65bdfc | 2606 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2607 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2608 | err = -EINVAL; |
2609 | goto errout_skb; | |
1da177e4 LT |
2610 | } |
2611 | ||
16b304f3 | 2612 | cb = &nlk->cb; |
c7ac8679 GR |
2613 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2614 | ||
f9c22888 PM |
2615 | if (!netlink_rx_is_mmaped(sk) && |
2616 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2617 | goto errout_skb; | |
9063e21f ED |
2618 | |
2619 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2620 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2621 | * to reduce number of system calls on dump operations, if user | |
2622 | * ever provided a big enough buffer. | |
2623 | */ | |
2624 | if (alloc_size < nlk->max_recvmsg_len) { | |
2625 | skb = netlink_alloc_skb(sk, | |
2626 | nlk->max_recvmsg_len, | |
2627 | nlk->portid, | |
2628 | GFP_KERNEL | | |
2629 | __GFP_NOWARN | | |
2630 | __GFP_NORETRY); | |
2631 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2632 | if (skb) | |
2633 | skb_reserve(skb, skb_tailroom(skb) - | |
2634 | nlk->max_recvmsg_len); | |
2635 | } | |
2636 | if (!skb) | |
2637 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2638 | GFP_KERNEL); | |
c7ac8679 | 2639 | if (!skb) |
c63d6ea3 | 2640 | goto errout_skb; |
f9c22888 | 2641 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2642 | |
1da177e4 LT |
2643 | len = cb->dump(skb, cb); |
2644 | ||
2645 | if (len > 0) { | |
af65bdfc | 2646 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2647 | |
2648 | if (sk_filter(sk, skb)) | |
2649 | kfree_skb(skb); | |
4a7e7c2a ED |
2650 | else |
2651 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2652 | return 0; |
2653 | } | |
2654 | ||
bf8b79e4 TG |
2655 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2656 | if (!nlh) | |
2657 | goto errout_skb; | |
2658 | ||
670dc283 JB |
2659 | nl_dump_check_consistent(cb, nlh); |
2660 | ||
bf8b79e4 TG |
2661 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2662 | ||
b1153f29 SH |
2663 | if (sk_filter(sk, skb)) |
2664 | kfree_skb(skb); | |
4a7e7c2a ED |
2665 | else |
2666 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2667 | |
a8f74b22 TG |
2668 | if (cb->done) |
2669 | cb->done(cb); | |
1da177e4 | 2670 | |
16b304f3 PS |
2671 | nlk->cb_running = false; |
2672 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2673 | module_put(cb->module); |
16b304f3 | 2674 | consume_skb(cb->skb); |
1da177e4 | 2675 | return 0; |
1797754e | 2676 | |
bf8b79e4 | 2677 | errout_skb: |
af65bdfc | 2678 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2679 | kfree_skb(skb); |
bf8b79e4 | 2680 | return err; |
1da177e4 LT |
2681 | } |
2682 | ||
6dc878a8 G |
2683 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2684 | const struct nlmsghdr *nlh, | |
2685 | struct netlink_dump_control *control) | |
1da177e4 LT |
2686 | { |
2687 | struct netlink_callback *cb; | |
2688 | struct sock *sk; | |
2689 | struct netlink_sock *nlk; | |
b44d211e | 2690 | int ret; |
1da177e4 | 2691 | |
f9c22888 PM |
2692 | /* Memory mapped dump requests need to be copied to avoid looping |
2693 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2694 | * a reference to the skb. | |
2695 | */ | |
2696 | if (netlink_skb_is_mmaped(skb)) { | |
2697 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2698 | if (skb == NULL) |
f9c22888 | 2699 | return -ENOBUFS; |
f9c22888 PM |
2700 | } else |
2701 | atomic_inc(&skb->users); | |
2702 | ||
15e47304 | 2703 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2704 | if (sk == NULL) { |
16b304f3 PS |
2705 | ret = -ECONNREFUSED; |
2706 | goto error_free; | |
1da177e4 | 2707 | } |
6dc878a8 | 2708 | |
16b304f3 | 2709 | nlk = nlk_sk(sk); |
af65bdfc | 2710 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2711 | /* A dump is in progress... */ |
16b304f3 | 2712 | if (nlk->cb_running) { |
