]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * NETLINK Kernel-user communication protocol. | |
3 | * | |
4 | * Authors: Alan Cox <alan@redhat.com> | |
5 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
746fac4d | 11 | * |
1da177e4 LT |
12 | * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith |
13 | * added netlink_proto_exit | |
14 | * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br> | |
15 | * use nlk_sk, as sk->protinfo is on a diet 8) | |
4fdb3bb7 HW |
16 | * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org> |
17 | * - inc module use count of module that owns | |
18 | * the kernel socket in case userspace opens | |
19 | * socket of same protocol | |
20 | * - remove all module support, since netlink is | |
21 | * mandatory if CONFIG_NET=y these days | |
1da177e4 LT |
22 | */ |
23 | ||
1da177e4 LT |
24 | #include <linux/module.h> |
25 | ||
4fc268d2 | 26 | #include <linux/capability.h> |
1da177e4 LT |
27 | #include <linux/kernel.h> |
28 | #include <linux/init.h> | |
1da177e4 LT |
29 | #include <linux/signal.h> |
30 | #include <linux/sched.h> | |
31 | #include <linux/errno.h> | |
32 | #include <linux/string.h> | |
33 | #include <linux/stat.h> | |
34 | #include <linux/socket.h> | |
35 | #include <linux/un.h> | |
36 | #include <linux/fcntl.h> | |
37 | #include <linux/termios.h> | |
38 | #include <linux/sockios.h> | |
39 | #include <linux/net.h> | |
40 | #include <linux/fs.h> | |
41 | #include <linux/slab.h> | |
42 | #include <asm/uaccess.h> | |
43 | #include <linux/skbuff.h> | |
44 | #include <linux/netdevice.h> | |
45 | #include <linux/rtnetlink.h> | |
46 | #include <linux/proc_fs.h> | |
47 | #include <linux/seq_file.h> | |
1da177e4 LT |
48 | #include <linux/notifier.h> |
49 | #include <linux/security.h> | |
50 | #include <linux/jhash.h> | |
51 | #include <linux/jiffies.h> | |
52 | #include <linux/random.h> | |
53 | #include <linux/bitops.h> | |
54 | #include <linux/mm.h> | |
55 | #include <linux/types.h> | |
54e0f520 | 56 | #include <linux/audit.h> |
e7c34970 | 57 | #include <linux/selinux.h> |
af65bdfc | 58 | #include <linux/mutex.h> |
54e0f520 | 59 | |
1da177e4 LT |
60 | #include <net/sock.h> |
61 | #include <net/scm.h> | |
82ace47a | 62 | #include <net/netlink.h> |
1da177e4 | 63 | |
f7fa9b10 | 64 | #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8) |
b4ff4f04 | 65 | #define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long)) |
1da177e4 LT |
66 | |
67 | struct netlink_sock { | |
68 | /* struct sock has to be the first member of netlink_sock */ | |
69 | struct sock sk; | |
70 | u32 pid; | |
1da177e4 | 71 | u32 dst_pid; |
d629b836 | 72 | u32 dst_group; |
f7fa9b10 PM |
73 | u32 flags; |
74 | u32 subscriptions; | |
75 | u32 ngroups; | |
76 | unsigned long *groups; | |
1da177e4 LT |
77 | unsigned long state; |
78 | wait_queue_head_t wait; | |
79 | struct netlink_callback *cb; | |
af65bdfc PM |
80 | struct mutex *cb_mutex; |
81 | struct mutex cb_def_mutex; | |
1da177e4 | 82 | void (*data_ready)(struct sock *sk, int bytes); |
77247bbb | 83 | struct module *module; |
1da177e4 LT |
84 | }; |
85 | ||
77247bbb | 86 | #define NETLINK_KERNEL_SOCKET 0x1 |
9a4595bc | 87 | #define NETLINK_RECV_PKTINFO 0x2 |
77247bbb | 88 | |
1da177e4 LT |
89 | static inline struct netlink_sock *nlk_sk(struct sock *sk) |
90 | { | |
91 | return (struct netlink_sock *)sk; | |
92 | } | |
93 | ||
94 | struct nl_pid_hash { | |
95 | struct hlist_head *table; | |
96 | unsigned long rehash_time; | |
97 | ||
98 | unsigned int mask; | |
99 | unsigned int shift; | |
100 | ||
101 | unsigned int entries; | |
102 | unsigned int max_shift; | |
103 | ||
104 | u32 rnd; | |
105 | }; | |
106 | ||
107 | struct netlink_table { | |
108 | struct nl_pid_hash hash; | |
109 | struct hlist_head mc_list; | |
4277a083 | 110 | unsigned long *listeners; |
1da177e4 | 111 | unsigned int nl_nonroot; |
f7fa9b10 | 112 | unsigned int groups; |
af65bdfc | 113 | struct mutex *cb_mutex; |
77247bbb | 114 | struct module *module; |
ab33a171 | 115 | int registered; |
1da177e4 LT |
116 | }; |
117 | ||
118 | static struct netlink_table *nl_table; | |
119 | ||
120 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
121 | ||
122 | static int netlink_dump(struct sock *sk); | |
123 | static void netlink_destroy_callback(struct netlink_callback *cb); | |
