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 | |
0f29c768 AV |
99 | DEFINE_RWLOCK(nl_table_lock); |
100 | EXPORT_SYMBOL_GPL(nl_table_lock); | |
1da177e4 LT |
101 | static atomic_t nl_table_users = ATOMIC_INIT(0); |
102 | ||
6d772ac5 ED |
103 | #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock)); |
104 | ||
e341694e TG |
105 | /* Protects netlink socket hash table mutations */ |
106 | DEFINE_MUTEX(nl_sk_hash_lock); | |
107 | ||
108 | static int lockdep_nl_sk_hash_is_held(void) | |
109 | { | |
110 | #ifdef CONFIG_LOCKDEP | |
111 | return (debug_locks) ? lockdep_is_held(&nl_sk_hash_lock) : 1; | |
112 | #else | |
113 | return 1; | |
114 | #endif | |
115 | } | |
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 | ||
136 | if (nt->module) | |
137 | __module_get(nt->module); | |
138 | ||
139 | return 0; | |
140 | } | |
141 | EXPORT_SYMBOL_GPL(netlink_add_tap); | |
142 | ||
2173f8d9 | 143 | static int __netlink_remove_tap(struct netlink_tap *nt) |
bcbde0d4 DB |
144 | { |
145 | bool found = false; | |
146 | struct netlink_tap *tmp; | |
147 | ||
148 | spin_lock(&netlink_tap_lock); | |
149 | ||
150 | list_for_each_entry(tmp, &netlink_tap_all, list) { | |
151 | if (nt == tmp) { | |
152 | list_del_rcu(&nt->list); | |
153 | found = true; | |
154 | goto out; | |
155 | } | |
156 | } | |
157 | ||
158 | pr_warn("__netlink_remove_tap: %p not found\n", nt); | |
159 | out: | |
160 | spin_unlock(&netlink_tap_lock); | |
161 | ||
162 | if (found && nt->module) | |
163 | module_put(nt->module); | |
164 | ||
165 | return found ? 0 : -ENODEV; | |
166 | } | |
bcbde0d4 DB |
167 | |
168 | int netlink_remove_tap(struct netlink_tap *nt) | |
169 | { | |
170 | int ret; | |
171 | ||
172 | ret = __netlink_remove_tap(nt); | |
173 | synchronize_net(); | |
174 | ||
175 | return ret; | |
176 | } | |
177 | EXPORT_SYMBOL_GPL(netlink_remove_tap); | |
178 | ||
5ffd5cdd DB |
179 | static bool netlink_filter_tap(const struct sk_buff *skb) |
180 | { | |
181 | struct sock *sk = skb->sk; | |
5ffd5cdd DB |
182 | |
183 | /* We take the more conservative approach and | |
184 | * whitelist socket protocols that may pass. | |
185 | */ | |
186 | switch (sk->sk_protocol) { | |
187 | case NETLINK_ROUTE: | |
188 | case NETLINK_USERSOCK: | |
189 | case NETLINK_SOCK_DIAG: | |
190 | case NETLINK_NFLOG: | |
191 | case NETLINK_XFRM: | |
192 | case NETLINK_FIB_LOOKUP: | |
193 | case NETLINK_NETFILTER: | |
194 | case NETLINK_GENERIC: | |
498044bb | 195 | return true; |
5ffd5cdd DB |
196 | } |
197 | ||
498044bb | 198 | return false; |
5ffd5cdd DB |
199 | } |
200 | ||
bcbde0d4 DB |
201 | static int __netlink_deliver_tap_skb(struct sk_buff *skb, |
202 | struct net_device *dev) | |
203 | { | |
204 | struct sk_buff *nskb; | |
5ffd5cdd | 205 | struct sock *sk = skb->sk; |
bcbde0d4 DB |
206 | int ret = -ENOMEM; |
207 | ||
208 | dev_hold(dev); | |
209 | nskb = skb_clone(skb, GFP_ATOMIC); | |
210 | if (nskb) { | |
211 | nskb->dev = dev; | |
5ffd5cdd | 212 | nskb->protocol = htons((u16) sk->sk_protocol); |
604d13c9 DB |
213 | nskb->pkt_type = netlink_is_kernel(sk) ? |
214 | PACKET_KERNEL : PACKET_USER; | |
5ffd5cdd | 215 | |
bcbde0d4 DB |
216 | ret = dev_queue_xmit(nskb); |
217 | if (unlikely(ret > 0)) | |
218 | ret = net_xmit_errno(ret); | |
219 | } | |
220 | ||
221 | dev_put(dev); | |
222 | return ret; | |
223 | } | |
224 | ||
225 | static void __netlink_deliver_tap(struct sk_buff *skb) | |
226 | { | |
227 | int ret; | |
228 | struct netlink_tap *tmp; | |
229 | ||
5ffd5cdd DB |
230 | if (!netlink_filter_tap(skb)) |
231 | return; | |
232 | ||
bcbde0d4 DB |
233 | list_for_each_entry_rcu(tmp, &netlink_tap_all, list) { |
234 | ret = __netlink_deliver_tap_skb(skb, tmp->dev); | |
235 | if (unlikely(ret)) | |
236 | break; | |
237 | } | |
238 | } | |
239 | ||
240 | static void netlink_deliver_tap(struct sk_buff *skb) | |
241 | { | |
242 | rcu_read_lock(); | |
243 | ||
244 | if (unlikely(!list_empty(&netlink_tap_all))) | |
245 | __netlink_deliver_tap(skb); | |
246 | ||
247 | rcu_read_unlock(); | |
248 | } | |
249 | ||
73bfd370 DB |
250 | static void netlink_deliver_tap_kernel(struct sock *dst, struct sock *src, |
251 | struct sk_buff *skb) | |
252 | { | |
253 | if (!(netlink_is_kernel(dst) && netlink_is_kernel(src))) | |
254 | netlink_deliver_tap(skb); | |
255 | } | |
256 | ||
cd1df525 PM |
257 | static void netlink_overrun(struct sock *sk) |
258 | { | |
259 | struct netlink_sock *nlk = nlk_sk(sk); | |
260 | ||
261 | if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) { | |
262 | if (!test_and_set_bit(NETLINK_CONGESTED, &nlk_sk(sk)->state)) { | |
263 | sk->sk_err = ENOBUFS; | |
264 | sk->sk_error_report(sk); | |
265 | } | |
266 | } | |
267 | atomic_inc(&sk->sk_drops); | |
268 | } | |
269 | ||
270 | static void netlink_rcv_wake(struct sock *sk) | |
271 | { | |
272 | struct netlink_sock *nlk = nlk_sk(sk); | |
273 | ||
274 | if (skb_queue_empty(&sk->sk_receive_queue)) | |
275 | clear_bit(NETLINK_CONGESTED, &nlk->state); | |
276 | if (!test_bit(NETLINK_CONGESTED, &nlk->state)) | |
277 | wake_up_interruptible(&nlk->wait); | |
278 | } | |
279 | ||
ccdfcc39 | 280 | #ifdef CONFIG_NETLINK_MMAP |
9652e931 PM |
281 | static bool netlink_skb_is_mmaped(const struct sk_buff *skb) |
282 | { | |
283 | return NETLINK_CB(skb).flags & NETLINK_SKB_MMAPED; | |
284 | } | |
285 | ||
f9c22888 PM |
286 | static bool netlink_rx_is_mmaped(struct sock *sk) |
287 | { | |
288 | return nlk_sk(sk)->rx_ring.pg_vec != NULL; | |
289 | } | |
290 | ||
5fd96123 PM |
291 | static bool netlink_tx_is_mmaped(struct sock *sk) |
292 | { | |
293 | return nlk_sk(sk)->tx_ring.pg_vec != NULL; | |
294 | } | |
295 | ||
ccdfcc39 PM |
296 | static __pure struct page *pgvec_to_page(const void *addr) |
297 | { | |
298 | if (is_vmalloc_addr(addr)) | |
299 | return vmalloc_to_page(addr); | |
300 | else | |
301 | return virt_to_page(addr); | |
302 | } | |
303 | ||
304 | static void free_pg_vec(void **pg_vec, unsigned int order, unsigned int len) | |
305 | { | |
306 | unsigned int i; | |
307 | ||
308 | for (i = 0; i < len; i++) { | |
309 | if (pg_vec[i] != NULL) { | |
310 | if (is_vmalloc_addr(pg_vec[i])) | |
311 | vfree(pg_vec[i]); | |
312 | else | |
313 | free_pages((unsigned long)pg_vec[i], order); | |
314 | } | |
315 | } | |
316 | kfree(pg_vec); | |
317 | } | |
318 | ||
319 | static void *alloc_one_pg_vec_page(unsigned long order) | |
320 | { | |
321 | void *buffer; | |
322 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | | |
323 | __GFP_NOWARN | __GFP_NORETRY; | |
324 | ||
325 | buffer = (void *)__get_free_pages(gfp_flags, order); | |
326 | if (buffer != NULL) | |
327 | return buffer; | |
328 | ||
329 | buffer = vzalloc((1 << order) * PAGE_SIZE); | |
330 | if (buffer != NULL) | |
331 | return buffer; | |
332 | ||
333 | gfp_flags &= ~__GFP_NORETRY; | |
334 | return (void *)__get_free_pages(gfp_flags, order); | |
335 | } | |
336 | ||
337 | static void **alloc_pg_vec(struct netlink_sock *nlk, | |
338 | struct nl_mmap_req *req, unsigned int order) | |
339 | { | |
340 | unsigned int block_nr = req->nm_block_nr; | |
341 | unsigned int i; | |
8a849bb7 | 342 | void **pg_vec; |
ccdfcc39 PM |
343 | |
344 | pg_vec = kcalloc(block_nr, sizeof(void *), GFP_KERNEL); | |
345 | if (pg_vec == NULL) | |
346 | return NULL; | |
347 | ||
348 | for (i = 0; i < block_nr; i++) { | |
8a849bb7 | 349 | pg_vec[i] = alloc_one_pg_vec_page(order); |
ccdfcc39 PM |
350 | if (pg_vec[i] == NULL) |
351 | goto err1; | |
352 | } | |
353 | ||
354 | return pg_vec; | |
355 | err1: | |
356 | free_pg_vec(pg_vec, order, block_nr); | |
357 | return NULL; | |
358 | } | |
359 | ||
360 | static int netlink_set_ring(struct sock *sk, struct nl_mmap_req *req, | |
361 | bool closing, bool tx_ring) | |
362 | { | |
363 | struct netlink_sock *nlk = nlk_sk(sk); | |
364 | struct netlink_ring *ring; | |
365 | struct sk_buff_head *queue; | |
366 | void **pg_vec = NULL; | |
367 | unsigned int order = 0; | |
368 | int err; | |
369 | ||
370 | ring = tx_ring ? &nlk->tx_ring : &nlk->rx_ring; | |
371 | queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
372 | ||
373 | if (!closing) { | |
374 | if (atomic_read(&nlk->mapped)) | |
375 | return -EBUSY; | |
376 | if (atomic_read(&ring->pending)) | |
377 | return -EBUSY; | |
378 | } | |
379 | ||
380 | if (req->nm_block_nr) { | |
381 | if (ring->pg_vec != NULL) | |
382 | return -EBUSY; | |
383 | ||
384 | if ((int)req->nm_block_size <= 0) | |
385 | return -EINVAL; | |
74e83b23 | 386 | if (!PAGE_ALIGNED(req->nm_block_size)) |
ccdfcc39 PM |
387 | return -EINVAL; |
388 | if (req->nm_frame_size < NL_MMAP_HDRLEN) | |
389 | return -EINVAL; | |
390 | if (!IS_ALIGNED(req->nm_frame_size, NL_MMAP_MSG_ALIGNMENT)) | |
391 | return -EINVAL; | |
392 | ||
393 | ring->frames_per_block = req->nm_block_size / | |
394 | req->nm_frame_size; | |
395 | if (ring->frames_per_block == 0) | |
396 | return -EINVAL; | |
397 | if (ring->frames_per_block * req->nm_block_nr != | |
398 | req->nm_frame_nr) | |
399 | return -EINVAL; | |
400 | ||
401 | order = get_order(req->nm_block_size); | |
402 | pg_vec = alloc_pg_vec(nlk, req, order); | |
403 | if (pg_vec == NULL) | |
404 | return -ENOMEM; | |
405 | } else { | |
406 | if (req->nm_frame_nr) | |
407 | return -EINVAL; | |
408 | } | |
409 | ||
410 | err = -EBUSY; | |
411 | mutex_lock(&nlk->pg_vec_lock); | |
412 | if (closing || atomic_read(&nlk->mapped) == 0) { | |
413 | err = 0; | |
414 | spin_lock_bh(&queue->lock); | |
415 | ||
416 | ring->frame_max = req->nm_frame_nr - 1; | |
417 | ring->head = 0; | |
418 | ring->frame_size = req->nm_frame_size; | |
419 | ring->pg_vec_pages = req->nm_block_size / PAGE_SIZE; | |
420 | ||
421 | swap(ring->pg_vec_len, req->nm_block_nr); | |
422 | swap(ring->pg_vec_order, order); | |
423 | swap(ring->pg_vec, pg_vec); | |
424 | ||
425 | __skb_queue_purge(queue); | |
426 | spin_unlock_bh(&queue->lock); | |
427 | ||
428 | WARN_ON(atomic_read(&nlk->mapped)); | |
429 | } | |
430 | mutex_unlock(&nlk->pg_vec_lock); | |
431 | ||
432 | if (pg_vec) | |
433 | free_pg_vec(pg_vec, order, req->nm_block_nr); | |
434 | return err; | |
435 | } | |
436 | ||
437 | static void netlink_mm_open(struct vm_area_struct *vma) | |
438 | { | |
439 | struct file *file = vma->vm_file; | |
440 | struct socket *sock = file->private_data; | |
441 | struct sock *sk = sock->sk; | |
442 | ||
443 | if (sk) | |
444 | atomic_inc(&nlk_sk(sk)->mapped); | |
445 | } | |
446 | ||
447 | static void netlink_mm_close(struct vm_area_struct *vma) | |
448 | { | |
449 | struct file *file = vma->vm_file; | |
450 | struct socket *sock = file->private_data; | |
451 | struct sock *sk = sock->sk; | |
452 | ||
453 | if (sk) | |
454 | atomic_dec(&nlk_sk(sk)->mapped); | |
455 | } | |
456 | ||
457 | static const struct vm_operations_struct netlink_mmap_ops = { | |
458 | .open = netlink_mm_open, | |
459 | .close = netlink_mm_close, | |
460 | }; | |
461 | ||
462 | static int netlink_mmap(struct file *file, struct socket *sock, | |
463 | struct vm_area_struct *vma) | |
464 | { | |
465 | struct sock *sk = sock->sk; | |
466 | struct netlink_sock *nlk = nlk_sk(sk); | |
467 | struct netlink_ring *ring; | |
468 | unsigned long start, size, expected; | |
469 | unsigned int i; | |
470 | int err = -EINVAL; | |
471 | ||
472 | if (vma->vm_pgoff) | |
473 | return -EINVAL; | |
474 | ||
475 | mutex_lock(&nlk->pg_vec_lock); | |
476 | ||
477 | expected = 0; | |
478 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
479 | if (ring->pg_vec == NULL) | |
480 | continue; | |
481 | expected += ring->pg_vec_len * ring->pg_vec_pages * PAGE_SIZE; | |
482 | } | |
483 | ||
484 | if (expected == 0) | |
485 | goto out; | |
486 | ||
487 | size = vma->vm_end - vma->vm_start; | |
488 | if (size != expected) | |
489 | goto out; | |
490 | ||
491 | start = vma->vm_start; | |
492 | for (ring = &nlk->rx_ring; ring <= &nlk->tx_ring; ring++) { | |
493 | if (ring->pg_vec == NULL) | |
494 | continue; | |
495 | ||
496 | for (i = 0; i < ring->pg_vec_len; i++) { | |
497 | struct page *page; | |
498 | void *kaddr = ring->pg_vec[i]; | |
499 | unsigned int pg_num; | |
500 | ||
501 | for (pg_num = 0; pg_num < ring->pg_vec_pages; pg_num++) { | |
502 | page = pgvec_to_page(kaddr); | |
503 | err = vm_insert_page(vma, start, page); | |
504 | if (err < 0) | |
505 | goto out; | |
506 | start += PAGE_SIZE; | |
507 | kaddr += PAGE_SIZE; | |
508 | } | |
509 | } | |
510 | } | |
511 | ||
512 | atomic_inc(&nlk->mapped); | |
513 | vma->vm_ops = &netlink_mmap_ops; | |
514 | err = 0; | |
515 | out: | |
516 | mutex_unlock(&nlk->pg_vec_lock); | |
7cdbac71 | 517 | return err; |
ccdfcc39 | 518 | } |
9652e931 PM |
519 | |
520 | static void netlink_frame_flush_dcache(const struct nl_mmap_hdr *hdr) | |
521 | { | |
522 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
523 | struct page *p_start, *p_end; | |
524 | ||
525 | /* First page is flushed through netlink_{get,set}_status */ | |
526 | p_start = pgvec_to_page(hdr + PAGE_SIZE); | |
1d5085cb | 527 | p_end = pgvec_to_page((void *)hdr + NL_MMAP_HDRLEN + hdr->nm_len - 1); |
9652e931 PM |
528 | while (p_start <= p_end) { |
529 | flush_dcache_page(p_start); | |
530 | p_start++; | |
531 | } | |
532 | #endif | |
533 | } | |
534 | ||
535 | static enum nl_mmap_status netlink_get_status(const struct nl_mmap_hdr *hdr) | |
536 | { | |
537 | smp_rmb(); | |
538 | flush_dcache_page(pgvec_to_page(hdr)); | |
539 | return hdr->nm_status; | |
540 | } | |
541 | ||
542 | static void netlink_set_status(struct nl_mmap_hdr *hdr, | |
543 | enum nl_mmap_status status) | |
544 | { | |
545 | hdr->nm_status = status; | |
546 | flush_dcache_page(pgvec_to_page(hdr)); | |