6dc878a8 | 2713 | ret = -EBUSY; |
16b304f3 | 2714 | goto error_unlock; |
1da177e4 | 2715 | } |
6dc878a8 | 2716 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2717 | if (!try_module_get(control->module)) { |
6dc878a8 | 2718 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2719 | goto error_unlock; |
6dc878a8 G |
2720 | } |
2721 | ||
16b304f3 PS |
2722 | cb = &nlk->cb; |
2723 | memset(cb, 0, sizeof(*cb)); | |
2724 | cb->dump = control->dump; | |
2725 | cb->done = control->done; | |
2726 | cb->nlh = nlh; | |
2727 | cb->data = control->data; | |
2728 | cb->module = control->module; | |
2729 | cb->min_dump_alloc = control->min_dump_alloc; | |
2730 | cb->skb = skb; | |
2731 | ||
2732 | nlk->cb_running = true; | |
2733 | ||
af65bdfc | 2734 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2735 | |
b44d211e | 2736 | ret = netlink_dump(sk); |
1da177e4 | 2737 | sock_put(sk); |
5c58298c | 2738 | |
b44d211e AV |
2739 | if (ret) |
2740 | return ret; | |
2741 | ||
5c58298c DL |
2742 | /* We successfully started a dump, by returning -EINTR we |
2743 | * signal not to send ACK even if it was requested. | |
2744 | */ | |
2745 | return -EINTR; | |
16b304f3 PS |
2746 | |
2747 | error_unlock: | |
2748 | sock_put(sk); | |
2749 | mutex_unlock(nlk->cb_mutex); | |
2750 | error_free: | |
2751 | kfree_skb(skb); | |
2752 | return ret; | |
1da177e4 | 2753 | } |
6dc878a8 | 2754 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2755 | |
2756 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2757 | { | |
2758 | struct sk_buff *skb; | |
2759 | struct nlmsghdr *rep; | |
2760 | struct nlmsgerr *errmsg; | |
339bf98f | 2761 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2762 | |
339bf98f TG |
2763 | /* error messages get the original request appened */ |
2764 | if (err) | |
2765 | payload += nlmsg_len(nlh); | |
1da177e4 | 2766 | |
f9c22888 PM |
2767 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2768 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2769 | if (!skb) { |
2770 | struct sock *sk; | |
2771 | ||
3b1e0a65 | 2772 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2773 | in_skb->sk->sk_protocol, |
15e47304 | 2774 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2775 | if (sk) { |
2776 | sk->sk_err = ENOBUFS; | |
2777 | sk->sk_error_report(sk); | |
2778 | sock_put(sk); | |
2779 | } | |
2780 | return; | |
2781 | } | |
2782 | ||
15e47304 | 2783 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2784 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2785 | errmsg = nlmsg_data(rep); |
1da177e4 | 2786 | errmsg->error = err; |
bf8b79e4 | 2787 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2788 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2789 | } |
6ac552fd | 2790 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2791 | |
cd40b7d3 | 2792 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2793 | struct nlmsghdr *)) |
82ace47a | 2794 | { |
82ace47a TG |
2795 | struct nlmsghdr *nlh; |
2796 | int err; | |
2797 | ||
2798 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2799 | int msglen; |
2800 | ||
b529ccf2 | 2801 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2802 | err = 0; |
82ace47a | 2803 | |
ad8e4b75 | 2804 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2805 | return 0; |
2806 | ||
d35b6856 TG |
2807 | /* Only requests are handled by the kernel */ |
2808 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2809 | goto ack; |
45e7ae7f TG |
2810 | |
2811 | /* Skip control messages */ | |
2812 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2813 | goto ack; |
d35b6856 | 2814 | |
1d00a4eb | 2815 | err = cb(skb, nlh); |
5c58298c DL |
2816 | if (err == -EINTR) |
2817 | goto skip; | |
2818 | ||
2819 | ack: | |
d35b6856 | 2820 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2821 | netlink_ack(skb, nlh, err); |
82ace47a | 2822 | |