42bad1da | 124 | static void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb); |
1da177e4 LT |
125 | |
126 | static DEFINE_RWLOCK(nl_table_lock); | |
127 | static atomic_t nl_table_users = ATOMIC_INIT(0); | |
128 | ||
e041c683 | 129 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 130 | |
d629b836 PM |
131 | static u32 netlink_group_mask(u32 group) |
132 | { | |
133 | return group ? 1 << (group - 1) : 0; | |
134 | } | |
135 | ||
1da177e4 LT |
136 | static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid) |
137 | { | |
138 | return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask]; | |
139 | } | |
140 | ||
141 | static void netlink_sock_destruct(struct sock *sk) | |
142 | { | |
3f660d66 HX |
143 | struct netlink_sock *nlk = nlk_sk(sk); |
144 | ||
3f660d66 HX |
145 | if (nlk->cb) { |
146 | if (nlk->cb->done) | |
147 | nlk->cb->done(nlk->cb); | |
148 | netlink_destroy_callback(nlk->cb); | |
149 | } | |
150 | ||
1da177e4 LT |
151 | skb_queue_purge(&sk->sk_receive_queue); |
152 | ||
153 | if (!sock_flag(sk, SOCK_DEAD)) { | |
154 | printk("Freeing alive netlink socket %p\n", sk); | |
155 | return; | |
156 | } | |
157 | BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); | |
158 | BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); | |
f7fa9b10 | 159 | BUG_TRAP(!nlk_sk(sk)->groups); |
1da177e4 LT |
160 | } |
161 | ||
162 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP. | |
163 | * Look, when several writers sleep and reader wakes them up, all but one | |
164 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves | |
165 | * this, _but_ remember, it adds useless work on UP machines. | |
166 | */ | |
167 | ||
168 | static void netlink_table_grab(void) | |
169 | { | |
6abd219c | 170 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
171 | |
172 | if (atomic_read(&nl_table_users)) { | |
173 | DECLARE_WAITQUEUE(wait, current); | |
174 | ||
175 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
176 | for(;;) { | |
177 | set_current_state(TASK_UNINTERRUPTIBLE); | |
178 | if (atomic_read(&nl_table_users) == 0) | |
179 | break; | |
6abd219c | 180 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 181 | schedule(); |
6abd219c | 182 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
183 | } |
184 | ||
185 | __set_current_state(TASK_RUNNING); | |
186 | remove_wait_queue(&nl_table_wait, &wait); | |
187 | } | |
188 | } | |
189 | ||
190 | static __inline__ void netlink_table_ungrab(void) | |
191 | { | |
6abd219c | 192 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
193 | wake_up(&nl_table_wait); |
194 | } | |
195 | ||
196 | static __inline__ void | |
197 | netlink_lock_table(void) | |
198 | { | |
199 | /* read_lock() synchronizes us to netlink_table_grab */ | |
200 | ||
201 | read_lock(&nl_table_lock); | |
202 | atomic_inc(&nl_table_users); | |
203 | read_unlock(&nl_table_lock); | |
204 | } | |
205 | ||
206 | static __inline__ void | |
207 | netlink_unlock_table(void) | |
208 | { | |
209 | if (atomic_dec_and_test(&nl_table_users)) | |
210 | wake_up(&nl_table_wait); | |
211 | } | |
212 | ||
213 | static __inline__ struct sock *netlink_lookup(int protocol, u32 pid) | |
214 | { | |
215 | struct nl_pid_hash *hash = &nl_table[protocol].hash; | |
216 | struct hlist_head *head; | |
217 | struct sock *sk; | |
218 | struct hlist_node *node; | |
219 | ||
220 | read_lock(&nl_table_lock); | |
221 | head = nl_pid_hashfn(hash, pid); | |
222 | sk_for_each(sk, node, head) { | |
223 | if (nlk_sk(sk)->pid == pid) { | |
224 | sock_hold(sk); | |
225 | goto found; | |
226 | } | |
227 | } | |
228 | sk = NULL; | |
229 | found: | |
230 | read_unlock(&nl_table_lock); | |
231 | return sk; | |
232 | } | |
233 | ||
234 | static inline struct hlist_head *nl_pid_hash_alloc(size_t size) | |
235 | { | |
236 | if (size <= PAGE_SIZE) | |
237 | return kmalloc(size, GFP_ATOMIC); | |
238 | else | |
239 | return (struct hlist_head *) | |
240 | __get_free_pages(GFP_ATOMIC, get_order(size)); | |
241 | } | |
242 | ||
243 | static inline void nl_pid_hash_free(struct hlist_head *table, size_t size) | |
244 | { | |
245 | if (size <= PAGE_SIZE) | |
246 | kfree(table); | |
247 | else | |
248 | free_pages((unsigned long)table, get_order(size)); | |
249 | } | |
250 | ||
251 | static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow) | |