547 | smp_wmb(); | |
548 | } | |
549 | ||
550 | static struct nl_mmap_hdr * | |
551 | __netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos) | |
552 | { | |
553 | unsigned int pg_vec_pos, frame_off; | |
554 | ||
555 | pg_vec_pos = pos / ring->frames_per_block; | |
556 | frame_off = pos % ring->frames_per_block; | |
557 | ||
558 | return ring->pg_vec[pg_vec_pos] + (frame_off * ring->frame_size); | |
559 | } | |
560 | ||
561 | static struct nl_mmap_hdr * | |
562 | netlink_lookup_frame(const struct netlink_ring *ring, unsigned int pos, | |
563 | enum nl_mmap_status status) | |
564 | { | |
565 | struct nl_mmap_hdr *hdr; | |
566 | ||
567 | hdr = __netlink_lookup_frame(ring, pos); | |
568 | if (netlink_get_status(hdr) != status) | |
569 | return NULL; | |
570 | ||
571 | return hdr; | |
572 | } | |
573 | ||
574 | static struct nl_mmap_hdr * | |
575 | netlink_current_frame(const struct netlink_ring *ring, | |
576 | enum nl_mmap_status status) | |
577 | { | |
578 | return netlink_lookup_frame(ring, ring->head, status); | |
579 | } | |
580 | ||
581 | static struct nl_mmap_hdr * | |
582 | netlink_previous_frame(const struct netlink_ring *ring, | |
583 | enum nl_mmap_status status) | |
584 | { | |
585 | unsigned int prev; | |
586 | ||
587 | prev = ring->head ? ring->head - 1 : ring->frame_max; | |
588 | return netlink_lookup_frame(ring, prev, status); | |
589 | } | |
590 | ||
591 | static void netlink_increment_head(struct netlink_ring *ring) | |
592 | { | |
593 | ring->head = ring->head != ring->frame_max ? ring->head + 1 : 0; | |
594 | } | |
595 | ||
596 | static void netlink_forward_ring(struct netlink_ring *ring) | |
597 | { | |
598 | unsigned int head = ring->head, pos = head; | |
599 | const struct nl_mmap_hdr *hdr; | |
600 | ||
601 | do { | |
602 | hdr = __netlink_lookup_frame(ring, pos); | |
603 | if (hdr->nm_status == NL_MMAP_STATUS_UNUSED) | |
604 | break; | |
605 | if (hdr->nm_status != NL_MMAP_STATUS_SKIP) | |
606 | break; | |
607 | netlink_increment_head(ring); | |
608 | } while (ring->head != head); | |
609 | } | |
610 | ||
cd1df525 PM |
611 | static bool netlink_dump_space(struct netlink_sock *nlk) |
612 | { | |
613 | struct netlink_ring *ring = &nlk->rx_ring; | |
614 | struct nl_mmap_hdr *hdr; | |
615 | unsigned int n; | |
616 | ||
617 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
618 | if (hdr == NULL) | |
619 | return false; | |
620 | ||
621 | n = ring->head + ring->frame_max / 2; | |
622 | if (n > ring->frame_max) | |
623 | n -= ring->frame_max; | |
624 | ||
625 | hdr = __netlink_lookup_frame(ring, n); | |
626 | ||
627 | return hdr->nm_status == NL_MMAP_STATUS_UNUSED; | |
628 | } | |
629 | ||
9652e931 PM |
630 | static unsigned int netlink_poll(struct file *file, struct socket *sock, |
631 | poll_table *wait) | |
632 | { | |
633 | struct sock *sk = sock->sk; | |
634 | struct netlink_sock *nlk = nlk_sk(sk); | |
635 | unsigned int mask; | |
cd1df525 | 636 | int err; |
9652e931 | 637 | |
cd1df525 PM |
638 | if (nlk->rx_ring.pg_vec != NULL) { |
639 | /* Memory mapped sockets don't call recvmsg(), so flow control | |
640 | * for dumps is performed here. A dump is allowed to continue | |
641 | * if at least half the ring is unused. | |
642 | */ | |
16b304f3 | 643 | while (nlk->cb_running && netlink_dump_space(nlk)) { |
cd1df525 PM |
644 | err = netlink_dump(sk); |
645 | if (err < 0) { | |
ac30ef83 | 646 | sk->sk_err = -err; |
cd1df525 PM |
647 | sk->sk_error_report(sk); |
648 | break; | |
649 | } | |
650 | } | |
651 | netlink_rcv_wake(sk); | |
652 | } | |
5fd96123 | 653 | |
9652e931 PM |
654 | mask = datagram_poll(file, sock, wait); |
655 | ||
656 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
657 | if (nlk->rx_ring.pg_vec) { | |
658 | netlink_forward_ring(&nlk->rx_ring); | |
659 | if (!netlink_previous_frame(&nlk->rx_ring, NL_MMAP_STATUS_UNUSED)) | |
660 | mask |= POLLIN | POLLRDNORM; | |
661 | } | |
662 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
663 | ||
664 | spin_lock_bh(&sk->sk_write_queue.lock); | |
665 | if (nlk->tx_ring.pg_vec) { | |
666 | if (netlink_current_frame(&nlk->tx_ring, NL_MMAP_STATUS_UNUSED)) | |
667 | mask |= POLLOUT | POLLWRNORM; | |
668 | } | |
669 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
670 | ||
671 | return mask; | |
672 | } | |
673 | ||
674 | static struct nl_mmap_hdr *netlink_mmap_hdr(struct sk_buff *skb) | |
675 | { | |
676 | return (struct nl_mmap_hdr *)(skb->head - NL_MMAP_HDRLEN); | |
677 | } | |
678 | ||
679 | static void netlink_ring_setup_skb(struct sk_buff *skb, struct sock *sk, | |
680 | struct netlink_ring *ring, | |
681 | struct nl_mmap_hdr *hdr) | |
682 | { | |
683 | unsigned int size; | |
684 | void *data; | |
685 | ||
686 | size = ring->frame_size - NL_MMAP_HDRLEN; | |
687 | data = (void *)hdr + NL_MMAP_HDRLEN; | |
688 | ||
689 | skb->head = data; | |
690 | skb->data = data; | |
691 | skb_reset_tail_pointer(skb); | |
692 | skb->end = skb->tail + size; | |
693 | skb->len = 0; | |
694 | ||
695 | skb->destructor = netlink_skb_destructor; | |
696 | NETLINK_CB(skb).flags |= NETLINK_SKB_MMAPED; | |
697 | NETLINK_CB(skb).sk = sk; | |
698 | } | |
5fd96123 PM |
699 | |
700 | static int netlink_mmap_sendmsg(struct sock *sk, struct msghdr *msg, | |
701 | u32 dst_portid, u32 dst_group, | |
702 | struct sock_iocb *siocb) | |
703 | { | |
704 | struct netlink_sock *nlk = nlk_sk(sk); | |
705 | struct netlink_ring *ring; | |
706 | struct nl_mmap_hdr *hdr; | |
707 | struct sk_buff *skb; | |
708 | unsigned int maxlen; | |
709 | bool excl = true; | |
710 | int err = 0, len = 0; | |
711 | ||
712 | /* Netlink messages are validated by the receiver before processing. | |
713 | * In order to avoid userspace changing the contents of the message | |
714 | * after validation, the socket and the ring may only be used by a | |
715 | * single process, otherwise we fall back to copying. | |
716 | */ | |
717 | if (atomic_long_read(&sk->sk_socket->file->f_count) > 2 || | |
718 | atomic_read(&nlk->mapped) > 1) | |
719 | excl = false; | |
720 | ||
721 | mutex_lock(&nlk->pg_vec_lock); | |
722 | ||
723 | ring = &nlk->tx_ring; | |
724 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; | |
725 | ||
726 | do { | |
727 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_VALID); | |
728 | if (hdr == NULL) { | |
729 | if (!(msg->msg_flags & MSG_DONTWAIT) && | |
730 | atomic_read(&nlk->tx_ring.pending)) | |
731 | schedule(); | |
732 | continue; | |
733 | } | |
734 | if (hdr->nm_len > maxlen) { | |
735 | err = -EINVAL; | |
736 | goto out; | |
737 | } | |
738 | ||
739 | netlink_frame_flush_dcache(hdr); | |
740 | ||
741 | if (likely(dst_portid == 0 && dst_group == 0 && excl)) { | |
742 | skb = alloc_skb_head(GFP_KERNEL); | |
743 | if (skb == NULL) { | |
744 | err = -ENOBUFS; | |
745 | goto out; | |
746 | } | |
747 | sock_hold(sk); | |
748 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
749 | NETLINK_CB(skb).flags |= NETLINK_SKB_TX; | |
750 | __skb_put(skb, hdr->nm_len); | |
751 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
752 | atomic_inc(&ring->pending); | |
753 | } else { | |
754 | skb = alloc_skb(hdr->nm_len, GFP_KERNEL); | |
755 | if (skb == NULL) { | |
756 | err = -ENOBUFS; | |
757 | goto out; | |
758 | } | |
759 | __skb_put(skb, hdr->nm_len); | |
760 | memcpy(skb->data, (void *)hdr + NL_MMAP_HDRLEN, hdr->nm_len); | |
761 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
762 | } | |
763 | ||
764 | netlink_increment_head(ring); | |
765 | ||
766 | NETLINK_CB(skb).portid = nlk->portid; | |
767 | NETLINK_CB(skb).dst_group = dst_group; | |
768 | NETLINK_CB(skb).creds = siocb->scm->creds; | |
769 | ||
770 | err = security_netlink_send(sk, skb); | |
771 | if (err) { | |
772 | kfree_skb(skb); | |
773 | goto out; | |
774 | } | |
775 | ||
776 | if (unlikely(dst_group)) { | |
777 | atomic_inc(&skb->users); | |
778 | netlink_broadcast(sk, skb, dst_portid, dst_group, | |
779 | GFP_KERNEL); | |
780 | } | |
781 | err = netlink_unicast(sk, skb, dst_portid, | |
782 | msg->msg_flags & MSG_DONTWAIT); | |
783 | if (err < 0) | |
784 | goto out; | |
785 | len += err; | |
786 | ||
787 | } while (hdr != NULL || | |
788 | (!(msg->msg_flags & MSG_DONTWAIT) && | |
789 | atomic_read(&nlk->tx_ring.pending))); | |
790 | ||
791 | if (len > 0) | |
792 | err = len; | |
793 | out: | |
794 | mutex_unlock(&nlk->pg_vec_lock); | |
795 | return err; | |
796 | } | |
f9c22888 PM |
797 | |
798 | static void netlink_queue_mmaped_skb(struct sock *sk, struct sk_buff *skb) | |
799 | { | |
800 | struct nl_mmap_hdr *hdr; | |
801 | ||
802 | hdr = netlink_mmap_hdr(skb); | |
803 | hdr->nm_len = skb->len; | |
804 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
805 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
806 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
807 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
808 | netlink_frame_flush_dcache(hdr); |
809 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
810 | ||
811 | NETLINK_CB(skb).flags |= NETLINK_SKB_DELIVERED; | |
812 | kfree_skb(skb); | |
813 | } | |
814 | ||
815 | static void netlink_ring_set_copied(struct sock *sk, struct sk_buff *skb) | |
816 | { | |
817 | struct netlink_sock *nlk = nlk_sk(sk); | |
818 | struct netlink_ring *ring = &nlk->rx_ring; | |
819 | struct nl_mmap_hdr *hdr; | |
820 | ||
821 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
822 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
823 | if (hdr == NULL) { | |
824 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
825 | kfree_skb(skb); | |
cd1df525 | 826 | netlink_overrun(sk); |
f9c22888 PM |
827 | return; |
828 | } | |
829 | netlink_increment_head(ring); | |
830 | __skb_queue_tail(&sk->sk_receive_queue, skb); | |
831 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
832 | ||
833 | hdr->nm_len = skb->len; | |
834 | hdr->nm_group = NETLINK_CB(skb).dst_group; | |
835 | hdr->nm_pid = NETLINK_CB(skb).creds.pid; | |
1bf9310a ND |
836 | hdr->nm_uid = from_kuid(sk_user_ns(sk), NETLINK_CB(skb).creds.uid); |
837 | hdr->nm_gid = from_kgid(sk_user_ns(sk), NETLINK_CB(skb).creds.gid); | |
f9c22888 PM |
838 | netlink_set_status(hdr, NL_MMAP_STATUS_COPY); |
839 | } | |
840 | ||
ccdfcc39 | 841 | #else /* CONFIG_NETLINK_MMAP */ |
9652e931 | 842 | #define netlink_skb_is_mmaped(skb) false |
f9c22888 | 843 | #define netlink_rx_is_mmaped(sk) false |
5fd96123 | 844 | #define netlink_tx_is_mmaped(sk) false |
ccdfcc39 | 845 | #define netlink_mmap sock_no_mmap |
9652e931 | 846 | #define netlink_poll datagram_poll |
5fd96123 | 847 | #define netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, siocb) 0 |
ccdfcc39 PM |
848 | #endif /* CONFIG_NETLINK_MMAP */ |
849 | ||
cf0a018a PM |
850 | static void netlink_skb_destructor(struct sk_buff *skb) |
851 | { | |
9652e931 PM |
852 | #ifdef CONFIG_NETLINK_MMAP |
853 | struct nl_mmap_hdr *hdr; | |
854 | struct netlink_ring *ring; | |
855 | struct sock *sk; | |
856 | ||
857 | /* If a packet from the kernel to userspace was freed because of an | |
858 | * error without being delivered to userspace, the kernel must reset | |
859 | * the status. In the direction userspace to kernel, the status is | |
860 | * always reset here after the packet was processed and freed. | |
861 | */ | |
862 | if (netlink_skb_is_mmaped(skb)) { | |
863 | hdr = netlink_mmap_hdr(skb); | |
864 | sk = NETLINK_CB(skb).sk; | |
865 | ||
5fd96123 PM |
866 | if (NETLINK_CB(skb).flags & NETLINK_SKB_TX) { |
867 | netlink_set_status(hdr, NL_MMAP_STATUS_UNUSED); | |
868 | ring = &nlk_sk(sk)->tx_ring; | |
869 | } else { | |
870 | if (!(NETLINK_CB(skb).flags & NETLINK_SKB_DELIVERED)) { | |
871 | hdr->nm_len = 0; | |
872 | netlink_set_status(hdr, NL_MMAP_STATUS_VALID); | |
873 | } | |
874 | ring = &nlk_sk(sk)->rx_ring; | |
9652e931 | 875 | } |
9652e931 PM |
876 | |
877 | WARN_ON(atomic_read(&ring->pending) == 0); | |
878 | atomic_dec(&ring->pending); | |
879 | sock_put(sk); | |
880 | ||
5e71d9d7 | 881 | skb->head = NULL; |
9652e931 PM |
882 | } |
883 | #endif | |
c05cdb1b | 884 | if (is_vmalloc_addr(skb->head)) { |
3a36515f PN |
885 | if (!skb->cloned || |
886 | !atomic_dec_return(&(skb_shinfo(skb)->dataref))) | |
887 | vfree(skb->head); | |
888 | ||
c05cdb1b PNA |
889 | skb->head = NULL; |
890 | } | |
9652e931 PM |
891 | if (skb->sk != NULL) |
892 | sock_rfree(skb); | |
cf0a018a PM |
893 | } |
894 | ||
895 | static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
896 | { | |
897 | WARN_ON(skb->sk != NULL); | |
898 | skb->sk = sk; | |
899 | skb->destructor = netlink_skb_destructor; | |
900 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
901 | sk_mem_charge(sk, skb->truesize); | |
902 | } | |
903 | ||
1da177e4 LT |
904 | static void netlink_sock_destruct(struct sock *sk) |
905 | { | |
3f660d66 HX |
906 | struct netlink_sock *nlk = nlk_sk(sk); |
907 | ||
16b304f3 PS |
908 | if (nlk->cb_running) { |
909 | if (nlk->cb.done) | |
910 | nlk->cb.done(&nlk->cb); | |
6dc878a8 | 911 | |
16b304f3 PS |
912 | module_put(nlk->cb.module); |
913 | kfree_skb(nlk->cb.skb); | |
3f660d66 HX |
914 | } |
915 | ||
1da177e4 | 916 | skb_queue_purge(&sk->sk_receive_queue); |
ccdfcc39 PM |
917 | #ifdef CONFIG_NETLINK_MMAP |
918 | if (1) { | |
919 | struct nl_mmap_req req; | |
920 | ||
921 | memset(&req, 0, sizeof(req)); | |
922 | if (nlk->rx_ring.pg_vec) | |
923 | netlink_set_ring(sk, &req, true, false); | |
924 | memset(&req, 0, sizeof(req)); | |
925 | if (nlk->tx_ring.pg_vec) | |
926 | netlink_set_ring(sk, &req, true, true); | |
927 | } | |
928 | #endif /* CONFIG_NETLINK_MMAP */ | |
1da177e4 LT |
929 | |
930 | if (!sock_flag(sk, SOCK_DEAD)) { | |