5c58298c | 2823 | skip: |
6ac552fd | 2824 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2825 | if (msglen > skb->len) |
2826 | msglen = skb->len; | |
2827 | skb_pull(skb, msglen); | |
82ace47a TG |
2828 | } |
2829 | ||
2830 | return 0; | |
2831 | } | |
6ac552fd | 2832 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2833 | |
d387f6ad TG |
2834 | /** |
2835 | * nlmsg_notify - send a notification netlink message | |
2836 | * @sk: netlink socket to use | |
2837 | * @skb: notification message | |
15e47304 | 2838 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2839 | * @group: destination multicast group or 0 |
2840 | * @report: 1 to report back, 0 to disable | |
2841 | * @flags: allocation flags | |
2842 | */ | |
15e47304 | 2843 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2844 | unsigned int group, int report, gfp_t flags) |
2845 | { | |
2846 | int err = 0; | |
2847 | ||
2848 | if (group) { | |
15e47304 | 2849 | int exclude_portid = 0; |
d387f6ad TG |
2850 | |
2851 | if (report) { | |
2852 | atomic_inc(&skb->users); | |
15e47304 | 2853 | exclude_portid = portid; |
d387f6ad TG |
2854 | } |
2855 | ||
1ce85fe4 PNA |
2856 | /* errors reported via destination sk->sk_err, but propagate |
2857 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2858 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2859 | } |
2860 | ||
1ce85fe4 PNA |
2861 | if (report) { |
2862 | int err2; | |
2863 | ||
15e47304 | 2864 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2865 | if (!err || err == -ESRCH) |
2866 | err = err2; | |
2867 | } | |
d387f6ad TG |
2868 | |
2869 | return err; | |
2870 | } | |
6ac552fd | 2871 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2872 | |
1da177e4 LT |
2873 | #ifdef CONFIG_PROC_FS |
2874 | struct nl_seq_iter { | |
e372c414 | 2875 | struct seq_net_private p; |
1da177e4 LT |
2876 | int link; |
2877 | int hash_idx; | |
2878 | }; | |
2879 | ||
2880 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2881 | { | |
2882 | struct nl_seq_iter *iter = seq->private; | |
2883 | int i, j; | |
e341694e | 2884 | struct netlink_sock *nlk; |
1da177e4 | 2885 | struct sock *s; |
1da177e4 LT |
2886 | loff_t off = 0; |
2887 | ||
6ac552fd | 2888 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e | 2889 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2890 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
e341694e TG |
2891 | |
2892 | for (j = 0; j < tbl->size; j++) { | |
88d6ed15 TG |
2893 | struct rhash_head *node; |
2894 | ||
2895 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { | |
e341694e | 2896 | s = (struct sock *)nlk; |
1da177e4 | 2897 | |
1218854a | 2898 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2899 | continue; |
1da177e4 LT |
2900 | if (off == pos) { |
2901 | iter->link = i; | |
2902 | iter->hash_idx = j; | |
2903 | return s; | |
2904 | } | |
2905 | ++off; | |
2906 | } | |
2907 | } | |
2908 | } | |
2909 | return NULL; | |
2910 | } | |
2911 | ||
2912 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
78fd1d0a | 2913 | __acquires(nl_table_lock) __acquires(RCU) |
1da177e4 | 2914 | { |
78fd1d0a | 2915 | read_lock(&nl_table_lock); |
e341694e | 2916 | rcu_read_lock(); |
1da177e4 LT |
2917 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
2918 | } | |
2919 | ||
2920 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2921 | { | |
78fd1d0a | 2922 | struct rhashtable *ht; |
88d6ed15 TG |
2923 | const struct bucket_table *tbl; |
2924 | struct rhash_head *node; | |
e341694e | 2925 | struct netlink_sock *nlk; |
1da177e4 | 2926 | struct nl_seq_iter *iter; |
da12c90e | 2927 | struct net *net; |
1da177e4 LT |
2928 | int i, j; |
2929 | ||
2930 | ++*pos; | |
2931 | ||
2932 | if (v == SEQ_START_TOKEN) | |
2933 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2934 | |
da12c90e | 2935 | net = seq_file_net(seq); |