252 | { | |
253 | unsigned int omask, mask, shift; | |
254 | size_t osize, size; | |
255 | struct hlist_head *otable, *table; | |
256 | int i; | |
257 | ||
258 | omask = mask = hash->mask; | |
259 | osize = size = (mask + 1) * sizeof(*table); | |
260 | shift = hash->shift; | |
261 | ||
262 | if (grow) { | |
263 | if (++shift > hash->max_shift) | |
264 | return 0; | |
265 | mask = mask * 2 + 1; | |
266 | size *= 2; | |
267 | } | |
268 | ||
269 | table = nl_pid_hash_alloc(size); | |
270 | if (!table) | |
271 | return 0; | |
272 | ||
273 | memset(table, 0, size); | |
274 | otable = hash->table; | |
275 | hash->table = table; | |
276 | hash->mask = mask; | |
277 | hash->shift = shift; | |
278 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
279 | ||
280 | for (i = 0; i <= omask; i++) { | |
281 | struct sock *sk; | |
282 | struct hlist_node *node, *tmp; | |
283 | ||
284 | sk_for_each_safe(sk, node, tmp, &otable[i]) | |
285 | __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid)); | |
286 | } | |
287 | ||
288 | nl_pid_hash_free(otable, osize); | |
289 | hash->rehash_time = jiffies + 10 * 60 * HZ; | |
290 | return 1; | |
291 | } | |
292 | ||
293 | static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len) | |
294 | { | |
295 | int avg = hash->entries >> hash->shift; | |
296 | ||
297 | if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1)) | |
298 | return 1; | |
299 | ||
300 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
301 | nl_pid_hash_rehash(hash, 0); | |
302 | return 1; | |
303 | } | |
304 | ||
305 | return 0; | |
306 | } | |
307 | ||
90ddc4f0 | 308 | static const struct proto_ops netlink_ops; |
1da177e4 | 309 | |
4277a083 PM |
310 | static void |
311 | netlink_update_listeners(struct sock *sk) | |
312 | { | |
313 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
314 | struct hlist_node *node; | |
315 | unsigned long mask; | |
316 | unsigned int i; | |
317 | ||
b4ff4f04 | 318 | for (i = 0; i < NLGRPLONGS(tbl->groups); i++) { |
4277a083 | 319 | mask = 0; |
b4ff4f04 JB |
320 | sk_for_each_bound(sk, node, &tbl->mc_list) { |
321 | if (i < NLGRPLONGS(nlk_sk(sk)->ngroups)) | |
322 | mask |= nlk_sk(sk)->groups[i]; | |
323 | } | |
4277a083 PM |
324 | tbl->listeners[i] = mask; |
325 | } | |
326 | /* this function is only called with the netlink table "grabbed", which | |
327 | * makes sure updates are visible before bind or setsockopt return. */ | |
328 | } | |
329 | ||
1da177e4 LT |
330 | static int netlink_insert(struct sock *sk, u32 pid) |
331 | { | |
332 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
333 | struct hlist_head *head; | |
334 | int err = -EADDRINUSE; | |
335 | struct sock *osk; | |
336 | struct hlist_node *node; | |
337 | int len; | |
338 | ||
339 | netlink_table_grab(); | |
340 | head = nl_pid_hashfn(hash, pid); | |
341 | len = 0; | |
342 | sk_for_each(osk, node, head) { | |
343 | if (nlk_sk(osk)->pid == pid) | |
344 | break; | |
345 | len++; | |
346 | } | |
347 | if (node) | |
348 | goto err; | |
349 | ||
350 | err = -EBUSY; | |
351 | if (nlk_sk(sk)->pid) | |
352 | goto err; | |
353 | ||
354 | err = -ENOMEM; | |
355 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
356 | goto err; | |
357 | ||
358 | if (len && nl_pid_hash_dilute(hash, len)) | |
359 | head = nl_pid_hashfn(hash, pid); | |
360 | hash->entries++; | |
361 | nlk_sk(sk)->pid = pid; | |
362 | sk_add_node(sk, head); | |
363 | err = 0; | |
364 | ||
365 | err: | |
366 | netlink_table_ungrab(); | |
367 | return err; | |
368 | } | |
369 | ||
370 | static void netlink_remove(struct sock *sk) | |
371 | { | |
372 | netlink_table_grab(); | |
d470e3b4 DM |
373 | if (sk_del_node_init(sk)) |
374 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 375 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
376 | __sk_del_bind_node(sk); |
377 | netlink_table_ungrab(); | |
378 | } | |
379 | ||
380 | static struct proto netlink_proto = { | |
381 | .name = "NETLINK", | |
382 | .owner = THIS_MODULE, | |
383 | .obj_size = sizeof(struct netlink_sock), | |
384 | }; | |
385 | ||
af65bdfc PM |
386 | static int __netlink_create(struct socket *sock, struct mutex *cb_mutex, |
387 | int protocol) | |
1da177e4 LT |
388 | { |
389 | struct sock *sk; | |
390 | struct netlink_sock *nlk; | |
ab33a171 PM |