6ac552fd | 931 | printk(KERN_ERR "Freeing alive netlink socket %p\n", sk); |
1da177e4 LT |
932 | return; |
933 | } | |
547b792c IJ |
934 | |
935 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); | |
936 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
937 | WARN_ON(nlk_sk(sk)->groups); | |
1da177e4 LT |
938 | } |
939 | ||
6ac552fd PM |
940 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on |
941 | * SMP. Look, when several writers sleep and reader wakes them up, all but one | |
1da177e4 LT |
942 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves |
943 | * this, _but_ remember, it adds useless work on UP machines. | |
944 | */ | |
945 | ||
d136f1bd | 946 | void netlink_table_grab(void) |
9a429c49 | 947 | __acquires(nl_table_lock) |
1da177e4 | 948 | { |
d136f1bd JB |
949 | might_sleep(); |
950 | ||
6abd219c | 951 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
952 | |
953 | if (atomic_read(&nl_table_users)) { | |
954 | DECLARE_WAITQUEUE(wait, current); | |
955 | ||
956 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
6ac552fd | 957 | for (;;) { |
1da177e4 LT |
958 | set_current_state(TASK_UNINTERRUPTIBLE); |
959 | if (atomic_read(&nl_table_users) == 0) | |
960 | break; | |
6abd219c | 961 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 962 | schedule(); |
6abd219c | 963 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
964 | } |
965 | ||
966 | __set_current_state(TASK_RUNNING); | |
967 | remove_wait_queue(&nl_table_wait, &wait); | |
968 | } | |
969 | } | |
970 | ||
d136f1bd | 971 | void netlink_table_ungrab(void) |
9a429c49 | 972 | __releases(nl_table_lock) |
1da177e4 | 973 | { |
6abd219c | 974 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
975 | wake_up(&nl_table_wait); |
976 | } | |
977 | ||
6ac552fd | 978 | static inline void |
1da177e4 LT |
979 | netlink_lock_table(void) |
980 | { | |
981 | /* read_lock() synchronizes us to netlink_table_grab */ | |
982 | ||
983 | read_lock(&nl_table_lock); | |
984 | atomic_inc(&nl_table_users); | |
985 | read_unlock(&nl_table_lock); | |
986 | } | |
987 | ||
6ac552fd | 988 | static inline void |
1da177e4 LT |
989 | netlink_unlock_table(void) |
990 | { | |
991 | if (atomic_dec_and_test(&nl_table_users)) | |
992 | wake_up(&nl_table_wait); | |
993 | } | |
994 | ||
e341694e | 995 | struct netlink_compare_arg |
1da177e4 | 996 | { |
e341694e TG |
997 | struct net *net; |
998 | u32 portid; | |
999 | }; | |
1da177e4 | 1000 | |
e341694e | 1001 | static bool netlink_compare(void *ptr, void *arg) |
1da177e4 | 1002 | { |
e341694e TG |
1003 | struct netlink_compare_arg *x = arg; |
1004 | struct sock *sk = ptr; | |
1da177e4 | 1005 | |
e341694e TG |
1006 | return nlk_sk(sk)->portid == x->portid && |
1007 | net_eq(sock_net(sk), x->net); | |
1da177e4 LT |
1008 | } |
1009 | ||
e341694e TG |
1010 | static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid, |
1011 | struct net *net) | |
1da177e4 | 1012 | { |
e341694e TG |
1013 | struct netlink_compare_arg arg = { |
1014 | .net = net, | |
1015 | .portid = portid, | |
1016 | }; | |
1017 | u32 hash; | |
1da177e4 | 1018 | |
e341694e | 1019 | hash = rhashtable_hashfn(&table->hash, &portid, sizeof(portid)); |
1da177e4 | 1020 | |
e341694e TG |
1021 | return rhashtable_lookup_compare(&table->hash, hash, |
1022 | &netlink_compare, &arg); | |
1da177e4 LT |
1023 | } |
1024 | ||
e341694e | 1025 | static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid) |
1da177e4 | 1026 | { |
e341694e TG |
1027 | struct netlink_table *table = &nl_table[protocol]; |
1028 | struct sock *sk; | |
1da177e4 | 1029 | |
e341694e TG |
1030 | rcu_read_lock(); |
1031 | sk = __netlink_lookup(table, portid, net); | |
1032 | if (sk) | |
1033 | sock_hold(sk); | |
1034 | rcu_read_unlock(); | |
1da177e4 | 1035 | |
e341694e | 1036 | return sk; |
1da177e4 LT |
1037 | } |
1038 | ||
90ddc4f0 | 1039 | static const struct proto_ops netlink_ops; |
1da177e4 | 1040 | |
4277a083 PM |
1041 | static void |
1042 | netlink_update_listeners(struct sock *sk) | |
1043 | { | |
1044 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
4277a083 PM |
1045 | unsigned long mask; |
1046 | unsigned int i; | |
6d772ac5 ED |
1047 | struct listeners *listeners; |
1048 | ||
1049 | listeners = nl_deref_protected(tbl->listeners); | |
1050 | if (!listeners) | |
1051 | return; | |
4277a083 | 1052 | |
b4ff4f04 | 1053 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 1054 | mask = 0; |
b67bfe0d | 1055 | sk_for_each_bound(sk, &tbl->mc_list) { |
b4ff4f04 JB |
1056 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) |
1057 | mask |= nlk_sk(sk)->groups[i]; | |
1058 | } | |
6d772ac5 | 1059 | listeners->masks[i] = mask; |
4277a083 PM |
1060 | } |
1061 | /* this function is only called with the netlink table "grabbed", which | |
1062 | * makes sure updates are visible before bind or setsockopt return. */ | |
1063 | } | |
1064 | ||
15e47304 | 1065 | static int netlink_insert(struct sock *sk, struct net *net, u32 portid) |
1da177e4 | 1066 | { |
da12c90e | 1067 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
1da177e4 | 1068 | int err = -EADDRINUSE; |
1da177e4 | 1069 | |
e341694e TG |
1070 | mutex_lock(&nl_sk_hash_lock); |
1071 | if (__netlink_lookup(table, portid, net)) | |
1da177e4 LT |
1072 | goto err; |
1073 | ||
1074 | err = -EBUSY; | |
15e47304 | 1075 | if (nlk_sk(sk)->portid) |
1da177e4 LT |
1076 | goto err; |
1077 | ||
1078 | err = -ENOMEM; | |
e341694e | 1079 | if (BITS_PER_LONG > 32 && unlikely(table->hash.nelems >= UINT_MAX)) |
1da177e4 LT |
1080 | goto err; |
1081 | ||
15e47304 | 1082 | nlk_sk(sk)->portid = portid; |
e341694e TG |
1083 | sock_hold(sk); |
1084 | rhashtable_insert(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL); | |
1da177e4 | 1085 | err = 0; |
1da177e4 | 1086 | err: |
e341694e | 1087 | mutex_unlock(&nl_sk_hash_lock); |
1da177e4 LT |
1088 | return err; |
1089 | } | |
1090 | ||
1091 | static void netlink_remove(struct sock *sk) | |
1092 | { | |
e341694e TG |
1093 | struct netlink_table *table; |
1094 | ||
1095 | mutex_lock(&nl_sk_hash_lock); | |
1096 | table = &nl_table[sk->sk_protocol]; | |
1097 | if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL)) { | |
1098 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
1099 | __sock_put(sk); | |
1100 | } | |
1101 | mutex_unlock(&nl_sk_hash_lock); | |
1102 | ||
1da177e4 | 1103 | netlink_table_grab(); |
f7fa9b10 | 1104 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
1105 | __sk_del_bind_node(sk); |
1106 | netlink_table_ungrab(); | |
1107 | } | |
1108 | ||
1109 | static struct proto netlink_proto = { | |
1110 | .name = "NETLINK", | |
1111 | .owner = THIS_MODULE, | |
1112 | .obj_size = sizeof(struct netlink_sock), | |
1113 | }; | |
1114 | ||
1b8d7ae4 EB |
1115 | static int __netlink_create(struct net *net, struct socket *sock, |
1116 | struct mutex *cb_mutex, int protocol) | |
1da177e4 LT |
1117 | { |
1118 | struct sock *sk; | |
1119 | struct netlink_sock *nlk; | |
ab33a171 PM |
1120 | |
1121 | sock->ops = &netlink_ops; | |
1122 | ||
6257ff21 | 1123 | sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto); |
ab33a171 PM |
1124 | if (!sk) |
1125 | return -ENOMEM; | |
1126 | ||
1127 | sock_init_data(sock, sk); | |
1128 | ||
1129 | nlk = nlk_sk(sk); | |
658cb354 | 1130 | if (cb_mutex) { |
ffa4d721 | 1131 | nlk->cb_mutex = cb_mutex; |
658cb354 | 1132 | } else { |
ffa4d721 PM |
1133 | nlk->cb_mutex = &nlk->cb_def_mutex; |
1134 | mutex_init(nlk->cb_mutex); | |
1135 | } | |
ab33a171 | 1136 | init_waitqueue_head(&nlk->wait); |
ccdfcc39 PM |
1137 | #ifdef CONFIG_NETLINK_MMAP |
1138 | mutex_init(&nlk->pg_vec_lock); | |
1139 | #endif | |
ab33a171 PM |
1140 | |
1141 | sk->sk_destruct = netlink_sock_destruct; | |
1142 | sk->sk_protocol = protocol; | |
1143 | return 0; | |
1144 | } | |
1145 | ||
3f378b68 EP |
1146 | static int netlink_create(struct net *net, struct socket *sock, int protocol, |
1147 | int kern) | |
ab33a171 PM |
1148 | { |
1149 | struct module *module = NULL; | |
af65bdfc | 1150 | struct mutex *cb_mutex; |
f7fa9b10 | 1151 | struct netlink_sock *nlk; |
4f520900 RGB |
1152 | int (*bind)(int group); |
1153 | void (*unbind)(int group); | |
ab33a171 | 1154 | int err = 0; |
1da177e4 LT |
1155 | |
1156 | sock->state = SS_UNCONNECTED; | |
1157 | ||
1158 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
1159 | return -ESOCKTNOSUPPORT; | |
1160 | ||
6ac552fd | 1161 | if (protocol < 0 || protocol >= MAX_LINKS) |
1da177e4 LT |
1162 | return -EPROTONOSUPPORT; |
1163 | ||
77247bbb | 1164 | netlink_lock_table(); |
95a5afca | 1165 | #ifdef CONFIG_MODULES |
ab33a171 | 1166 | if (!nl_table[protocol].registered) { |
77247bbb | 1167 | netlink_unlock_table(); |
4fdb3bb7 | 1168 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 1169 | netlink_lock_table(); |
4fdb3bb7 | 1170 | } |
ab33a171 PM |
1171 | #endif |
1172 | if (nl_table[protocol].registered && | |
1173 | try_module_get(nl_table[protocol].module)) | |
1174 | module = nl_table[protocol].module; | |
974c37e9 AD |
1175 | else |
1176 | err = -EPROTONOSUPPORT; | |
af65bdfc | 1177 | cb_mutex = nl_table[protocol].cb_mutex; |
03292745 | 1178 | bind = nl_table[protocol].bind; |
4f520900 | 1179 | unbind = nl_table[protocol].unbind; |
77247bbb | 1180 | netlink_unlock_table(); |
4fdb3bb7 | 1181 | |
974c37e9 AD |
1182 | if (err < 0) |
1183 | goto out; | |
1184 | ||
6ac552fd PM |
1185 | err = __netlink_create(net, sock, cb_mutex, protocol); |
1186 | if (err < 0) | |
f7fa9b10 PM |
1187 | goto out_module; |
1188 | ||
6f756a8c | 1189 | local_bh_disable(); |
c1fd3b94 | 1190 | sock_prot_inuse_add(net, &netlink_proto, 1); |
6f756a8c DM |
1191 | local_bh_enable(); |
1192 | ||
f7fa9b10 | 1193 | nlk = nlk_sk(sock->sk); |
f7fa9b10 | 1194 | nlk->module = module; |
03292745 | 1195 | nlk->netlink_bind = bind; |
4f520900 | 1196 | nlk->netlink_unbind = unbind; |
ab33a171 PM |
1197 | out: |
1198 | return err; | |
1da177e4 | 1199 | |
ab33a171 PM |
1200 | out_module: |
1201 | module_put(module); | |
1202 | goto out; | |
1da177e4 LT |
1203 | } |
1204 | ||
1205 | static int netlink_release(struct socket *sock) | |
1206 | { | |
1207 | struct sock *sk = sock->sk; | |
1208 | struct netlink_sock *nlk; | |
1209 | ||
1210 | if (!sk) | |
1211 | return 0; | |
1212 | ||
1213 | netlink_remove(sk); | |
ac57b3a9 | 1214 | sock_orphan(sk); |
1da177e4 LT |
1215 | nlk = nlk_sk(sk); |
1216 | ||
3f660d66 HX |
1217 | /* |
1218 | * OK. Socket is unlinked, any packets that arrive now | |
1219 | * will be purged. | |
1220 | */ | |
1da177e4 | 1221 | |
1da177e4 LT |
1222 | sock->sk = NULL; |
1223 | wake_up_interruptible_all(&nlk->wait); | |
1224 | ||
1225 | skb_queue_purge(&sk->sk_write_queue); | |
1226 | ||
15e47304 | 1227 | if (nlk->portid) { |
1da177e4 | 1228 | struct netlink_notify n = { |
3b1e0a65 | 1229 | .net = sock_net(sk), |
1da177e4 | 1230 | .protocol = sk->sk_protocol, |
15e47304 | 1231 | .portid = nlk->portid, |
1da177e4 | 1232 | }; |
e041c683 AS |
1233 | atomic_notifier_call_chain(&netlink_chain, |
1234 | NETLINK_URELEASE, &n); | |
746fac4d | 1235 | } |
4fdb3bb7 | 1236 | |
5e7c001c | 1237 | module_put(nlk->module); |
4fdb3bb7 | 1238 | |
4277a083 | 1239 | netlink_table_grab(); |
aed81560 | 1240 | if (netlink_is_kernel(sk)) { |
869e58f8 DL |
1241 | BUG_ON(nl_table[sk->sk_protocol].registered == 0); |
1242 | if (--nl_table[sk->sk_protocol].registered == 0) { | |
6d772ac5 ED |
1243 | struct listeners *old; |
1244 | ||
1245 | old = nl_deref_protected(nl_table[sk->sk_protocol].listeners); | |
1246 | RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL); | |
1247 | kfree_rcu(old, rcu); | |
869e58f8 | 1248 | nl_table[sk->sk_protocol].module = NULL; |
9785e10a | 1249 | nl_table[sk->sk_protocol].bind = NULL; |
4f520900 | 1250 | nl_table[sk->sk_protocol].unbind = NULL; |
9785e10a | 1251 | nl_table[sk->sk_protocol].flags = 0; |
869e58f8 DL |
1252 | nl_table[sk->sk_protocol].registered = 0; |
1253 | } | |
658cb354 | 1254 | } else if (nlk->subscriptions) { |
4277a083 | 1255 | netlink_update_listeners(sk); |
658cb354 | 1256 | } |
4277a083 | 1257 | netlink_table_ungrab(); |
77247bbb | 1258 | |
e341694e TG |
1259 | /* Wait for readers to complete */ |
1260 | synchronize_net(); | |
1261 | ||
f7fa9b10 PM |
1262 | kfree(nlk->groups); |
1263 | nlk->groups = NULL; | |
1264 | ||
3755810c | 1265 | local_bh_disable(); |
c1fd3b94 | 1266 | sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1); |
3755810c | 1267 | local_bh_enable(); |
1da177e4 LT |
1268 | sock_put(sk); |
1269 | return 0; | |
1270 | } | |
1271 | ||
1272 | static int netlink_autobind(struct socket *sock) | |
1273 | { | |
1274 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1275 | struct net *net = sock_net(sk); |
da12c90e | 1276 | struct netlink_table *table = &nl_table[sk->sk_protocol]; |
15e47304 | 1277 | s32 portid = task_tgid_vnr(current); |
1da177e4 LT |
1278 | int err; |
1279 | static s32 rover = -4097; | |
1280 | ||
1281 | retry: | |
1282 | cond_resched(); | |
e341694e TG |
1283 | rcu_read_lock(); |
1284 | if (__netlink_lookup(table, portid, net)) { | |
1285 | /* Bind collision, search negative portid values. */ | |
1286 | portid = rover--; | |
1287 | if (rover > -4097) | |
1288 | rover = -4097; | |
1289 | rcu_read_unlock(); | |
1290 | goto retry; | |
1da177e4 | 1291 | } |
e341694e | 1292 | rcu_read_unlock(); |
1da177e4 | 1293 | |
15e47304 | 1294 | err = netlink_insert(sk, net, portid); |
1da177e4 LT |
1295 | if (err == -EADDRINUSE) |
1296 | goto retry; | |
d470e3b4 DM |
1297 | |
1298 | /* If 2 threads race to autobind, that is fine. */ | |
1299 | if (err == -EBUSY) | |
1300 | err = 0; | |
1301 | ||
1302 | return err; | |
1da177e4 LT |