b4b51029 | 2936 | iter = seq->private; |
e341694e TG |
2937 | nlk = v; |
2938 | ||
78fd1d0a TG |
2939 | i = iter->link; |
2940 | ht = &nl_table[i].hash; | |
88d6ed15 TG |
2941 | tbl = rht_dereference_rcu(ht->tbl, ht); |
2942 | rht_for_each_entry_rcu_continue(nlk, node, nlk->node.next, tbl, iter->hash_idx, node) | |
e341694e TG |
2943 | if (net_eq(sock_net((struct sock *)nlk), net)) |
2944 | return nlk; | |
1da177e4 | 2945 | |
1da177e4 LT |
2946 | j = iter->hash_idx + 1; |
2947 | ||
2948 | do { | |
da12c90e | 2949 | |
e341694e | 2950 | for (; j < tbl->size; j++) { |
88d6ed15 | 2951 | rht_for_each_entry_rcu(nlk, node, tbl, j, node) { |
e341694e TG |
2952 | if (net_eq(sock_net((struct sock *)nlk), net)) { |
2953 | iter->link = i; | |
2954 | iter->hash_idx = j; | |
2955 | return nlk; | |
2956 | } | |
1da177e4 LT |
2957 | } |
2958 | } | |
2959 | ||
2960 | j = 0; | |
2961 | } while (++i < MAX_LINKS); | |
2962 | ||
2963 | return NULL; | |
2964 | } | |
2965 | ||
2966 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
78fd1d0a | 2967 | __releases(RCU) __releases(nl_table_lock) |
1da177e4 | 2968 | { |
e341694e | 2969 | rcu_read_unlock(); |
78fd1d0a | 2970 | read_unlock(&nl_table_lock); |
1da177e4 LT |
2971 | } |
2972 | ||
2973 | ||
2974 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2975 | { | |
658cb354 | 2976 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2977 | seq_puts(seq, |
2978 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2979 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2980 | } else { |
1da177e4 LT |
2981 | struct sock *s = v; |
2982 | struct netlink_sock *nlk = nlk_sk(s); | |
2983 | ||
16b304f3 | 2984 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2985 | s, |
2986 | s->sk_protocol, | |
15e47304 | 2987 | nlk->portid, |
513c2500 | 2988 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2989 | sk_rmem_alloc_get(s), |
2990 | sk_wmem_alloc_get(s), | |
16b304f3 | 2991 | nlk->cb_running, |
38938bfe | 2992 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2993 | atomic_read(&s->sk_drops), |
2994 | sock_i_ino(s) | |
1da177e4 LT |
2995 | ); |
2996 | ||
2997 | } | |
2998 | return 0; | |
2999 | } | |
3000 | ||
56b3d975 | 3001 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3002 | .start = netlink_seq_start, |
3003 | .next = netlink_seq_next, | |
3004 | .stop = netlink_seq_stop, | |
3005 | .show = netlink_seq_show, | |
3006 | }; | |
3007 | ||
3008 | ||
3009 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3010 | { | |
e372c414 DL |
3011 | return seq_open_net(inode, file, &netlink_seq_ops, |
3012 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3013 | } |
3014 | ||
da7071d7 | 3015 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3016 | .owner = THIS_MODULE, |
3017 | .open = netlink_seq_open, | |
3018 | .read = seq_read, | |
3019 | .llseek = seq_lseek, | |
e372c414 | 3020 | .release = seq_release_net, |
1da177e4 LT |
3021 | }; |
3022 | ||
3023 | #endif | |
3024 | ||
3025 | int netlink_register_notifier(struct notifier_block *nb) | |
3026 | { | |
e041c683 | 3027 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3028 | } |
6ac552fd | 3029 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3030 | |
3031 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3032 | { | |
e041c683 | 3033 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3034 | } |
6ac552fd | 3035 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3036 | |
90ddc4f0 | 3037 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3038 | .family = PF_NETLINK, |
3039 | .owner = THIS_MODULE, | |
3040 | .release = netlink_release, | |
3041 | .bind = netlink_bind, | |
3042 | .connect = netlink_connect, | |
3043 | .socketpair = sock_no_socketpair, | |
3044 | .accept = sock_no_accept, | |
3045 | .getname = netlink_getname, | |
9652e931 | 3046 | .poll = netlink_poll, |
1da177e4 LT |
3047 | .ioctl = sock_no_ioctl, |