391 | |
392 | sock->ops = &netlink_ops; | |
393 | ||
394 | sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1); | |
395 | if (!sk) | |
396 | return -ENOMEM; | |
397 | ||
398 | sock_init_data(sock, sk); | |
399 | ||
400 | nlk = nlk_sk(sk); | |
ffa4d721 PM |
401 | if (cb_mutex) |
402 | nlk->cb_mutex = cb_mutex; | |
403 | else { | |
404 | nlk->cb_mutex = &nlk->cb_def_mutex; | |
405 | mutex_init(nlk->cb_mutex); | |
406 | } | |
ab33a171 PM |
407 | init_waitqueue_head(&nlk->wait); |
408 | ||
409 | sk->sk_destruct = netlink_sock_destruct; | |
410 | sk->sk_protocol = protocol; | |
411 | return 0; | |
412 | } | |
413 | ||
414 | static int netlink_create(struct socket *sock, int protocol) | |
415 | { | |
416 | struct module *module = NULL; | |
af65bdfc | 417 | struct mutex *cb_mutex; |
f7fa9b10 | 418 | struct netlink_sock *nlk; |
ab33a171 | 419 | int err = 0; |
1da177e4 LT |
420 | |
421 | sock->state = SS_UNCONNECTED; | |
422 | ||
423 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
424 | return -ESOCKTNOSUPPORT; | |
425 | ||
426 | if (protocol<0 || protocol >= MAX_LINKS) | |
427 | return -EPROTONOSUPPORT; | |
428 | ||
77247bbb | 429 | netlink_lock_table(); |
4fdb3bb7 | 430 | #ifdef CONFIG_KMOD |
ab33a171 | 431 | if (!nl_table[protocol].registered) { |
77247bbb | 432 | netlink_unlock_table(); |
4fdb3bb7 | 433 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 434 | netlink_lock_table(); |
4fdb3bb7 | 435 | } |
ab33a171 PM |
436 | #endif |
437 | if (nl_table[protocol].registered && | |
438 | try_module_get(nl_table[protocol].module)) | |
439 | module = nl_table[protocol].module; | |
af65bdfc | 440 | cb_mutex = nl_table[protocol].cb_mutex; |
77247bbb | 441 | netlink_unlock_table(); |
4fdb3bb7 | 442 | |
af65bdfc | 443 | if ((err = __netlink_create(sock, cb_mutex, protocol)) < 0) |
f7fa9b10 PM |
444 | goto out_module; |
445 | ||
446 | nlk = nlk_sk(sock->sk); | |
f7fa9b10 | 447 | nlk->module = module; |
ab33a171 PM |
448 | out: |
449 | return err; | |
1da177e4 | 450 | |
ab33a171 PM |
451 | out_module: |
452 | module_put(module); | |
453 | goto out; | |
1da177e4 LT |
454 | } |
455 | ||
456 | static int netlink_release(struct socket *sock) | |
457 | { | |
458 | struct sock *sk = sock->sk; | |
459 | struct netlink_sock *nlk; | |
460 | ||
461 | if (!sk) | |
462 | return 0; | |
463 | ||
464 | netlink_remove(sk); | |
ac57b3a9 | 465 | sock_orphan(sk); |
1da177e4 LT |
466 | nlk = nlk_sk(sk); |
467 | ||
3f660d66 HX |
468 | /* |
469 | * OK. Socket is unlinked, any packets that arrive now | |
470 | * will be purged. | |
471 | */ | |
1da177e4 | 472 | |
1da177e4 LT |
473 | sock->sk = NULL; |
474 | wake_up_interruptible_all(&nlk->wait); | |
475 | ||
476 | skb_queue_purge(&sk->sk_write_queue); | |
477 | ||
f7fa9b10 | 478 | if (nlk->pid && !nlk->subscriptions) { |
1da177e4 LT |
479 | struct netlink_notify n = { |
480 | .protocol = sk->sk_protocol, | |
481 | .pid = nlk->pid, | |
482 | }; | |
e041c683 AS |
483 | atomic_notifier_call_chain(&netlink_chain, |
484 | NETLINK_URELEASE, &n); | |
746fac4d | 485 | } |
4fdb3bb7 | 486 | |
5e7c001c | 487 | module_put(nlk->module); |
4fdb3bb7 | 488 | |
4277a083 | 489 | netlink_table_grab(); |
77247bbb | 490 | if (nlk->flags & NETLINK_KERNEL_SOCKET) { |
4277a083 | 491 | kfree(nl_table[sk->sk_protocol].listeners); |
77247bbb | 492 | nl_table[sk->sk_protocol].module = NULL; |
ab33a171 | 493 | nl_table[sk->sk_protocol].registered = 0; |
4277a083 PM |
494 | } else if (nlk->subscriptions) |
495 | netlink_update_listeners(sk); | |
496 | netlink_table_ungrab(); | |
77247bbb | 497 | |
f7fa9b10 PM |
498 | kfree(nlk->groups); |
499 | nlk->groups = NULL; | |
500 | ||
1da177e4 LT |
501 | sock_put(sk); |
502 | return 0; | |
503 | } | |
504 | ||
505 | static int netlink_autobind(struct socket *sock) | |
506 | { | |
507 | struct sock *sk = sock->sk; | |
508 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
509 | struct hlist_head *head; | |
510 | struct sock *osk; | |
511 | struct hlist_node *node; | |
c27bd492 | 512 | s32 pid = current->tgid; |
1da177e4 LT |
513 | int err; |
514 | static s32 rover = -4097; | |
515 | ||
516 | retry: | |
517 | cond_resched(); | |
518 | netlink_table_grab(); | |
519 | head = nl_pid_hashfn(hash, pid); | |
520 | sk_for_each(osk, node, head) { | |
521 | if (nlk_sk(osk)->pid == pid) { | |