1303 | } |
1304 | ||
aa4cf945 EB |
1305 | /** |
1306 | * __netlink_ns_capable - General netlink message capability test | |
1307 | * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace. | |
1308 | * @user_ns: The user namespace of the capability to use | |
1309 | * @cap: The capability to use | |
1310 | * | |
1311 | * Test to see if the opener of the socket we received the message | |
1312 | * from had when the netlink socket was created and the sender of the | |
1313 | * message has has the capability @cap in the user namespace @user_ns. | |
1314 | */ | |
1315 | bool __netlink_ns_capable(const struct netlink_skb_parms *nsp, | |
1316 | struct user_namespace *user_ns, int cap) | |
1317 | { | |
2d7a85f4 EB |
1318 | return ((nsp->flags & NETLINK_SKB_DST) || |
1319 | file_ns_capable(nsp->sk->sk_socket->file, user_ns, cap)) && | |
1320 | ns_capable(user_ns, cap); | |
aa4cf945 EB |
1321 | } |
1322 | EXPORT_SYMBOL(__netlink_ns_capable); | |
1323 | ||
1324 | /** | |
1325 | * netlink_ns_capable - General netlink message capability test | |
1326 | * @skb: socket buffer holding a netlink command from userspace | |
1327 | * @user_ns: The user namespace of the capability to use | |
1328 | * @cap: The capability to use | |
1329 | * | |
1330 | * Test to see if the opener of the socket we received the message | |
1331 | * from had when the netlink socket was created and the sender of the | |
1332 | * message has has the capability @cap in the user namespace @user_ns. | |
1333 | */ | |
1334 | bool netlink_ns_capable(const struct sk_buff *skb, | |
1335 | struct user_namespace *user_ns, int cap) | |
1336 | { | |
1337 | return __netlink_ns_capable(&NETLINK_CB(skb), user_ns, cap); | |
1338 | } | |
1339 | EXPORT_SYMBOL(netlink_ns_capable); | |
1340 | ||
1341 | /** | |
1342 | * netlink_capable - Netlink global message capability test | |
1343 | * @skb: socket buffer holding a netlink command from userspace | |
1344 | * @cap: The capability to use | |
1345 | * | |
1346 | * Test to see if the opener of the socket we received the message | |
1347 | * from had when the netlink socket was created and the sender of the | |
1348 | * message has has the capability @cap in all user namespaces. | |
1349 | */ | |
1350 | bool netlink_capable(const struct sk_buff *skb, int cap) | |
1351 | { | |
1352 | return netlink_ns_capable(skb, &init_user_ns, cap); | |
1353 | } | |
1354 | EXPORT_SYMBOL(netlink_capable); | |
1355 | ||
1356 | /** | |
1357 | * netlink_net_capable - Netlink network namespace message capability test | |
1358 | * @skb: socket buffer holding a netlink command from userspace | |
1359 | * @cap: The capability to use | |
1360 | * | |
1361 | * Test to see if the opener of the socket we received the message | |
1362 | * from had when the netlink socket was created and the sender of the | |
1363 | * message has has the capability @cap over the network namespace of | |
1364 | * the socket we received the message from. | |
1365 | */ | |
1366 | bool netlink_net_capable(const struct sk_buff *skb, int cap) | |
1367 | { | |
1368 | return netlink_ns_capable(skb, sock_net(skb->sk)->user_ns, cap); | |
1369 | } | |
1370 | EXPORT_SYMBOL(netlink_net_capable); | |
1371 | ||
5187cd05 | 1372 | static inline int netlink_allowed(const struct socket *sock, unsigned int flag) |
746fac4d | 1373 | { |
9785e10a | 1374 | return (nl_table[sock->sk->sk_protocol].flags & flag) || |
df008c91 | 1375 | ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN); |
746fac4d | 1376 | } |
1da177e4 | 1377 | |
f7fa9b10 PM |
1378 | static void |
1379 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
1380 | { | |
1381 | struct netlink_sock *nlk = nlk_sk(sk); | |
1382 | ||
1383 | if (nlk->subscriptions && !subscriptions) | |
1384 | __sk_del_bind_node(sk); | |
1385 | else if (!nlk->subscriptions && subscriptions) | |
1386 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
1387 | nlk->subscriptions = subscriptions; | |
1388 | } | |
1389 | ||
b4ff4f04 | 1390 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
1391 | { |
1392 | struct netlink_sock *nlk = nlk_sk(sk); | |
1393 | unsigned int groups; | |
b4ff4f04 | 1394 | unsigned long *new_groups; |
513c2500 PM |
1395 | int err = 0; |
1396 | ||
b4ff4f04 JB |
1397 | netlink_table_grab(); |
1398 | ||
513c2500 | 1399 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 1400 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 1401 | err = -ENOENT; |
b4ff4f04 JB |
1402 | goto out_unlock; |
1403 | } | |
513c2500 | 1404 | |
b4ff4f04 JB |
1405 | if (nlk->ngroups >= groups) |
1406 | goto out_unlock; | |
513c2500 | 1407 | |
b4ff4f04 JB |
1408 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
1409 | if (new_groups == NULL) { | |
1410 | err = -ENOMEM; | |
1411 | goto out_unlock; | |
1412 | } | |
6ac552fd | 1413 | memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0, |
b4ff4f04 JB |
1414 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); |
1415 | ||
1416 | nlk->groups = new_groups; | |
513c2500 | 1417 | nlk->ngroups = groups; |
b4ff4f04 JB |
1418 | out_unlock: |
1419 | netlink_table_ungrab(); | |
1420 | return err; | |
513c2500 PM |
1421 | } |
1422 | ||
4f520900 RGB |
1423 | static void netlink_unbind(int group, long unsigned int groups, |
1424 | struct netlink_sock *nlk) | |
1425 | { | |
1426 | int undo; | |
1427 | ||
1428 | if (!nlk->netlink_unbind) | |
1429 | return; | |
1430 | ||
1431 | for (undo = 0; undo < group; undo++) | |
1432 | if (test_bit(group, &groups)) | |
1433 | nlk->netlink_unbind(undo); | |
1434 | } | |
1435 | ||
6ac552fd PM |
1436 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, |
1437 | int addr_len) | |
1da177e4 LT |
1438 | { |
1439 | struct sock *sk = sock->sk; | |
3b1e0a65 | 1440 | struct net *net = sock_net(sk); |
1da177e4 LT |
1441 | struct netlink_sock *nlk = nlk_sk(sk); |
1442 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
1443 | int err; | |
4f520900 | 1444 | long unsigned int groups = nladdr->nl_groups; |
746fac4d | 1445 | |
4e4b5376 HFS |
1446 | if (addr_len < sizeof(struct sockaddr_nl)) |
1447 | return -EINVAL; | |
1448 | ||
1da177e4 LT |
1449 | if (nladdr->nl_family != AF_NETLINK) |
1450 | return -EINVAL; | |
1451 | ||
1452 | /* Only superuser is allowed to listen multicasts */ | |
4f520900 | 1453 | if (groups) { |
5187cd05 | 1454 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
513c2500 | 1455 | return -EPERM; |
b4ff4f04 JB |
1456 | err = netlink_realloc_groups(sk); |
1457 | if (err) | |
1458 | return err; | |
513c2500 | 1459 | } |
1da177e4 | 1460 | |
4f520900 | 1461 | if (nlk->portid) |
15e47304 | 1462 | if (nladdr->nl_pid != nlk->portid) |
1da177e4 | 1463 | return -EINVAL; |
4f520900 RGB |
1464 | |
1465 | if (nlk->netlink_bind && groups) { | |
1466 | int group; | |
1467 | ||
1468 | for (group = 0; group < nlk->ngroups; group++) { | |
1469 | if (!test_bit(group, &groups)) | |
1470 | continue; | |
1471 | err = nlk->netlink_bind(group); | |
1472 | if (!err) | |
1473 | continue; | |
1474 | netlink_unbind(group, groups, nlk); | |
1475 | return err; | |
1476 | } | |
1477 | } | |
1478 | ||
1479 | if (!nlk->portid) { | |
1da177e4 | 1480 | err = nladdr->nl_pid ? |
b4b51029 | 1481 | netlink_insert(sk, net, nladdr->nl_pid) : |
1da177e4 | 1482 | netlink_autobind(sock); |
4f520900 RGB |
1483 | if (err) { |
1484 | netlink_unbind(nlk->ngroups - 1, groups, nlk); | |
1da177e4 | 1485 | return err; |
4f520900 | 1486 | } |
1da177e4 LT |
1487 | } |
1488 | ||
4f520900 | 1489 | if (!groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
1490 | return 0; |
1491 | ||
1492 | netlink_table_grab(); | |
f7fa9b10 | 1493 | netlink_update_subscriptions(sk, nlk->subscriptions + |
4f520900 | 1494 | hweight32(groups) - |
746fac4d | 1495 | hweight32(nlk->groups[0])); |
4f520900 | 1496 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | groups; |
4277a083 | 1497 | netlink_update_listeners(sk); |
1da177e4 LT |
1498 | netlink_table_ungrab(); |
1499 | ||
1500 | return 0; | |
1501 | } | |
1502 | ||
1503 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
1504 | int alen, int flags) | |
1505 | { | |
1506 | int err = 0; | |
1507 | struct sock *sk = sock->sk; | |
1508 | struct netlink_sock *nlk = nlk_sk(sk); | |
6ac552fd | 1509 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; |
1da177e4 | 1510 | |
6503d961 CG |
1511 | if (alen < sizeof(addr->sa_family)) |
1512 | return -EINVAL; | |
1513 | ||
1da177e4 LT |
1514 | if (addr->sa_family == AF_UNSPEC) { |
1515 | sk->sk_state = NETLINK_UNCONNECTED; | |
15e47304 | 1516 | nlk->dst_portid = 0; |
d629b836 | 1517 | nlk->dst_group = 0; |
1da177e4 LT |
1518 | return 0; |
1519 | } | |
1520 | if (addr->sa_family != AF_NETLINK) | |
1521 | return -EINVAL; | |
1522 | ||
46833a86 | 1523 | if ((nladdr->nl_groups || nladdr->nl_pid) && |
5187cd05 | 1524 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
1da177e4 LT |
1525 | return -EPERM; |
1526 | ||
15e47304 | 1527 | if (!nlk->portid) |
1da177e4 LT |
1528 | err = netlink_autobind(sock); |
1529 | ||
1530 | if (err == 0) { | |
1531 | sk->sk_state = NETLINK_CONNECTED; | |
15e47304 | 1532 | nlk->dst_portid = nladdr->nl_pid; |
d629b836 | 1533 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
1534 | } |
1535 | ||
1536 | return err; | |
1537 | } | |
1538 | ||
6ac552fd PM |
1539 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, |
1540 | int *addr_len, int peer) | |
1da177e4 LT |
1541 | { |
1542 | struct sock *sk = sock->sk; | |
1543 | struct netlink_sock *nlk = nlk_sk(sk); | |
13cfa97b | 1544 | DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr); |
746fac4d | 1545 | |
1da177e4 LT |
1546 | nladdr->nl_family = AF_NETLINK; |
1547 | nladdr->nl_pad = 0; | |
1548 | *addr_len = sizeof(*nladdr); | |
1549 | ||
1550 | if (peer) { | |
15e47304 | 1551 | nladdr->nl_pid = nlk->dst_portid; |
d629b836 | 1552 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 | 1553 | } else { |
15e47304 | 1554 | nladdr->nl_pid = nlk->portid; |
513c2500 | 1555 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
1556 | } |
1557 | return 0; | |
1558 | } | |
1559 | ||
15e47304 | 1560 | static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid) |
1da177e4 | 1561 | { |
1da177e4 LT |
1562 | struct sock *sock; |
1563 | struct netlink_sock *nlk; | |
1564 | ||
15e47304 | 1565 | sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid); |
1da177e4 LT |
1566 | if (!sock) |
1567 | return ERR_PTR(-ECONNREFUSED); | |
1568 | ||
1569 | /* Don't bother queuing skb if kernel socket has no input function */ | |
1570 | nlk = nlk_sk(sock); | |
cd40b7d3 | 1571 | if (sock->sk_state == NETLINK_CONNECTED && |
15e47304 | 1572 | nlk->dst_portid != nlk_sk(ssk)->portid) { |
1da177e4 LT |
1573 | sock_put(sock); |
1574 | return ERR_PTR(-ECONNREFUSED); | |
1575 | } | |
1576 | return sock; | |
1577 | } | |
1578 | ||
1579 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
1580 | { | |
496ad9aa | 1581 | struct inode *inode = file_inode(filp); |
1da177e4 LT |
1582 | struct sock *sock; |
1583 | ||
1584 | if (!S_ISSOCK(inode->i_mode)) | |
1585 | return ERR_PTR(-ENOTSOCK); | |
1586 | ||
1587 | sock = SOCKET_I(inode)->sk; | |
1588 | if (sock->sk_family != AF_NETLINK) | |
1589 | return ERR_PTR(-EINVAL); | |
1590 | ||
1591 | sock_hold(sock); | |
1592 | return sock; | |
1593 | } | |
1594 | ||
3a36515f PN |
1595 | static struct sk_buff *netlink_alloc_large_skb(unsigned int size, |
1596 | int broadcast) | |
c05cdb1b PNA |
1597 | { |
1598 | struct sk_buff *skb; | |
1599 | void *data; | |
1600 | ||
3a36515f | 1601 | if (size <= NLMSG_GOODSIZE || broadcast) |
c05cdb1b PNA |
1602 | return alloc_skb(size, GFP_KERNEL); |
1603 | ||
3a36515f PN |
1604 | size = SKB_DATA_ALIGN(size) + |
1605 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); | |
c05cdb1b PNA |
1606 | |
1607 | data = vmalloc(size); | |
1608 | if (data == NULL) | |
3a36515f | 1609 | return NULL; |
c05cdb1b | 1610 | |
3a36515f PN |
1611 | skb = build_skb(data, size); |
1612 | if (skb == NULL) | |
1613 | vfree(data); | |
1614 | else { | |
1615 | skb->head_frag = 0; | |
1616 | skb->destructor = netlink_skb_destructor; | |
1617 | } | |
c05cdb1b PNA |
1618 | |
1619 | return skb; | |
c05cdb1b PNA |
1620 | } |
1621 | ||
1da177e4 LT |
1622 | /* |
1623 | * Attach a skb to a netlink socket. | |
1624 | * The caller must hold a reference to the destination socket. On error, the | |
1625 | * reference is dropped. The skb is not send to the destination, just all | |
1626 | * all error checks are performed and memory in the queue is reserved. | |
1627 | * Return values: | |
1628 | * < 0: error. skb freed, reference to sock dropped. | |
1629 | * 0: continue | |
1630 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
1631 | */ | |
9457afee | 1632 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, |
c3d8d1e3 | 1633 | long *timeo, struct sock *ssk) |
1da177e4 LT |
1634 | { |
1635 | struct netlink_sock *nlk; | |
1636 | ||
1637 | nlk = nlk_sk(sk); | |
1638 | ||
5fd96123 PM |
1639 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || |
1640 | test_bit(NETLINK_CONGESTED, &nlk->state)) && | |
1641 | !netlink_skb_is_mmaped(skb)) { | |
1da177e4 | 1642 | DECLARE_WAITQUEUE(wait, current); |
c3d8d1e3 | 1643 | if (!*timeo) { |
aed81560 | 1644 | if (!ssk || netlink_is_kernel(ssk)) |
1da177e4 LT |
1645 | netlink_overrun(sk); |
1646 | sock_put(sk); | |
1647 | kfree_skb(skb); | |
1648 | return -EAGAIN; | |
1649 | } | |
1650 | ||
1651 | __set_current_state(TASK_INTERRUPTIBLE); | |
1652 | add_wait_queue(&nlk->wait, &wait); | |
1653 | ||
1654 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
cd967e05 | 1655 | test_bit(NETLINK_CONGESTED, &nlk->state)) && |