3048 | .listen = sock_no_listen, | |
3049 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3050 | .setsockopt = netlink_setsockopt, |
3051 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3052 | .sendmsg = netlink_sendmsg, |
3053 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3054 | .mmap = netlink_mmap, |
1da177e4 LT |
3055 | .sendpage = sock_no_sendpage, |
3056 | }; | |
3057 | ||
ec1b4cf7 | 3058 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3059 | .family = PF_NETLINK, |
3060 | .create = netlink_create, | |
3061 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3062 | }; | |
3063 | ||
4665079c | 3064 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3065 | { |
3066 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3067 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3068 | return -ENOMEM; |
3069 | #endif | |
3070 | return 0; | |
3071 | } | |
3072 | ||
4665079c | 3073 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3074 | { |
3075 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3076 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3077 | #endif |
3078 | } | |
3079 | ||
b963ea89 DM |
3080 | static void __init netlink_add_usersock_entry(void) |
3081 | { | |
5c398dc8 | 3082 | struct listeners *listeners; |
b963ea89 DM |
3083 | int groups = 32; |
3084 | ||
5c398dc8 | 3085 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3086 | if (!listeners) |
5c398dc8 | 3087 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3088 | |
3089 | netlink_table_grab(); | |
3090 | ||
3091 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3092 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3093 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3094 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3095 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3096 | |
3097 | netlink_table_ungrab(); | |
3098 | } | |
3099 | ||
022cbae6 | 3100 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3101 | .init = netlink_net_init, |
3102 | .exit = netlink_net_exit, | |
3103 | }; | |
3104 | ||
1da177e4 LT |
3105 | static int __init netlink_proto_init(void) |
3106 | { | |
1da177e4 | 3107 | int i; |
1da177e4 | 3108 | int err = proto_register(&netlink_proto, 0); |
e341694e TG |
3109 | struct rhashtable_params ht_params = { |
3110 | .head_offset = offsetof(struct netlink_sock, node), | |
3111 | .key_offset = offsetof(struct netlink_sock, portid), | |
3112 | .key_len = sizeof(u32), /* portid */ | |
7f19fc5e | 3113 | .hashfn = jhash, |
e341694e TG |
3114 | .max_shift = 16, /* 64K */ |
3115 | .grow_decision = rht_grow_above_75, | |
3116 | .shrink_decision = rht_shrink_below_30, | |
e341694e | 3117 | }; |
1da177e4 LT |
3118 | |
3119 | if (err != 0) | |
3120 | goto out; | |
3121 | ||
fab25745 | 3122 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3123 | |
0da974f4 | 3124 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3125 | if (!nl_table) |
3126 | goto panic; | |
1da177e4 | 3127 | |
1da177e4 | 3128 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e TG |
3129 | if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) { |
3130 | while (--i > 0) | |
3131 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3132 | kfree(nl_table); |
fab2caf6 | 3133 | goto panic; |
1da177e4 | 3134 | } |
1da177e4 LT |
3135 | } |
3136 | ||
bcbde0d4 DB |
3137 | INIT_LIST_HEAD(&netlink_tap_all); |
3138 | ||
b963ea89 DM |
3139 | netlink_add_usersock_entry(); |
3140 | ||
1da177e4 | 3141 | sock_register(&netlink_family_ops); |
b4b51029 | 3142 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3143 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3144 | rtnetlink_init(); |
3145 | out: | |
3146 | return err; | |
fab2caf6 AM |
3147 | panic: |
3148 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3149 | } |
3150 | ||
1da177e4 | 3151 | core_initcall(netlink_proto_init); |