522 | /* Bind collision, search negative pid values. */ | |
523 | pid = rover--; | |
524 | if (rover > -4097) | |
525 | rover = -4097; | |
526 | netlink_table_ungrab(); | |
527 | goto retry; | |
528 | } | |
529 | } | |
530 | netlink_table_ungrab(); | |
531 | ||
532 | err = netlink_insert(sk, pid); | |
533 | if (err == -EADDRINUSE) | |
534 | goto retry; | |
d470e3b4 DM |
535 | |
536 | /* If 2 threads race to autobind, that is fine. */ | |
537 | if (err == -EBUSY) | |
538 | err = 0; | |
539 | ||
540 | return err; | |
1da177e4 LT |
541 | } |
542 | ||
746fac4d YH |
543 | static inline int netlink_capable(struct socket *sock, unsigned int flag) |
544 | { | |
1da177e4 LT |
545 | return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) || |
546 | capable(CAP_NET_ADMIN); | |
746fac4d | 547 | } |
1da177e4 | 548 | |
f7fa9b10 PM |
549 | static void |
550 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
551 | { | |
552 | struct netlink_sock *nlk = nlk_sk(sk); | |
553 | ||
554 | if (nlk->subscriptions && !subscriptions) | |
555 | __sk_del_bind_node(sk); | |
556 | else if (!nlk->subscriptions && subscriptions) | |
557 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
558 | nlk->subscriptions = subscriptions; | |
559 | } | |
560 | ||
b4ff4f04 | 561 | static int netlink_realloc_groups(struct sock *sk) |
513c2500 PM |
562 | { |
563 | struct netlink_sock *nlk = nlk_sk(sk); | |
564 | unsigned int groups; | |
b4ff4f04 | 565 | unsigned long *new_groups; |
513c2500 PM |
566 | int err = 0; |
567 | ||
b4ff4f04 JB |
568 | netlink_table_grab(); |
569 | ||
513c2500 | 570 | groups = nl_table[sk->sk_protocol].groups; |
b4ff4f04 | 571 | if (!nl_table[sk->sk_protocol].registered) { |
513c2500 | 572 | err = -ENOENT; |
b4ff4f04 JB |
573 | goto out_unlock; |
574 | } | |
513c2500 | 575 | |
b4ff4f04 JB |
576 | if (nlk->ngroups >= groups) |
577 | goto out_unlock; | |
513c2500 | 578 | |
b4ff4f04 JB |
579 | new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC); |
580 | if (new_groups == NULL) { | |
581 | err = -ENOMEM; | |
582 | goto out_unlock; | |
583 | } | |
584 | memset((char*)new_groups + NLGRPSZ(nlk->ngroups), 0, | |
585 | NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups)); | |
586 | ||
587 | nlk->groups = new_groups; | |
513c2500 | 588 | nlk->ngroups = groups; |
b4ff4f04 JB |
589 | out_unlock: |
590 | netlink_table_ungrab(); | |
591 | return err; | |
513c2500 PM |
592 | } |
593 | ||
1da177e4 LT |
594 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len) |
595 | { | |
596 | struct sock *sk = sock->sk; | |
597 | struct netlink_sock *nlk = nlk_sk(sk); | |
598 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
599 | int err; | |
746fac4d | 600 | |
1da177e4 LT |
601 | if (nladdr->nl_family != AF_NETLINK) |
602 | return -EINVAL; | |
603 | ||
604 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 PM |
605 | if (nladdr->nl_groups) { |
606 | if (!netlink_capable(sock, NL_NONROOT_RECV)) | |
607 | return -EPERM; | |
b4ff4f04 JB |
608 | err = netlink_realloc_groups(sk); |
609 | if (err) | |
610 | return err; | |
513c2500 | 611 | } |
1da177e4 LT |
612 | |
613 | if (nlk->pid) { | |
614 | if (nladdr->nl_pid != nlk->pid) | |
615 | return -EINVAL; | |
616 | } else { | |
617 | err = nladdr->nl_pid ? | |
618 | netlink_insert(sk, nladdr->nl_pid) : | |
619 | netlink_autobind(sock); | |
620 | if (err) | |
621 | return err; | |
622 | } | |
623 | ||
513c2500 | 624 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
625 | return 0; |
626 | ||
627 | netlink_table_grab(); | |
f7fa9b10 | 628 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
629 | hweight32(nladdr->nl_groups) - |
630 | hweight32(nlk->groups[0])); | |
631 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 632 | netlink_update_listeners(sk); |
1da177e4 LT |
633 | netlink_table_ungrab(); |
634 | ||
635 | return 0; | |
636 | } | |
637 | ||
638 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
639 | int alen, int flags) | |
640 | { | |
641 | int err = 0; | |
642 | struct sock *sk = sock->sk; | |
643 | struct netlink_sock *nlk = nlk_sk(sk); | |
644 | struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr; | |
645 | ||
646 | if (addr->sa_family == AF_UNSPEC) { | |
647 | sk->sk_state = NETLINK_UNCONNECTED; | |
648 | nlk->dst_pid = 0; | |