1da177e4 | 1656 | !sock_flag(sk, SOCK_DEAD)) |
c3d8d1e3 | 1657 | *timeo = schedule_timeout(*timeo); |
1da177e4 LT |
1658 | |
1659 | __set_current_state(TASK_RUNNING); | |
1660 | remove_wait_queue(&nlk->wait, &wait); | |
1661 | sock_put(sk); | |
1662 | ||
1663 | if (signal_pending(current)) { | |
1664 | kfree_skb(skb); | |
c3d8d1e3 | 1665 | return sock_intr_errno(*timeo); |
1da177e4 LT |
1666 | } |
1667 | return 1; | |
1668 | } | |
cf0a018a | 1669 | netlink_skb_set_owner_r(skb, sk); |
1da177e4 LT |
1670 | return 0; |
1671 | } | |
1672 | ||
4a7e7c2a | 1673 | static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 1674 | { |
1da177e4 LT |
1675 | int len = skb->len; |
1676 | ||
bcbde0d4 DB |
1677 | netlink_deliver_tap(skb); |
1678 | ||
f9c22888 PM |
1679 | #ifdef CONFIG_NETLINK_MMAP |
1680 | if (netlink_skb_is_mmaped(skb)) | |
1681 | netlink_queue_mmaped_skb(sk, skb); | |
1682 | else if (netlink_rx_is_mmaped(sk)) | |
1683 | netlink_ring_set_copied(sk, skb); | |
1684 | else | |
1685 | #endif /* CONFIG_NETLINK_MMAP */ | |
1686 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
676d2369 | 1687 | sk->sk_data_ready(sk); |
4a7e7c2a ED |
1688 | return len; |
1689 | } | |
1690 | ||
1691 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb) | |
1692 | { | |
1693 | int len = __netlink_sendskb(sk, skb); | |
1694 | ||
1da177e4 LT |
1695 | sock_put(sk); |
1696 | return len; | |
1697 | } | |
1698 | ||
1699 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
1700 | { | |
1701 | kfree_skb(skb); | |
1702 | sock_put(sk); | |
1703 | } | |
1704 | ||
b57ef81f | 1705 | static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation) |
1da177e4 LT |
1706 | { |
1707 | int delta; | |
1708 | ||
1298ca46 | 1709 | WARN_ON(skb->sk != NULL); |
5fd96123 PM |
1710 | if (netlink_skb_is_mmaped(skb)) |
1711 | return skb; | |
1da177e4 | 1712 | |
4305b541 | 1713 | delta = skb->end - skb->tail; |
c05cdb1b | 1714 | if (is_vmalloc_addr(skb->head) || delta * 2 < skb->truesize) |
1da177e4 LT |
1715 | return skb; |
1716 | ||
1717 | if (skb_shared(skb)) { | |
1718 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
1719 | if (!nskb) | |
1720 | return skb; | |
8460c00f | 1721 | consume_skb(skb); |
1da177e4 LT |
1722 | skb = nskb; |
1723 | } | |
1724 | ||
1725 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
1726 | skb->truesize -= delta; | |
1727 | ||
1728 | return skb; | |
1729 | } | |
1730 | ||
3fbc2905 EB |
1731 | static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, |
1732 | struct sock *ssk) | |
cd40b7d3 DL |
1733 | { |
1734 | int ret; | |
1735 | struct netlink_sock *nlk = nlk_sk(sk); | |
1736 | ||
1737 | ret = -ECONNREFUSED; | |
1738 | if (nlk->netlink_rcv != NULL) { | |
1739 | ret = skb->len; | |
cf0a018a | 1740 | netlink_skb_set_owner_r(skb, sk); |
e32123e5 | 1741 | NETLINK_CB(skb).sk = ssk; |
73bfd370 | 1742 | netlink_deliver_tap_kernel(sk, ssk, skb); |
cd40b7d3 | 1743 | nlk->netlink_rcv(skb); |
bfb253c9 ED |
1744 | consume_skb(skb); |
1745 | } else { | |
1746 | kfree_skb(skb); | |
cd40b7d3 | 1747 | } |
cd40b7d3 DL |
1748 | sock_put(sk); |
1749 | return ret; | |
1750 | } | |
1751 | ||
1752 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, | |
15e47304 | 1753 | u32 portid, int nonblock) |
1da177e4 LT |
1754 | { |
1755 | struct sock *sk; | |
1756 | int err; | |
1757 | long timeo; | |
1758 | ||
1759 | skb = netlink_trim(skb, gfp_any()); | |
1760 | ||
1761 | timeo = sock_sndtimeo(ssk, nonblock); | |
1762 | retry: | |
15e47304 | 1763 | sk = netlink_getsockbyportid(ssk, portid); |
1da177e4 LT |
1764 | if (IS_ERR(sk)) { |
1765 | kfree_skb(skb); | |
1766 | return PTR_ERR(sk); | |
1767 | } | |
cd40b7d3 | 1768 | if (netlink_is_kernel(sk)) |
3fbc2905 | 1769 | return netlink_unicast_kernel(sk, skb, ssk); |
cd40b7d3 | 1770 | |
b1153f29 | 1771 | if (sk_filter(sk, skb)) { |
84874607 | 1772 | err = skb->len; |
b1153f29 SH |
1773 | kfree_skb(skb); |
1774 | sock_put(sk); | |
1775 | return err; | |
1776 | } | |
1777 | ||
9457afee | 1778 | err = netlink_attachskb(sk, skb, &timeo, ssk); |
1da177e4 LT |
1779 | if (err == 1) |
1780 | goto retry; | |
1781 | if (err) | |
1782 | return err; | |
1783 | ||
7ee015e0 | 1784 | return netlink_sendskb(sk, skb); |
1da177e4 | 1785 | } |
6ac552fd | 1786 | EXPORT_SYMBOL(netlink_unicast); |
1da177e4 | 1787 | |
f9c22888 PM |
1788 | struct sk_buff *netlink_alloc_skb(struct sock *ssk, unsigned int size, |
1789 | u32 dst_portid, gfp_t gfp_mask) | |
1790 | { | |
1791 | #ifdef CONFIG_NETLINK_MMAP | |
1792 | struct sock *sk = NULL; | |
1793 | struct sk_buff *skb; | |
1794 | struct netlink_ring *ring; | |
1795 | struct nl_mmap_hdr *hdr; | |
1796 | unsigned int maxlen; | |
1797 | ||
1798 | sk = netlink_getsockbyportid(ssk, dst_portid); | |
1799 | if (IS_ERR(sk)) | |
1800 | goto out; | |
1801 | ||
1802 | ring = &nlk_sk(sk)->rx_ring; | |
1803 | /* fast-path without atomic ops for common case: non-mmaped receiver */ | |
1804 | if (ring->pg_vec == NULL) | |
1805 | goto out_put; | |
1806 | ||
aae9f0e2 TG |
1807 | if (ring->frame_size - NL_MMAP_HDRLEN < size) |
1808 | goto out_put; | |
1809 | ||
f9c22888 PM |
1810 | skb = alloc_skb_head(gfp_mask); |
1811 | if (skb == NULL) | |
1812 | goto err1; | |
1813 | ||
1814 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
1815 | /* check again under lock */ | |
1816 | if (ring->pg_vec == NULL) | |
1817 | goto out_free; | |
1818 | ||
aae9f0e2 | 1819 | /* check again under lock */ |
f9c22888 PM |
1820 | maxlen = ring->frame_size - NL_MMAP_HDRLEN; |
1821 | if (maxlen < size) | |
1822 | goto out_free; | |
1823 | ||
1824 | netlink_forward_ring(ring); | |
1825 | hdr = netlink_current_frame(ring, NL_MMAP_STATUS_UNUSED); | |
1826 | if (hdr == NULL) | |
1827 | goto err2; | |
1828 | netlink_ring_setup_skb(skb, sk, ring, hdr); | |
1829 | netlink_set_status(hdr, NL_MMAP_STATUS_RESERVED); | |
1830 | atomic_inc(&ring->pending); | |
1831 | netlink_increment_head(ring); | |
1832 | ||
1833 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1834 | return skb; | |
1835 | ||
1836 | err2: | |
1837 | kfree_skb(skb); | |
1838 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
cd1df525 | 1839 | netlink_overrun(sk); |
f9c22888 PM |
1840 | err1: |
1841 | sock_put(sk); | |
1842 | return NULL; | |
1843 | ||
1844 | out_free: | |
1845 | kfree_skb(skb); | |
1846 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
1847 | out_put: | |
1848 | sock_put(sk); | |
1849 | out: | |
1850 | #endif | |
1851 | return alloc_skb(size, gfp_mask); | |
1852 | } | |
1853 | EXPORT_SYMBOL_GPL(netlink_alloc_skb); | |
1854 | ||
4277a083 PM |
1855 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
1856 | { | |
1857 | int res = 0; | |
5c398dc8 | 1858 | struct listeners *listeners; |
4277a083 | 1859 | |
aed81560 | 1860 | BUG_ON(!netlink_is_kernel(sk)); |
b4ff4f04 JB |
1861 | |
1862 | rcu_read_lock(); | |
1863 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
1864 | ||
6d772ac5 | 1865 | if (listeners && group - 1 < nl_table[sk->sk_protocol].groups) |
5c398dc8 | 1866 | res = test_bit(group - 1, listeners->masks); |
b4ff4f04 JB |
1867 | |
1868 | rcu_read_unlock(); | |
1869 | ||
4277a083 PM |
1870 | return res; |
1871 | } | |
1872 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
1873 | ||
b57ef81f | 1874 | static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
1875 | { |
1876 | struct netlink_sock *nlk = nlk_sk(sk); | |
1877 | ||
1878 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
cd967e05 | 1879 | !test_bit(NETLINK_CONGESTED, &nlk->state)) { |
cf0a018a | 1880 | netlink_skb_set_owner_r(skb, sk); |
4a7e7c2a | 1881 | __netlink_sendskb(sk, skb); |
2c645800 | 1882 | return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); |
1da177e4 LT |
1883 | } |
1884 | return -1; | |
1885 | } | |
1886 | ||
1887 | struct netlink_broadcast_data { | |
1888 | struct sock *exclude_sk; | |
b4b51029 | 1889 | struct net *net; |
15e47304 | 1890 | u32 portid; |
1da177e4 LT |
1891 | u32 group; |
1892 | int failure; | |
ff491a73 | 1893 | int delivery_failure; |
1da177e4 LT |
1894 | int congested; |
1895 | int delivered; | |
7d877f3b | 1896 | gfp_t allocation; |
1da177e4 | 1897 | struct sk_buff *skb, *skb2; |
910a7e90 EB |
1898 | int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data); |
1899 | void *tx_data; | |
1da177e4 LT |
1900 | }; |
1901 | ||
46c9521f RR |
1902 | static void do_one_broadcast(struct sock *sk, |
1903 | struct netlink_broadcast_data *p) | |
1da177e4 LT |
1904 | { |
1905 | struct netlink_sock *nlk = nlk_sk(sk); | |
1906 | int val; | |
1907 | ||
1908 | if (p->exclude_sk == sk) | |
46c9521f | 1909 | return; |
1da177e4 | 1910 | |
15e47304 | 1911 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 1912 | !test_bit(p->group - 1, nlk->groups)) |
46c9521f | 1913 | return; |
1da177e4 | 1914 | |
878628fb | 1915 | if (!net_eq(sock_net(sk), p->net)) |
46c9521f | 1916 | return; |
b4b51029 | 1917 | |
1da177e4 LT |
1918 | if (p->failure) { |
1919 | netlink_overrun(sk); | |
46c9521f | 1920 | return; |
1da177e4 LT |
1921 | } |
1922 | ||
1923 | sock_hold(sk); | |
1924 | if (p->skb2 == NULL) { | |
68acc024 | 1925 | if (skb_shared(p->skb)) { |
1da177e4 LT |
1926 | p->skb2 = skb_clone(p->skb, p->allocation); |
1927 | } else { | |
68acc024 TC |
1928 | p->skb2 = skb_get(p->skb); |
1929 | /* | |
1930 | * skb ownership may have been set when | |
1931 | * delivered to a previous socket. | |
1932 | */ | |
1933 | skb_orphan(p->skb2); | |
1da177e4 LT |
1934 | } |
1935 | } | |
1936 | if (p->skb2 == NULL) { | |
1937 | netlink_overrun(sk); | |
1938 | /* Clone failed. Notify ALL listeners. */ | |
1939 | p->failure = 1; | |
be0c22a4 PNA |
1940 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1941 | p->delivery_failure = 1; | |
910a7e90 EB |
1942 | } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) { |
1943 | kfree_skb(p->skb2); | |
1944 | p->skb2 = NULL; | |
b1153f29 SH |
1945 | } else if (sk_filter(sk, p->skb2)) { |
1946 | kfree_skb(p->skb2); | |
1947 | p->skb2 = NULL; | |
1da177e4 LT |
1948 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { |
1949 | netlink_overrun(sk); | |
be0c22a4 PNA |
1950 | if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR) |
1951 | p->delivery_failure = 1; | |
1da177e4 LT |
1952 | } else { |
1953 | p->congested |= val; | |
1954 | p->delivered = 1; | |
1955 | p->skb2 = NULL; | |
1956 | } | |
1957 | sock_put(sk); | |
1da177e4 LT |
1958 | } |
1959 | ||
15e47304 | 1960 | int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
1961 | u32 group, gfp_t allocation, |
1962 | int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data), | |
1963 | void *filter_data) | |
1da177e4 | 1964 | { |
3b1e0a65 | 1965 | struct net *net = sock_net(ssk); |
1da177e4 | 1966 | struct netlink_broadcast_data info; |
1da177e4 LT |
1967 | struct sock *sk; |
1968 | ||
1969 | skb = netlink_trim(skb, allocation); | |
1970 | ||
1971 | info.exclude_sk = ssk; | |
b4b51029 | 1972 | info.net = net; |
15e47304 | 1973 | info.portid = portid; |
1da177e4 LT |
1974 | info.group = group; |
1975 | info.failure = 0; | |
ff491a73 | 1976 | info.delivery_failure = 0; |
1da177e4 LT |
1977 | info.congested = 0; |
1978 | info.delivered = 0; | |
1979 | info.allocation = allocation; | |
1980 | info.skb = skb; | |
1981 | info.skb2 = NULL; | |
910a7e90 EB |
1982 | info.tx_filter = filter; |
1983 | info.tx_data = filter_data; | |
1da177e4 LT |
1984 | |
1985 | /* While we sleep in clone, do not allow to change socket list */ | |
1986 | ||
1987 | netlink_lock_table(); | |
1988 | ||
b67bfe0d | 1989 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1da177e4 LT |
1990 | do_one_broadcast(sk, &info); |
1991 | ||
70d4bf6d | 1992 | consume_skb(skb); |
aa1c6a6f | 1993 | |
1da177e4 LT |
1994 | netlink_unlock_table(); |
1995 | ||
70d4bf6d NH |
1996 | if (info.delivery_failure) { |
1997 | kfree_skb(info.skb2); | |
ff491a73 | 1998 | return -ENOBUFS; |
658cb354 ED |
1999 | } |
2000 | consume_skb(info.skb2); | |
ff491a73 | 2001 | |
1da177e4 LT |
2002 | if (info.delivered) { |
2003 | if (info.congested && (allocation & __GFP_WAIT)) | |
2004 | yield(); | |
2005 | return 0; | |
2006 | } | |
1da177e4 LT |
2007 | return -ESRCH; |
2008 | } | |
910a7e90 EB |
2009 | EXPORT_SYMBOL(netlink_broadcast_filtered); |
2010 | ||
15e47304 | 2011 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid, |
910a7e90 EB |
2012 | u32 group, gfp_t allocation) |
2013 | { | |
15e47304 | 2014 | return netlink_broadcast_filtered(ssk, skb, portid, group, allocation, |
910a7e90 EB |
2015 | NULL, NULL); |
2016 | } | |
6ac552fd | 2017 | EXPORT_SYMBOL(netlink_broadcast); |
1da177e4 LT |
2018 | |
2019 | struct netlink_set_err_data { | |
2020 | struct sock *exclude_sk; | |
15e47304 | 2021 | u32 portid; |
1da177e4 LT |
2022 | u32 group; |
2023 | int code; | |
2024 | }; | |
2025 | ||
b57ef81f | 2026 | static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p) |
1da177e4 LT |
2027 | { |
2028 | struct netlink_sock *nlk = nlk_sk(sk); | |
1a50307b | 2029 | int ret = 0; |
1da177e4 LT |
2030 | |
2031 | if (sk == p->exclude_sk) | |
2032 | goto out; | |
2033 | ||
09ad9bc7 | 2034 | if (!net_eq(sock_net(sk), sock_net(p->exclude_sk))) |
b4b51029 EB |
2035 | goto out; |
2036 | ||
15e47304 | 2037 | if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups || |
f7fa9b10 | 2038 | !test_bit(p->group - 1, nlk->groups)) |
1da177e4 LT |
2039 | goto out; |
2040 | ||
1a50307b PNA |
2041 | if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) { |
2042 | ret = 1; | |
2043 | goto out; | |
2044 | } | |
2045 | ||
1da177e4 LT |
2046 | sk->sk_err = p->code; |
2047 | sk->sk_error_report(sk); | |
2048 | out: | |
1a50307b | 2049 | return ret; |
1da177e4 LT |
2050 | } |
2051 | ||
4843b93c PNA |
2052 | /** |
2053 | * netlink_set_err - report error to broadcast listeners | |
2054 | * @ssk: the kernel netlink socket, as returned by netlink_kernel_create() | |
15e47304 | 2055 | * @portid: the PORTID of a process that we want to skip (if any) |
840e93f2 | 2056 | * @group: the broadcast group that will notice the error |
4843b93c | 2057 | * @code: error code, must be negative (as usual in kernelspace) |
1a50307b PNA |
2058 | * |
2059 | * This function returns the number of broadcast listeners that have set the | |
2060 | * NETLINK_RECV_NO_ENOBUFS socket option. | |
4843b93c | 2061 | */ |
15e47304 | 2062 | int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code) |
1da177e4 LT |
2063 | { |
2064 | struct netlink_set_err_data info; | |
1da177e4 | 2065 | struct sock *sk; |
1a50307b | 2066 | int ret = 0; |
1da177e4 LT |
2067 | |
2068 | info.exclude_sk = ssk; | |
15e47304 | 2069 | info.portid = portid; |
1da177e4 | 2070 | info.group = group; |
4843b93c PNA |
2071 | /* sk->sk_err wants a positive error value */ |
2072 | info.code = -code; | |
1da177e4 LT |
2073 | |
2074 | read_lock(&nl_table_lock); | |
2075 | ||
b67bfe0d | 2076 | sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list) |
1a50307b | 2077 | ret += do_one_set_err(sk, &info); |
1da177e4 LT |
2078 | |
2079 | read_unlock(&nl_table_lock); | |
1a50307b | 2080 | return ret; |
1da177e4 | 2081 | } |
dd5b6ce6 | 2082 | EXPORT_SYMBOL(netlink_set_err); |
1da177e4 | 2083 | |
84659eb5 JB |
2084 | /* must be called with netlink table grabbed */ |
2085 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
2086 | unsigned int group, | |
2087 | int is_new) | |
2088 | { | |
2089 | int old, new = !!is_new, subscriptions; | |
2090 | ||
2091 | old = test_bit(group - 1, nlk->groups); | |
2092 | subscriptions = nlk->subscriptions - old + new; | |
2093 | if (new) | |
2094 | __set_bit(group - 1, nlk->groups); | |
2095 | else | |
2096 | __clear_bit(group - 1, nlk->groups); | |
2097 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
2098 | netlink_update_listeners(&nlk->sk); | |
2099 | } | |
2100 | ||
9a4595bc | 2101 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 2102 | char __user *optval, unsigned int optlen) |
9a4595bc PM |
2103 | { |
2104 | struct sock *sk = sock->sk; | |
2105 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
2106 | unsigned int val = 0; |
2107 | int err; | |
9a4595bc PM |
2108 | |
2109 | if (level != SOL_NETLINK) | |
2110 | return -ENOPROTOOPT; | |
2111 | ||
ccdfcc39 PM |
2112 | if (optname != NETLINK_RX_RING && optname != NETLINK_TX_RING && |
2113 | optlen >= sizeof(int) && | |
eb496534 | 2114 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
2115 | return -EFAULT; |
2116 | ||
2117 | switch (optname) { | |
2118 | case NETLINK_PKTINFO: | |
2119 | if (val) | |
2120 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
2121 | else | |
2122 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
2123 | err = 0; | |
2124 | break; | |
2125 | case NETLINK_ADD_MEMBERSHIP: | |
2126 | case NETLINK_DROP_MEMBERSHIP: { | |
5187cd05 | 2127 | if (!netlink_allowed(sock, NL_CFG_F_NONROOT_RECV)) |
9a4595bc | 2128 | return -EPERM; |
b4ff4f04 JB |
2129 | err = netlink_realloc_groups(sk); |
2130 | if (err) | |
2131 | return err; | |
9a4595bc PM |
2132 | if (!val || val - 1 >= nlk->ngroups) |
2133 | return -EINVAL; | |
7774d5e0 | 2134 | if (optname == NETLINK_ADD_MEMBERSHIP && nlk->netlink_bind) { |
4f520900 RGB |
2135 | err = nlk->netlink_bind(val); |
2136 | if (err) | |
2137 | return err; | |
2138 | } | |
9a4595bc | 2139 | netlink_table_grab(); |
84659eb5 JB |
2140 | netlink_update_socket_mc(nlk, val, |
2141 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc | 2142 | netlink_table_ungrab(); |
7774d5e0 RGB |
2143 | if (optname == NETLINK_DROP_MEMBERSHIP && nlk->netlink_unbind) |
2144 | nlk->netlink_unbind(val); | |
03292745 | 2145 | |
9a4595bc PM |
2146 | err = 0; |
2147 | break; | |
2148 | } | |
be0c22a4 PNA |
2149 | case NETLINK_BROADCAST_ERROR: |
2150 | if (val) | |
2151 | nlk->flags |= NETLINK_BROADCAST_SEND_ERROR; | |
2152 | else | |
2153 | nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR; | |
2154 | err = 0; | |
2155 | break; | |
38938bfe PNA |
2156 | case NETLINK_NO_ENOBUFS: |
2157 | if (val) { | |
2158 | nlk->flags |= NETLINK_RECV_NO_ENOBUFS; | |
cd967e05 | 2159 | clear_bit(NETLINK_CONGESTED, &nlk->state); |
38938bfe | 2160 | wake_up_interruptible(&nlk->wait); |
658cb354 | 2161 | } else { |
38938bfe | 2162 | nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS; |
658cb354 | 2163 | } |
38938bfe PNA |
2164 | err = 0; |
2165 | break; | |
ccdfcc39 PM |
2166 | #ifdef CONFIG_NETLINK_MMAP |
2167 | case NETLINK_RX_RING: | |
2168 | case NETLINK_TX_RING: { | |
2169 | struct nl_mmap_req req; | |
2170 | ||
2171 | /* Rings might consume more memory than queue limits, require | |
2172 | * CAP_NET_ADMIN. | |
2173 | */ | |
2174 | if (!capable(CAP_NET_ADMIN)) | |
2175 | return -EPERM; | |
2176 | if (optlen < sizeof(req)) | |
2177 | return -EINVAL; | |
2178 | if (copy_from_user(&req, optval, sizeof(req))) | |
2179 | return -EFAULT; | |
2180 | err = netlink_set_ring(sk, &req, false, | |
2181 | optname == NETLINK_TX_RING); | |
2182 | break; | |
2183 | } | |
2184 | #endif /* CONFIG_NETLINK_MMAP */ | |
9a4595bc PM |
2185 | default: |
2186 | err = -ENOPROTOOPT; | |
2187 | } | |
2188 | return err; | |
2189 | } | |
2190 | ||
2191 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 2192 | char __user *optval, int __user *optlen) |
9a4595bc PM |
2193 | { |
2194 | struct sock *sk = sock->sk; | |
2195 | struct netlink_sock *nlk = nlk_sk(sk); | |
2196 | int len, val, err; | |
2197 | ||
2198 | if (level != SOL_NETLINK) | |
2199 | return -ENOPROTOOPT; | |
2200 | ||
2201 | if (get_user(len, optlen)) | |
2202 | return -EFAULT; | |
2203 | if (len < 0) | |
2204 | return -EINVAL; | |
2205 | ||
2206 | switch (optname) { | |
2207 | case NETLINK_PKTINFO: | |
2208 | if (len < sizeof(int)) | |
2209 | return -EINVAL; | |
2210 | len = sizeof(int); | |
2211 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
2212 | if (put_user(len, optlen) || |
2213 | put_user(val, optval)) | |
2214 | return -EFAULT; | |
9a4595bc PM |
2215 | err = 0; |
2216 | break; | |
be0c22a4 PNA |
2217 | case NETLINK_BROADCAST_ERROR: |
2218 | if (len < sizeof(int)) | |
2219 | return -EINVAL; | |
2220 | len = sizeof(int); | |
2221 | val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0; | |
2222 | if (put_user(len, optlen) || | |
2223 | put_user(val, optval)) | |
2224 | return -EFAULT; | |
2225 | err = 0; | |
2226 | break; | |
38938bfe PNA |
2227 | case NETLINK_NO_ENOBUFS: |
2228 | if (len < sizeof(int)) | |
2229 | return -EINVAL; | |
2230 | len = sizeof(int); | |
2231 | val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0; | |
2232 | if (put_user(len, optlen) || | |
2233 | put_user(val, optval)) | |
2234 | return -EFAULT; | |
2235 | err = 0; | |
2236 | break; | |
9a4595bc PM |
2237 | default: |
2238 | err = -ENOPROTOOPT; | |
2239 | } | |
2240 | return err; | |
2241 | } | |
2242 | ||
2243 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
2244 | { | |
2245 | struct nl_pktinfo info; | |
2246 | ||
2247 | info.group = NETLINK_CB(skb).dst_group; | |
2248 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
2249 | } | |
2250 | ||
1da177e4 LT |
2251 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, |
2252 | struct msghdr *msg, size_t len) | |
2253 | { | |
2254 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2255 | struct sock *sk = sock->sk; | |
2256 | struct netlink_sock *nlk = nlk_sk(sk); | |
342dfc30 | 2257 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
15e47304 | 2258 | u32 dst_portid; |
d629b836 | 2259 | u32 dst_group; |
1da177e4 LT |
2260 | struct sk_buff *skb; |
2261 | int err; | |
2262 | struct scm_cookie scm; | |
2d7a85f4 | 2263 | u32 netlink_skb_flags = 0; |
1da177e4 LT |
2264 | |
2265 | if (msg->msg_flags&MSG_OOB) | |
2266 | return -EOPNOTSUPP; | |
2267 | ||
16e57262 | 2268 | if (NULL == siocb->scm) |
1da177e4 | 2269 | siocb->scm = &scm; |
16e57262 | 2270 | |
e0e3cea4 | 2271 | err = scm_send(sock, msg, siocb->scm, true); |
1da177e4 LT |
2272 | if (err < 0) |
2273 | return err; | |
2274 | ||
2275 | if (msg->msg_namelen) { | |
b47030c7 | 2276 | err = -EINVAL; |
1da177e4 | 2277 | if (addr->nl_family != AF_NETLINK) |
b47030c7 | 2278 | goto out; |
15e47304 | 2279 | dst_portid = addr->nl_pid; |
d629b836 | 2280 | dst_group = ffs(addr->nl_groups); |
b47030c7 | 2281 | err = -EPERM; |
15e47304 | 2282 | if ((dst_group || dst_portid) && |
5187cd05 | 2283 | !netlink_allowed(sock, NL_CFG_F_NONROOT_SEND)) |
b47030c7 | 2284 | goto out; |
2d7a85f4 | 2285 | netlink_skb_flags |= NETLINK_SKB_DST; |
1da177e4 | 2286 | } else { |
15e47304 | 2287 | dst_portid = nlk->dst_portid; |
d629b836 | 2288 | dst_group = nlk->dst_group; |
1da177e4 LT |
2289 | } |
2290 | ||
15e47304 | 2291 | if (!nlk->portid) { |
1da177e4 LT |
2292 | err = netlink_autobind(sock); |
2293 | if (err) | |
2294 | goto out; | |
2295 | } | |
2296 | ||
5fd96123 PM |
2297 | if (netlink_tx_is_mmaped(sk) && |
2298 | msg->msg_iov->iov_base == NULL) { | |
2299 | err = netlink_mmap_sendmsg(sk, msg, dst_portid, dst_group, | |
2300 | siocb); | |
2301 | goto out; | |
2302 | } | |
2303 | ||
1da177e4 LT |
2304 | err = -EMSGSIZE; |
2305 | if (len > sk->sk_sndbuf - 32) | |
2306 | goto out; | |
2307 | err = -ENOBUFS; | |
3a36515f | 2308 | skb = netlink_alloc_large_skb(len, dst_group); |
6ac552fd | 2309 | if (skb == NULL) |
1da177e4 LT |
2310 | goto out; |
2311 | ||
15e47304 | 2312 | NETLINK_CB(skb).portid = nlk->portid; |
d629b836 | 2313 | NETLINK_CB(skb).dst_group = dst_group; |
dbe9a417 | 2314 | NETLINK_CB(skb).creds = siocb->scm->creds; |
2d7a85f4 | 2315 | NETLINK_CB(skb).flags = netlink_skb_flags; |
1da177e4 | 2316 | |
1da177e4 | 2317 | err = -EFAULT; |
6ac552fd | 2318 | if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) { |
1da177e4 LT |
2319 | kfree_skb(skb); |
2320 | goto out; | |
2321 | } | |
2322 | ||
2323 | err = security_netlink_send(sk, skb); | |
2324 | if (err) { | |
2325 | kfree_skb(skb); | |
2326 | goto out; | |
2327 | } | |
2328 | ||
d629b836 | 2329 | if (dst_group) { |
1da177e4 | 2330 | atomic_inc(&skb->users); |
15e47304 | 2331 | netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL); |
1da177e4 | 2332 | } |
15e47304 | 2333 | err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT); |
1da177e4 LT |
2334 | |
2335 | out: | |
b47030c7 | 2336 | scm_destroy(siocb->scm); |
1da177e4 LT |
2337 | return err; |
2338 | } | |
2339 | ||
2340 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
2341 | struct msghdr *msg, size_t len, | |
2342 | int flags) | |
2343 | { | |
2344 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
2345 | struct scm_cookie scm; | |
2346 | struct sock *sk = sock->sk; | |
2347 | struct netlink_sock *nlk = nlk_sk(sk); | |
2348 | int noblock = flags&MSG_DONTWAIT; | |
2349 | size_t copied; | |
68d6ac6d | 2350 | struct sk_buff *skb, *data_skb; |
b44d211e | 2351 | int err, ret; |
1da177e4 LT |
2352 | |
2353 | if (flags&MSG_OOB) | |
2354 | return -EOPNOTSUPP; | |
2355 | ||
2356 | copied = 0; | |
2357 | ||
6ac552fd PM |
2358 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
2359 | if (skb == NULL) | |
1da177e4 LT |
2360 | goto out; |
2361 | ||
68d6ac6d JB |
2362 | data_skb = skb; |
2363 | ||
1dacc76d JB |
2364 | #ifdef CONFIG_COMPAT_NETLINK_MESSAGES |
2365 | if (unlikely(skb_shinfo(skb)->frag_list)) { | |
1dacc76d | 2366 | /* |
68d6ac6d JB |
2367 | * If this skb has a frag_list, then here that means that we |
2368 | * will have to use the frag_list skb's data for compat tasks | |
2369 | * and the regular skb's data for normal (non-compat) tasks. | |
1dacc76d | 2370 | * |
68d6ac6d JB |
2371 | * If we need to send the compat skb, assign it to the |
2372 | * 'data_skb' variable so that it will be used below for data | |
2373 | * copying. We keep 'skb' for everything else, including | |
2374 | * freeing both later. | |
1dacc76d | 2375 | */ |
68d6ac6d JB |
2376 | if (flags & MSG_CMSG_COMPAT) |
2377 | data_skb = skb_shinfo(skb)->frag_list; | |
1dacc76d JB |
2378 | } |
2379 | #endif | |
2380 | ||
9063e21f ED |
2381 | /* Record the max length of recvmsg() calls for future allocations */ |
2382 | nlk->max_recvmsg_len = max(nlk->max_recvmsg_len, len); | |
2383 | nlk->max_recvmsg_len = min_t(size_t, nlk->max_recvmsg_len, | |
2384 | 16384); | |
2385 | ||
68d6ac6d | 2386 | copied = data_skb->len; |
1da177e4 LT |
2387 | if (len < copied) { |
2388 | msg->msg_flags |= MSG_TRUNC; | |
2389 | copied = len; | |
2390 | } | |
2391 | ||
68d6ac6d JB |
2392 | skb_reset_transport_header(data_skb); |
2393 | err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied); | |
1da177e4 LT |
2394 | |
2395 | if (msg->msg_name) { | |
342dfc30 | 2396 | DECLARE_SOCKADDR(struct sockaddr_nl *, addr, msg->msg_name); |
1da177e4 LT |
2397 | addr->nl_family = AF_NETLINK; |
2398 | addr->nl_pad = 0; | |
15e47304 | 2399 | addr->nl_pid = NETLINK_CB(skb).portid; |
d629b836 | 2400 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
2401 | msg->msg_namelen = sizeof(*addr); |
2402 | } | |
2403 | ||
cc9a06cd PM |
2404 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
2405 | netlink_cmsg_recv_pktinfo(msg, skb); | |
2406 | ||
1da177e4 LT |
2407 | if (NULL == siocb->scm) { |
2408 | memset(&scm, 0, sizeof(scm)); | |
2409 | siocb->scm = &scm; | |
2410 | } | |
2411 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 | 2412 | if (flags & MSG_TRUNC) |