d629b836 | 649 | nlk->dst_group = 0; |
1da177e4 LT |
650 | return 0; |
651 | } | |
652 | if (addr->sa_family != AF_NETLINK) | |
653 | return -EINVAL; | |
654 | ||
655 | /* Only superuser is allowed to send multicasts */ | |
656 | if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND)) | |
657 | return -EPERM; | |
658 | ||
659 | if (!nlk->pid) | |
660 | err = netlink_autobind(sock); | |
661 | ||
662 | if (err == 0) { | |
663 | sk->sk_state = NETLINK_CONNECTED; | |
664 | nlk->dst_pid = nladdr->nl_pid; | |
d629b836 | 665 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
666 | } |
667 | ||
668 | return err; | |
669 | } | |
670 | ||
671 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer) | |
672 | { | |
673 | struct sock *sk = sock->sk; | |
674 | struct netlink_sock *nlk = nlk_sk(sk); | |
675 | struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr; | |
746fac4d | 676 | |
1da177e4 LT |
677 | nladdr->nl_family = AF_NETLINK; |
678 | nladdr->nl_pad = 0; | |
679 | *addr_len = sizeof(*nladdr); | |
680 | ||
681 | if (peer) { | |
682 | nladdr->nl_pid = nlk->dst_pid; | |
d629b836 | 683 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 LT |
684 | } else { |
685 | nladdr->nl_pid = nlk->pid; | |
513c2500 | 686 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
687 | } |
688 | return 0; | |
689 | } | |
690 | ||
691 | static void netlink_overrun(struct sock *sk) | |
692 | { | |
693 | if (!test_and_set_bit(0, &nlk_sk(sk)->state)) { | |
694 | sk->sk_err = ENOBUFS; | |
695 | sk->sk_error_report(sk); | |
696 | } | |
697 | } | |
698 | ||
699 | static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid) | |
700 | { | |
701 | int protocol = ssk->sk_protocol; | |
702 | struct sock *sock; | |
703 | struct netlink_sock *nlk; | |
704 | ||
705 | sock = netlink_lookup(protocol, pid); | |
706 | if (!sock) | |
707 | return ERR_PTR(-ECONNREFUSED); | |
708 | ||
709 | /* Don't bother queuing skb if kernel socket has no input function */ | |
710 | nlk = nlk_sk(sock); | |
711 | if ((nlk->pid == 0 && !nlk->data_ready) || | |
712 | (sock->sk_state == NETLINK_CONNECTED && | |
713 | nlk->dst_pid != nlk_sk(ssk)->pid)) { | |
714 | sock_put(sock); | |
715 | return ERR_PTR(-ECONNREFUSED); | |
716 | } | |
717 | return sock; | |
718 | } | |
719 | ||
720 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
721 | { | |
6db5fc5d | 722 | struct inode *inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
723 | struct sock *sock; |
724 | ||
725 | if (!S_ISSOCK(inode->i_mode)) | |
726 | return ERR_PTR(-ENOTSOCK); | |
727 | ||
728 | sock = SOCKET_I(inode)->sk; | |
729 | if (sock->sk_family != AF_NETLINK) | |
730 | return ERR_PTR(-EINVAL); | |
731 | ||
732 | sock_hold(sock); | |
733 | return sock; | |
734 | } | |
735 | ||
736 | /* | |
737 | * Attach a skb to a netlink socket. | |
738 | * The caller must hold a reference to the destination socket. On error, the | |
739 | * reference is dropped. The skb is not send to the destination, just all | |
740 | * all error checks are performed and memory in the queue is reserved. | |
741 | * Return values: | |
742 | * < 0: error. skb freed, reference to sock dropped. | |
743 | * 0: continue | |
744 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
745 | */ | |
a70ea994 AK |
746 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock, |
747 | long timeo, struct sock *ssk) | |
1da177e4 LT |
748 | { |
749 | struct netlink_sock *nlk; | |
750 | ||
751 | nlk = nlk_sk(sk); | |
752 | ||
753 | if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
754 | test_bit(0, &nlk->state)) { | |
755 | DECLARE_WAITQUEUE(wait, current); | |
756 | if (!timeo) { | |
a70ea994 | 757 | if (!ssk || nlk_sk(ssk)->pid == 0) |
1da177e4 LT |
758 | netlink_overrun(sk); |
759 | sock_put(sk); | |
760 | kfree_skb(skb); | |
761 | return -EAGAIN; | |
762 | } | |
763 | ||
764 | __set_current_state(TASK_INTERRUPTIBLE); | |
765 | add_wait_queue(&nlk->wait, &wait); | |
766 | ||
767 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
768 | test_bit(0, &nlk->state)) && | |
769 | !sock_flag(sk, SOCK_DEAD)) | |
770 | timeo = schedule_timeout(timeo); | |
771 | ||
772 | __set_current_state(TASK_RUNNING); | |
773 | remove_wait_queue(&nlk->wait, &wait); | |
774 | sock_put(sk); | |
775 | ||
776 | if (signal_pending(current)) { | |
777 | kfree_skb(skb); | |