68d6ac6d | 2413 | copied = data_skb->len; |
daa3766e | 2414 | |
1da177e4 LT |
2415 | skb_free_datagram(sk, skb); |
2416 | ||
16b304f3 PS |
2417 | if (nlk->cb_running && |
2418 | atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) { | |
b44d211e AV |
2419 | ret = netlink_dump(sk); |
2420 | if (ret) { | |
ac30ef83 | 2421 | sk->sk_err = -ret; |
b44d211e AV |
2422 | sk->sk_error_report(sk); |
2423 | } | |
2424 | } | |
1da177e4 LT |
2425 | |
2426 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
2427 | out: |
2428 | netlink_rcv_wake(sk); | |
2429 | return err ? : copied; | |
2430 | } | |
2431 | ||
676d2369 | 2432 | static void netlink_data_ready(struct sock *sk) |
1da177e4 | 2433 | { |
cd40b7d3 | 2434 | BUG(); |
1da177e4 LT |
2435 | } |
2436 | ||
2437 | /* | |
746fac4d | 2438 | * We export these functions to other modules. They provide a |
1da177e4 LT |
2439 | * complete set of kernel non-blocking support for message |
2440 | * queueing. | |
2441 | */ | |
2442 | ||
2443 | struct sock * | |
9f00d977 PNA |
2444 | __netlink_kernel_create(struct net *net, int unit, struct module *module, |
2445 | struct netlink_kernel_cfg *cfg) | |
1da177e4 LT |
2446 | { |
2447 | struct socket *sock; | |
2448 | struct sock *sk; | |
77247bbb | 2449 | struct netlink_sock *nlk; |
5c398dc8 | 2450 | struct listeners *listeners = NULL; |
a31f2d17 PNA |
2451 | struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL; |
2452 | unsigned int groups; | |
1da177e4 | 2453 | |
fab2caf6 | 2454 | BUG_ON(!nl_table); |
1da177e4 | 2455 | |
6ac552fd | 2456 | if (unit < 0 || unit >= MAX_LINKS) |
1da177e4 LT |
2457 | return NULL; |
2458 | ||
2459 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
2460 | return NULL; | |
2461 | ||
23fe1866 PE |
2462 | /* |
2463 | * We have to just have a reference on the net from sk, but don't | |
2464 | * get_net it. Besides, we cannot get and then put the net here. | |
2465 | * So we create one inside init_net and the move it to net. | |
2466 | */ | |
2467 | ||
2468 | if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0) | |
2469 | goto out_sock_release_nosk; | |
2470 | ||
2471 | sk = sock->sk; | |
edf02087 | 2472 | sk_change_net(sk, net); |
4fdb3bb7 | 2473 | |
a31f2d17 | 2474 | if (!cfg || cfg->groups < 32) |
4277a083 | 2475 | groups = 32; |
a31f2d17 PNA |
2476 | else |
2477 | groups = cfg->groups; | |
4277a083 | 2478 | |
5c398dc8 | 2479 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
4277a083 PM |
2480 | if (!listeners) |
2481 | goto out_sock_release; | |
2482 | ||
1da177e4 | 2483 | sk->sk_data_ready = netlink_data_ready; |
a31f2d17 PNA |
2484 | if (cfg && cfg->input) |
2485 | nlk_sk(sk)->netlink_rcv = cfg->input; | |
1da177e4 | 2486 | |
b4b51029 | 2487 | if (netlink_insert(sk, net, 0)) |
77247bbb | 2488 | goto out_sock_release; |
4fdb3bb7 | 2489 | |
77247bbb PM |
2490 | nlk = nlk_sk(sk); |
2491 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 2492 | |
4fdb3bb7 | 2493 | netlink_table_grab(); |
b4b51029 EB |
2494 | if (!nl_table[unit].registered) { |
2495 | nl_table[unit].groups = groups; | |
5c398dc8 | 2496 | rcu_assign_pointer(nl_table[unit].listeners, listeners); |
b4b51029 EB |
2497 | nl_table[unit].cb_mutex = cb_mutex; |
2498 | nl_table[unit].module = module; | |
9785e10a PNA |
2499 | if (cfg) { |
2500 | nl_table[unit].bind = cfg->bind; | |
2501 | nl_table[unit].flags = cfg->flags; | |
da12c90e G |
2502 | if (cfg->compare) |
2503 | nl_table[unit].compare = cfg->compare; | |
9785e10a | 2504 | } |
b4b51029 | 2505 | nl_table[unit].registered = 1; |
f937f1f4 JJ |
2506 | } else { |
2507 | kfree(listeners); | |
869e58f8 | 2508 | nl_table[unit].registered++; |
b4b51029 | 2509 | } |
4fdb3bb7 | 2510 | netlink_table_ungrab(); |
77247bbb PM |
2511 | return sk; |
2512 | ||
4fdb3bb7 | 2513 | out_sock_release: |
4277a083 | 2514 | kfree(listeners); |
9dfbec1f | 2515 | netlink_kernel_release(sk); |
23fe1866 PE |
2516 | return NULL; |
2517 | ||
2518 | out_sock_release_nosk: | |
4fdb3bb7 | 2519 | sock_release(sock); |
77247bbb | 2520 | return NULL; |
1da177e4 | 2521 | } |
9f00d977 | 2522 | EXPORT_SYMBOL(__netlink_kernel_create); |
b7c6ba6e DL |
2523 | |
2524 | void | |
2525 | netlink_kernel_release(struct sock *sk) | |
2526 | { | |
edf02087 | 2527 | sk_release_kernel(sk); |
b7c6ba6e DL |
2528 | } |
2529 | EXPORT_SYMBOL(netlink_kernel_release); | |
2530 | ||
d136f1bd | 2531 | int __netlink_change_ngroups(struct sock *sk, unsigned int groups) |
b4ff4f04 | 2532 | { |
5c398dc8 | 2533 | struct listeners *new, *old; |
b4ff4f04 | 2534 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; |
b4ff4f04 JB |
2535 | |
2536 | if (groups < 32) | |
2537 | groups = 32; | |
2538 | ||
b4ff4f04 | 2539 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { |
5c398dc8 ED |
2540 | new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC); |
2541 | if (!new) | |
d136f1bd | 2542 | return -ENOMEM; |
6d772ac5 | 2543 | old = nl_deref_protected(tbl->listeners); |
5c398dc8 ED |
2544 | memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups)); |
2545 | rcu_assign_pointer(tbl->listeners, new); | |
2546 | ||
37b6b935 | 2547 | kfree_rcu(old, rcu); |
b4ff4f04 JB |
2548 | } |
2549 | tbl->groups = groups; | |
2550 | ||
d136f1bd JB |
2551 | return 0; |
2552 | } | |
2553 | ||
2554 | /** | |
2555 | * netlink_change_ngroups - change number of multicast groups | |
2556 | * | |
2557 | * This changes the number of multicast groups that are available | |
2558 | * on a certain netlink family. Note that it is not possible to | |
2559 | * change the number of groups to below 32. Also note that it does | |
2560 | * not implicitly call netlink_clear_multicast_users() when the | |
2561 | * number of groups is reduced. | |
2562 | * | |
2563 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
2564 | * @groups: The new number of groups. | |
2565 | */ | |
2566 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
2567 | { | |
2568 | int err; | |
2569 | ||
2570 | netlink_table_grab(); | |
2571 | err = __netlink_change_ngroups(sk, groups); | |
b4ff4f04 | 2572 | netlink_table_ungrab(); |
d136f1bd | 2573 | |
b4ff4f04 JB |
2574 | return err; |
2575 | } | |
b4ff4f04 | 2576 | |
b8273570 JB |
2577 | void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group) |
2578 | { | |
2579 | struct sock *sk; | |
b8273570 JB |
2580 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; |
2581 | ||
b67bfe0d | 2582 | sk_for_each_bound(sk, &tbl->mc_list) |
b8273570 JB |
2583 | netlink_update_socket_mc(nlk_sk(sk), group, 0); |
2584 | } | |
2585 | ||
a46621a3 | 2586 | struct nlmsghdr * |
15e47304 | 2587 | __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags) |
a46621a3 DV |
2588 | { |
2589 | struct nlmsghdr *nlh; | |
573ce260 | 2590 | int size = nlmsg_msg_size(len); |
a46621a3 | 2591 | |
23b45672 | 2592 | nlh = (struct nlmsghdr *)skb_put(skb, NLMSG_ALIGN(size)); |
a46621a3 DV |
2593 | nlh->nlmsg_type = type; |
2594 | nlh->nlmsg_len = size; | |
2595 | nlh->nlmsg_flags = flags; | |
15e47304 | 2596 | nlh->nlmsg_pid = portid; |
a46621a3 DV |
2597 | nlh->nlmsg_seq = seq; |
2598 | if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0) | |
573ce260 | 2599 | memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size); |
a46621a3 DV |
2600 | return nlh; |
2601 | } | |
2602 | EXPORT_SYMBOL(__nlmsg_put); | |
2603 | ||
1da177e4 LT |
2604 | /* |
2605 | * It looks a bit ugly. | |
2606 | * It would be better to create kernel thread. | |
2607 | */ | |
2608 | ||
2609 | static int netlink_dump(struct sock *sk) | |
2610 | { | |
2611 | struct netlink_sock *nlk = nlk_sk(sk); | |
2612 | struct netlink_callback *cb; | |
c7ac8679 | 2613 | struct sk_buff *skb = NULL; |
1da177e4 | 2614 | struct nlmsghdr *nlh; |
bf8b79e4 | 2615 | int len, err = -ENOBUFS; |
c7ac8679 | 2616 | int alloc_size; |
1da177e4 | 2617 | |
af65bdfc | 2618 | mutex_lock(nlk->cb_mutex); |
16b304f3 | 2619 | if (!nlk->cb_running) { |
bf8b79e4 TG |
2620 | err = -EINVAL; |
2621 | goto errout_skb; | |
1da177e4 LT |
2622 | } |
2623 | ||
16b304f3 | 2624 | cb = &nlk->cb; |
c7ac8679 GR |
2625 | alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE); |
2626 | ||
f9c22888 PM |
2627 | if (!netlink_rx_is_mmaped(sk) && |
2628 | atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) | |
2629 | goto errout_skb; | |
9063e21f ED |
2630 | |
2631 | /* NLMSG_GOODSIZE is small to avoid high order allocations being | |
2632 | * required, but it makes sense to _attempt_ a 16K bytes allocation | |
2633 | * to reduce number of system calls on dump operations, if user | |
2634 | * ever provided a big enough buffer. | |
2635 | */ | |
2636 | if (alloc_size < nlk->max_recvmsg_len) { | |
2637 | skb = netlink_alloc_skb(sk, | |
2638 | nlk->max_recvmsg_len, | |
2639 | nlk->portid, | |
2640 | GFP_KERNEL | | |
2641 | __GFP_NOWARN | | |
2642 | __GFP_NORETRY); | |
2643 | /* available room should be exact amount to avoid MSG_TRUNC */ | |
2644 | if (skb) | |
2645 | skb_reserve(skb, skb_tailroom(skb) - | |
2646 | nlk->max_recvmsg_len); | |
2647 | } | |
2648 | if (!skb) | |
2649 | skb = netlink_alloc_skb(sk, alloc_size, nlk->portid, | |
2650 | GFP_KERNEL); | |
c7ac8679 | 2651 | if (!skb) |
c63d6ea3 | 2652 | goto errout_skb; |
f9c22888 | 2653 | netlink_skb_set_owner_r(skb, sk); |
c7ac8679 | 2654 | |
1da177e4 LT |
2655 | len = cb->dump(skb, cb); |
2656 | ||
2657 | if (len > 0) { | |
af65bdfc | 2658 | mutex_unlock(nlk->cb_mutex); |
b1153f29 SH |
2659 | |
2660 | if (sk_filter(sk, skb)) | |
2661 | kfree_skb(skb); | |
4a7e7c2a ED |
2662 | else |
2663 | __netlink_sendskb(sk, skb); | |
1da177e4 LT |
2664 | return 0; |
2665 | } | |
2666 | ||
bf8b79e4 TG |
2667 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
2668 | if (!nlh) | |
2669 | goto errout_skb; | |
2670 | ||
670dc283 JB |
2671 | nl_dump_check_consistent(cb, nlh); |
2672 | ||
bf8b79e4 TG |
2673 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); |
2674 | ||
b1153f29 SH |
2675 | if (sk_filter(sk, skb)) |
2676 | kfree_skb(skb); | |
4a7e7c2a ED |
2677 | else |
2678 | __netlink_sendskb(sk, skb); | |
1da177e4 | 2679 | |
a8f74b22 TG |
2680 | if (cb->done) |
2681 | cb->done(cb); | |
1da177e4 | 2682 | |
16b304f3 PS |
2683 | nlk->cb_running = false; |
2684 | mutex_unlock(nlk->cb_mutex); | |
6dc878a8 | 2685 | module_put(cb->module); |
16b304f3 | 2686 | consume_skb(cb->skb); |
1da177e4 | 2687 | return 0; |
1797754e | 2688 | |
bf8b79e4 | 2689 | errout_skb: |
af65bdfc | 2690 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 | 2691 | kfree_skb(skb); |
bf8b79e4 | 2692 | return err; |
1da177e4 LT |
2693 | } |
2694 | ||
6dc878a8 G |
2695 | int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb, |
2696 | const struct nlmsghdr *nlh, | |
2697 | struct netlink_dump_control *control) | |
1da177e4 LT |
2698 | { |
2699 | struct netlink_callback *cb; | |
2700 | struct sock *sk; | |
2701 | struct netlink_sock *nlk; | |
b44d211e | 2702 | int ret; |
1da177e4 | 2703 | |
f9c22888 PM |
2704 | /* Memory mapped dump requests need to be copied to avoid looping |
2705 | * on the pending state in netlink_mmap_sendmsg() while the CB hold | |
2706 | * a reference to the skb. | |
2707 | */ | |
2708 | if (netlink_skb_is_mmaped(skb)) { | |
2709 | skb = skb_copy(skb, GFP_KERNEL); | |
16b304f3 | 2710 | if (skb == NULL) |
f9c22888 | 2711 | return -ENOBUFS; |
f9c22888 PM |
2712 | } else |
2713 | atomic_inc(&skb->users); | |
2714 | ||
15e47304 | 2715 | sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid); |
1da177e4 | 2716 | if (sk == NULL) { |
16b304f3 PS |
2717 | ret = -ECONNREFUSED; |
2718 | goto error_free; | |
1da177e4 | 2719 | } |
6dc878a8 | 2720 | |
16b304f3 | 2721 | nlk = nlk_sk(sk); |
af65bdfc | 2722 | mutex_lock(nlk->cb_mutex); |
6dc878a8 | 2723 | /* A dump is in progress... */ |
16b304f3 | 2724 | if (nlk->cb_running) { |
6dc878a8 | 2725 | ret = -EBUSY; |
16b304f3 | 2726 | goto error_unlock; |
1da177e4 | 2727 | } |
6dc878a8 | 2728 | /* add reference of module which cb->dump belongs to */ |
16b304f3 | 2729 | if (!try_module_get(control->module)) { |
6dc878a8 | 2730 | ret = -EPROTONOSUPPORT; |
16b304f3 | 2731 | goto error_unlock; |
6dc878a8 G |
2732 | } |
2733 | ||
16b304f3 PS |
2734 | cb = &nlk->cb; |
2735 | memset(cb, 0, sizeof(*cb)); | |
2736 | cb->dump = control->dump; | |
2737 | cb->done = control->done; | |
2738 | cb->nlh = nlh; | |
2739 | cb->data = control->data; | |
2740 | cb->module = control->module; | |
2741 | cb->min_dump_alloc = control->min_dump_alloc; | |
2742 | cb->skb = skb; | |
2743 | ||
2744 | nlk->cb_running = true; | |
2745 | ||
af65bdfc | 2746 | mutex_unlock(nlk->cb_mutex); |
1da177e4 | 2747 | |
b44d211e | 2748 | ret = netlink_dump(sk); |
1da177e4 | 2749 | sock_put(sk); |
5c58298c | 2750 | |
b44d211e AV |
2751 | if (ret) |
2752 | return ret; | |
2753 | ||
5c58298c DL |
2754 | /* We successfully started a dump, by returning -EINTR we |
2755 | * signal not to send ACK even if it was requested. | |
2756 | */ | |
2757 | return -EINTR; | |
16b304f3 PS |
2758 | |
2759 | error_unlock: | |
2760 | sock_put(sk); | |
2761 | mutex_unlock(nlk->cb_mutex); | |
2762 | error_free: | |
2763 | kfree_skb(skb); | |
2764 | return ret; | |
1da177e4 | 2765 | } |
6dc878a8 | 2766 | EXPORT_SYMBOL(__netlink_dump_start); |
1da177e4 LT |
2767 | |
2768 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
2769 | { | |
2770 | struct sk_buff *skb; | |
2771 | struct nlmsghdr *rep; | |
2772 | struct nlmsgerr *errmsg; | |
339bf98f | 2773 | size_t payload = sizeof(*errmsg); |
1da177e4 | 2774 | |
339bf98f TG |
2775 | /* error messages get the original request appened */ |
2776 | if (err) | |
2777 | payload += nlmsg_len(nlh); | |
1da177e4 | 2778 | |
f9c22888 PM |
2779 | skb = netlink_alloc_skb(in_skb->sk, nlmsg_total_size(payload), |
2780 | NETLINK_CB(in_skb).portid, GFP_KERNEL); | |
1da177e4 LT |
2781 | if (!skb) { |
2782 | struct sock *sk; | |
2783 | ||
3b1e0a65 | 2784 | sk = netlink_lookup(sock_net(in_skb->sk), |
b4b51029 | 2785 | in_skb->sk->sk_protocol, |
15e47304 | 2786 | NETLINK_CB(in_skb).portid); |
1da177e4 LT |
2787 | if (sk) { |
2788 | sk->sk_err = ENOBUFS; | |
2789 | sk->sk_error_report(sk); | |
2790 | sock_put(sk); | |
2791 | } | |
2792 | return; | |
2793 | } | |
2794 | ||
15e47304 | 2795 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, |
5dba93ae | 2796 | NLMSG_ERROR, payload, 0); |
bf8b79e4 | 2797 | errmsg = nlmsg_data(rep); |
1da177e4 | 2798 | errmsg->error = err; |
bf8b79e4 | 2799 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
15e47304 | 2800 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT); |
1da177e4 | 2801 | } |
6ac552fd | 2802 | EXPORT_SYMBOL(netlink_ack); |
1da177e4 | 2803 | |
cd40b7d3 | 2804 | int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 2805 | struct nlmsghdr *)) |
82ace47a | 2806 | { |
82ace47a TG |
2807 | struct nlmsghdr *nlh; |
2808 | int err; | |
2809 | ||
2810 | while (skb->len >= nlmsg_total_size(0)) { | |
cd40b7d3 DL |
2811 | int msglen; |
2812 | ||
b529ccf2 | 2813 | nlh = nlmsg_hdr(skb); |
d35b6856 | 2814 | err = 0; |
82ace47a | 2815 | |
ad8e4b75 | 2816 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
2817 | return 0; |
2818 | ||
d35b6856 TG |
2819 | /* Only requests are handled by the kernel */ |
2820 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
5c58298c | 2821 | goto ack; |
45e7ae7f TG |
2822 | |
2823 | /* Skip control messages */ | |
2824 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
5c58298c | 2825 | goto ack; |
d35b6856 | 2826 | |
1d00a4eb | 2827 | err = cb(skb, nlh); |
5c58298c DL |
2828 | if (err == -EINTR) |
2829 | goto skip; | |
2830 | ||
2831 | ack: | |
d35b6856 | 2832 | if (nlh->nlmsg_flags & NLM_F_ACK || err) |
82ace47a | 2833 | netlink_ack(skb, nlh, err); |
82ace47a | 2834 | |
5c58298c | 2835 | skip: |
6ac552fd | 2836 | msglen = NLMSG_ALIGN(nlh->nlmsg_len); |
cd40b7d3 DL |
2837 | if (msglen > skb->len) |
2838 | msglen = skb->len; | |
2839 | skb_pull(skb, msglen); | |
82ace47a TG |
2840 | } |
2841 | ||
2842 | return 0; | |
2843 | } | |
6ac552fd | 2844 | EXPORT_SYMBOL(netlink_rcv_skb); |
82ace47a | 2845 | |
d387f6ad TG |
2846 | /** |
2847 | * nlmsg_notify - send a notification netlink message | |
2848 | * @sk: netlink socket to use | |
2849 | * @skb: notification message | |
15e47304 | 2850 | * @portid: destination netlink portid for reports or 0 |
d387f6ad TG |
2851 | * @group: destination multicast group or 0 |
2852 | * @report: 1 to report back, 0 to disable | |
2853 | * @flags: allocation flags | |
2854 | */ | |
15e47304 | 2855 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid, |
d387f6ad TG |
2856 | unsigned int group, int report, gfp_t flags) |
2857 | { | |
2858 | int err = 0; | |
2859 | ||
2860 | if (group) { | |
15e47304 | 2861 | int exclude_portid = 0; |
d387f6ad TG |
2862 | |
2863 | if (report) { | |
2864 | atomic_inc(&skb->users); | |
15e47304 | 2865 | exclude_portid = portid; |
d387f6ad TG |
2866 | } |
2867 | ||
1ce85fe4 PNA |
2868 | /* errors reported via destination sk->sk_err, but propagate |
2869 | * delivery errors if NETLINK_BROADCAST_ERROR flag is set */ | |
15e47304 | 2870 | err = nlmsg_multicast(sk, skb, exclude_portid, group, flags); |
d387f6ad TG |
2871 | } |
2872 | ||
1ce85fe4 PNA |
2873 | if (report) { |
2874 | int err2; | |
2875 | ||
15e47304 | 2876 | err2 = nlmsg_unicast(sk, skb, portid); |
1ce85fe4 PNA |
2877 | if (!err || err == -ESRCH) |
2878 | err = err2; | |
2879 | } | |
d387f6ad TG |
2880 | |
2881 | return err; | |
2882 | } | |
6ac552fd | 2883 | EXPORT_SYMBOL(nlmsg_notify); |
d387f6ad | 2884 | |
1da177e4 LT |
2885 | #ifdef CONFIG_PROC_FS |
2886 | struct nl_seq_iter { | |
e372c414 | 2887 | struct seq_net_private p; |
1da177e4 LT |
2888 | int link; |
2889 | int hash_idx; | |
2890 | }; | |
2891 | ||
2892 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
2893 | { | |
2894 | struct nl_seq_iter *iter = seq->private; | |
2895 | int i, j; | |
e341694e | 2896 | struct netlink_sock *nlk; |
1da177e4 | 2897 | struct sock *s; |
1da177e4 LT |
2898 | loff_t off = 0; |
2899 | ||
6ac552fd | 2900 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e | 2901 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2902 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
e341694e TG |
2903 | |
2904 | for (j = 0; j < tbl->size; j++) { | |
2905 | rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) { | |
2906 | s = (struct sock *)nlk; | |
1da177e4 | 2907 | |
1218854a | 2908 | if (sock_net(s) != seq_file_net(seq)) |
b4b51029 | 2909 | continue; |
1da177e4 LT |
2910 | if (off == pos) { |
2911 | iter->link = i; | |
2912 | iter->hash_idx = j; | |
2913 | return s; | |
2914 | } | |
2915 | ++off; | |
2916 | } | |
2917 | } | |
2918 | } | |
2919 | return NULL; | |
2920 | } | |
2921 | ||
2922 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
2923 | { | |
e341694e | 2924 | rcu_read_lock(); |
1da177e4 LT |
2925 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
2926 | } | |
2927 | ||
2928 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
2929 | { | |
e341694e | 2930 | struct netlink_sock *nlk; |
1da177e4 | 2931 | struct nl_seq_iter *iter; |
da12c90e | 2932 | struct net *net; |
1da177e4 LT |
2933 | int i, j; |
2934 | ||
2935 | ++*pos; | |
2936 | ||
2937 | if (v == SEQ_START_TOKEN) | |
2938 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 2939 | |
da12c90e | 2940 | net = seq_file_net(seq); |
b4b51029 | 2941 | iter = seq->private; |
e341694e TG |
2942 | nlk = v; |
2943 | ||
2944 | rht_for_each_entry_rcu(nlk, nlk->node.next, node) | |
2945 | if (net_eq(sock_net((struct sock *)nlk), net)) | |
2946 | return nlk; | |
1da177e4 | 2947 | |
1da177e4 LT |
2948 | i = iter->link; |
2949 | j = iter->hash_idx + 1; | |
2950 | ||
2951 | do { | |
e341694e | 2952 | struct rhashtable *ht = &nl_table[i].hash; |
67a24ac1 | 2953 | const struct bucket_table *tbl = rht_dereference_rcu(ht->tbl, ht); |
da12c90e | 2954 | |
e341694e TG |
2955 | for (; j < tbl->size; j++) { |
2956 | rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) { | |
2957 | if (net_eq(sock_net((struct sock *)nlk), net)) { | |
2958 | iter->link = i; | |
2959 | iter->hash_idx = j; | |
2960 | return nlk; | |
2961 | } | |
1da177e4 LT |
2962 | } |
2963 | } | |
2964 | ||
2965 | j = 0; | |
2966 | } while (++i < MAX_LINKS); | |
2967 | ||
2968 | return NULL; | |
2969 | } | |
2970 | ||
2971 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
2972 | { | |
e341694e | 2973 | rcu_read_unlock(); |
1da177e4 LT |
2974 | } |
2975 | ||
2976 | ||
2977 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
2978 | { | |
658cb354 | 2979 | if (v == SEQ_START_TOKEN) { |
1da177e4 LT |
2980 | seq_puts(seq, |
2981 | "sk Eth Pid Groups " | |
cf0aa4e0 | 2982 | "Rmem Wmem Dump Locks Drops Inode\n"); |
658cb354 | 2983 | } else { |
1da177e4 LT |
2984 | struct sock *s = v; |
2985 | struct netlink_sock *nlk = nlk_sk(s); | |
2986 | ||
16b304f3 | 2987 | seq_printf(seq, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n", |
1da177e4 LT |
2988 | s, |
2989 | s->sk_protocol, | |
15e47304 | 2990 | nlk->portid, |
513c2500 | 2991 | nlk->groups ? (u32)nlk->groups[0] : 0, |
31e6d363 ED |
2992 | sk_rmem_alloc_get(s), |
2993 | sk_wmem_alloc_get(s), | |
16b304f3 | 2994 | nlk->cb_running, |
38938bfe | 2995 | atomic_read(&s->sk_refcnt), |
cf0aa4e0 MY |
2996 | atomic_read(&s->sk_drops), |
2997 | sock_i_ino(s) | |
1da177e4 LT |
2998 | ); |
2999 | ||
3000 | } | |
3001 | return 0; | |
3002 | } | |
3003 | ||
56b3d975 | 3004 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
3005 | .start = netlink_seq_start, |
3006 | .next = netlink_seq_next, | |
3007 | .stop = netlink_seq_stop, | |
3008 | .show = netlink_seq_show, | |
3009 | }; | |
3010 | ||
3011 | ||
3012 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
3013 | { | |
e372c414 DL |
3014 | return seq_open_net(inode, file, &netlink_seq_ops, |
3015 | sizeof(struct nl_seq_iter)); | |
b4b51029 EB |
3016 | } |
3017 | ||
da7071d7 | 3018 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
3019 | .owner = THIS_MODULE, |
3020 | .open = netlink_seq_open, | |
3021 | .read = seq_read, | |
3022 | .llseek = seq_lseek, | |
e372c414 | 3023 | .release = seq_release_net, |
1da177e4 LT |
3024 | }; |
3025 | ||
3026 | #endif | |
3027 | ||
3028 | int netlink_register_notifier(struct notifier_block *nb) | |
3029 | { | |
e041c683 | 3030 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 | 3031 | } |
6ac552fd | 3032 | EXPORT_SYMBOL(netlink_register_notifier); |
1da177e4 LT |
3033 | |
3034 | int netlink_unregister_notifier(struct notifier_block *nb) | |
3035 | { | |
e041c683 | 3036 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 3037 | } |
6ac552fd | 3038 | EXPORT_SYMBOL(netlink_unregister_notifier); |
746fac4d | 3039 | |
90ddc4f0 | 3040 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
3041 | .family = PF_NETLINK, |
3042 | .owner = THIS_MODULE, | |
3043 | .release = netlink_release, | |
3044 | .bind = netlink_bind, | |
3045 | .connect = netlink_connect, | |
3046 | .socketpair = sock_no_socketpair, | |
3047 | .accept = sock_no_accept, | |
3048 | .getname = netlink_getname, | |
9652e931 | 3049 | .poll = netlink_poll, |
1da177e4 LT |
3050 | .ioctl = sock_no_ioctl, |
3051 | .listen = sock_no_listen, | |
3052 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
3053 | .setsockopt = netlink_setsockopt, |
3054 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
3055 | .sendmsg = netlink_sendmsg, |
3056 | .recvmsg = netlink_recvmsg, | |
ccdfcc39 | 3057 | .mmap = netlink_mmap, |
1da177e4 LT |
3058 | .sendpage = sock_no_sendpage, |
3059 | }; | |
3060 | ||
ec1b4cf7 | 3061 | static const struct net_proto_family netlink_family_ops = { |
1da177e4 LT |
3062 | .family = PF_NETLINK, |
3063 | .create = netlink_create, | |
3064 | .owner = THIS_MODULE, /* for consistency 8) */ | |
3065 | }; | |
3066 | ||
4665079c | 3067 | static int __net_init netlink_net_init(struct net *net) |
b4b51029 EB |
3068 | { |
3069 | #ifdef CONFIG_PROC_FS | |
d4beaa66 | 3070 | if (!proc_create("netlink", 0, net->proc_net, &netlink_seq_fops)) |
b4b51029 EB |
3071 | return -ENOMEM; |
3072 | #endif | |
3073 | return 0; | |
3074 | } | |
3075 | ||
4665079c | 3076 | static void __net_exit netlink_net_exit(struct net *net) |
b4b51029 EB |
3077 | { |
3078 | #ifdef CONFIG_PROC_FS | |
ece31ffd | 3079 | remove_proc_entry("netlink", net->proc_net); |
b4b51029 EB |
3080 | #endif |
3081 | } | |
3082 | ||
b963ea89 DM |
3083 | static void __init netlink_add_usersock_entry(void) |
3084 | { | |
5c398dc8 | 3085 | struct listeners *listeners; |
b963ea89 DM |
3086 | int groups = 32; |
3087 | ||
5c398dc8 | 3088 | listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL); |
b963ea89 | 3089 | if (!listeners) |
5c398dc8 | 3090 | panic("netlink_add_usersock_entry: Cannot allocate listeners\n"); |
b963ea89 DM |
3091 | |
3092 | netlink_table_grab(); | |
3093 | ||
3094 | nl_table[NETLINK_USERSOCK].groups = groups; | |
5c398dc8 | 3095 | rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners); |
b963ea89 DM |
3096 | nl_table[NETLINK_USERSOCK].module = THIS_MODULE; |
3097 | nl_table[NETLINK_USERSOCK].registered = 1; | |
9785e10a | 3098 | nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND; |
b963ea89 DM |
3099 | |
3100 | netlink_table_ungrab(); | |
3101 | } | |
3102 | ||
022cbae6 | 3103 | static struct pernet_operations __net_initdata netlink_net_ops = { |
b4b51029 EB |
3104 | .init = netlink_net_init, |
3105 | .exit = netlink_net_exit, | |
3106 | }; | |
3107 | ||
1da177e4 LT |
3108 | static int __init netlink_proto_init(void) |
3109 | { | |
1da177e4 | 3110 | int i; |
1da177e4 | 3111 | int err = proto_register(&netlink_proto, 0); |
e341694e TG |
3112 | struct rhashtable_params ht_params = { |
3113 | .head_offset = offsetof(struct netlink_sock, node), | |
3114 | .key_offset = offsetof(struct netlink_sock, portid), | |
3115 | .key_len = sizeof(u32), /* portid */ | |
3116 | .hashfn = arch_fast_hash, | |
3117 | .max_shift = 16, /* 64K */ | |
3118 | .grow_decision = rht_grow_above_75, | |
3119 | .shrink_decision = rht_shrink_below_30, | |
3120 | .mutex_is_held = lockdep_nl_sk_hash_is_held, | |
3121 | }; | |
1da177e4 LT |
3122 | |
3123 | if (err != 0) | |
3124 | goto out; | |
3125 | ||
fab25745 | 3126 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 | 3127 | |
0da974f4 | 3128 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
3129 | if (!nl_table) |
3130 | goto panic; | |
1da177e4 | 3131 | |
1da177e4 | 3132 | for (i = 0; i < MAX_LINKS; i++) { |
e341694e TG |
3133 | if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) { |
3134 | while (--i > 0) | |
3135 | rhashtable_destroy(&nl_table[i].hash); | |
1da177e4 | 3136 | kfree(nl_table); |
fab2caf6 | 3137 | goto panic; |
1da177e4 | 3138 | } |
1da177e4 LT |
3139 | } |
3140 | ||
bcbde0d4 DB |
3141 | INIT_LIST_HEAD(&netlink_tap_all); |
3142 | ||
b963ea89 DM |
3143 | netlink_add_usersock_entry(); |
3144 | ||
1da177e4 | 3145 | sock_register(&netlink_family_ops); |
b4b51029 | 3146 | register_pernet_subsys(&netlink_net_ops); |
746fac4d | 3147 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
3148 | rtnetlink_init(); |
3149 | out: | |
3150 | return err; | |
fab2caf6 AM |
3151 | panic: |
3152 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
3153 | } |
3154 | ||
1da177e4 | 3155 | core_initcall(netlink_proto_init); |