778 | return sock_intr_errno(timeo); | |
779 | } | |
780 | return 1; | |
781 | } | |
782 | skb_set_owner_r(skb, sk); | |
783 | return 0; | |
784 | } | |
785 | ||
786 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol) | |
787 | { | |
1da177e4 LT |
788 | int len = skb->len; |
789 | ||
1da177e4 LT |
790 | skb_queue_tail(&sk->sk_receive_queue, skb); |
791 | sk->sk_data_ready(sk, len); | |
792 | sock_put(sk); | |
793 | return len; | |
794 | } | |
795 | ||
796 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
797 | { | |
798 | kfree_skb(skb); | |
799 | sock_put(sk); | |
800 | } | |
801 | ||
37da647d | 802 | static inline struct sk_buff *netlink_trim(struct sk_buff *skb, |
dd0fc66f | 803 | gfp_t allocation) |
1da177e4 LT |
804 | { |
805 | int delta; | |
806 | ||
807 | skb_orphan(skb); | |
808 | ||
4305b541 | 809 | delta = skb->end - skb->tail; |
1da177e4 LT |
810 | if (delta * 2 < skb->truesize) |
811 | return skb; | |
812 | ||
813 | if (skb_shared(skb)) { | |
814 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
815 | if (!nskb) | |
816 | return skb; | |
817 | kfree_skb(skb); | |
818 | skb = nskb; | |
819 | } | |
820 | ||
821 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
822 | skb->truesize -= delta; | |
823 | ||
824 | return skb; | |
825 | } | |
826 | ||
827 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock) | |
828 | { | |
829 | struct sock *sk; | |
830 | int err; | |
831 | long timeo; | |
832 | ||
833 | skb = netlink_trim(skb, gfp_any()); | |
834 | ||
835 | timeo = sock_sndtimeo(ssk, nonblock); | |
836 | retry: | |
837 | sk = netlink_getsockbypid(ssk, pid); | |
838 | if (IS_ERR(sk)) { | |
839 | kfree_skb(skb); | |
840 | return PTR_ERR(sk); | |
841 | } | |
a70ea994 | 842 | err = netlink_attachskb(sk, skb, nonblock, timeo, ssk); |
1da177e4 LT |
843 | if (err == 1) |
844 | goto retry; | |
845 | if (err) | |
846 | return err; | |
847 | ||
848 | return netlink_sendskb(sk, skb, ssk->sk_protocol); | |
849 | } | |
850 | ||
4277a083 PM |
851 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
852 | { | |
853 | int res = 0; | |
b4ff4f04 | 854 | unsigned long *listeners; |
4277a083 PM |
855 | |
856 | BUG_ON(!(nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET)); | |
b4ff4f04 JB |
857 | |
858 | rcu_read_lock(); | |
859 | listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners); | |
860 | ||
4277a083 | 861 | if (group - 1 < nl_table[sk->sk_protocol].groups) |
b4ff4f04 JB |
862 | res = test_bit(group - 1, listeners); |
863 | ||
864 | rcu_read_unlock(); | |
865 | ||
4277a083 PM |
866 | return res; |
867 | } | |
868 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
869 | ||
1da177e4 LT |
870 | static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
871 | { | |
872 | struct netlink_sock *nlk = nlk_sk(sk); | |
873 | ||
874 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
875 | !test_bit(0, &nlk->state)) { | |
876 | skb_set_owner_r(skb, sk); | |
877 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
878 | sk->sk_data_ready(sk, skb->len); | |
879 | return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf; | |
880 | } | |
881 | return -1; | |
882 | } | |
883 | ||
884 | struct netlink_broadcast_data { | |
885 | struct sock *exclude_sk; | |
886 | u32 pid; | |
887 | u32 group; | |
888 | int failure; | |
889 | int congested; | |
890 | int delivered; | |
7d877f3b | 891 | gfp_t allocation; |
1da177e4 LT |
892 | struct sk_buff *skb, *skb2; |
893 | }; | |
894 | ||
895 | static inline int do_one_broadcast(struct sock *sk, | |
896 | struct netlink_broadcast_data *p) | |
897 | { | |
898 | struct netlink_sock *nlk = nlk_sk(sk); | |
899 | int val; | |
900 | ||
901 | if (p->exclude_sk == sk) | |
902 | goto out; | |
903 | ||
f7fa9b10 PM |
904 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
905 | !test_bit(p->group - 1, nlk->groups)) | |
1da177e4 LT |
906 | goto out; |
907 | ||
908 | if (p->failure) { | |
909 | netlink_overrun(sk); | |
910 | goto out; | |
911 | } | |
912 | ||
913 | sock_hold(sk); | |
914 | if (p->skb2 == NULL) { | |
68acc024 | 915 | if (skb_shared(p->skb)) { |
1da177e4 LT |
916 | p->skb2 = skb_clone(p->skb, p->allocation); |
917 | } else { | |
68acc024 TC |
918 | p->skb2 = skb_get(p->skb); |
919 | /* | |
920 | * skb ownership may have been set when | |
921 | * delivered to a previous socket. | |
922 | */ | |
923 | skb_orphan(p->skb2); | |
1da177e4 LT |
924 | } |
925 | } | |
926 | if (p->skb2 == NULL) { | |
927 | netlink_overrun(sk); | |
928 | /* Clone failed. Notify ALL listeners. */ | |
929 | p->failure = 1; | |
930 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { | |
931 | netlink_overrun(sk); | |
932 | } else { | |
933 | p->congested |= val; | |
934 | p->delivered = 1; | |
935 | p->skb2 = NULL; | |
936 | } | |
937 | sock_put(sk); | |
938 | ||
939 | out: | |
940 | return 0; | |
941 | } | |
942 | ||
943 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid, | |
dd0fc66f | 944 | u32 group, gfp_t allocation) |
1da177e4 LT |
945 | { |
946 | struct netlink_broadcast_data info; | |
947 | struct hlist_node *node; | |
948 | struct sock *sk; | |
949 | ||
950 | skb = netlink_trim(skb, allocation); | |
951 | ||
952 | info.exclude_sk = ssk; | |
953 | info.pid = pid; | |
954 | info.group = group; | |
955 | info.failure = 0; | |
956 | info.congested = 0; | |
957 | info.delivered = 0; | |
958 | info.allocation = allocation; | |
959 | info.skb = skb; | |
960 | info.skb2 = NULL; | |
961 | ||
962 | /* While we sleep in clone, do not allow to change socket list */ | |
963 | ||
964 | netlink_lock_table(); | |
965 | ||
966 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
967 | do_one_broadcast(sk, &info); | |
968 | ||
aa1c6a6f TC |
969 | kfree_skb(skb); |
970 | ||
1da177e4 LT |
971 | netlink_unlock_table(); |
972 | ||
973 | if (info.skb2) | |
974 | kfree_skb(info.skb2); | |
1da177e4 LT |
975 | |
976 | if (info.delivered) { | |
977 | if (info.congested && (allocation & __GFP_WAIT)) | |
978 | yield(); | |
979 | return 0; | |
980 | } | |
981 | if (info.failure) | |
982 | return -ENOBUFS; | |
983 | return -ESRCH; | |
984 | } | |
985 | ||
986 | struct netlink_set_err_data { | |
987 | struct sock *exclude_sk; | |
988 | u32 pid; | |
989 | u32 group; | |
990 | int code; | |
991 | }; | |
992 | ||
993 | static inline int do_one_set_err(struct sock *sk, | |
994 | struct netlink_set_err_data *p) | |
995 | { | |
996 | struct netlink_sock *nlk = nlk_sk(sk); | |
997 | ||
998 | if (sk == p->exclude_sk) | |
999 | goto out; | |
1000 | ||
f7fa9b10 PM |
1001 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
1002 | !test_bit(p->group - 1, nlk->groups)) | |
1da177e4 LT |
1003 | goto out; |
1004 | ||
1005 | sk->sk_err = p->code; | |
1006 | sk->sk_error_report(sk); | |
1007 | out: | |
1008 | return 0; | |
1009 | } | |
1010 | ||
1011 | void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code) | |
1012 | { | |
1013 | struct netlink_set_err_data info; | |
1014 | struct hlist_node *node; | |
1015 | struct sock *sk; | |
1016 | ||
1017 | info.exclude_sk = ssk; | |
1018 | info.pid = pid; | |
1019 | info.group = group; | |
1020 | info.code = code; | |
1021 | ||
1022 | read_lock(&nl_table_lock); | |
1023 | ||
1024 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
1025 | do_one_set_err(sk, &info); | |
1026 | ||
1027 | read_unlock(&nl_table_lock); | |
1028 | } | |
1029 | ||
84659eb5 JB |
1030 | /* must be called with netlink table grabbed */ |
1031 | static void netlink_update_socket_mc(struct netlink_sock *nlk, | |
1032 | unsigned int group, | |
1033 | int is_new) | |
1034 | { | |
1035 | int old, new = !!is_new, subscriptions; | |
1036 | ||
1037 | old = test_bit(group - 1, nlk->groups); | |
1038 | subscriptions = nlk->subscriptions - old + new; | |
1039 | if (new) | |
1040 | __set_bit(group - 1, nlk->groups); | |
1041 | else | |
1042 | __clear_bit(group - 1, nlk->groups); | |
1043 | netlink_update_subscriptions(&nlk->sk, subscriptions); | |
1044 | netlink_update_listeners(&nlk->sk); | |
1045 | } | |
1046 | ||
9a4595bc | 1047 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
746fac4d | 1048 | char __user *optval, int optlen) |
9a4595bc PM |
1049 | { |
1050 | struct sock *sk = sock->sk; | |
1051 | struct netlink_sock *nlk = nlk_sk(sk); | |
eb496534 JB |
1052 | unsigned int val = 0; |
1053 | int err; | |
9a4595bc PM |
1054 | |
1055 | if (level != SOL_NETLINK) | |
1056 | return -ENOPROTOOPT; | |
1057 | ||
1058 | if (optlen >= sizeof(int) && | |
eb496534 | 1059 | get_user(val, (unsigned int __user *)optval)) |
9a4595bc PM |
1060 | return -EFAULT; |
1061 | ||
1062 | switch (optname) { | |
1063 | case NETLINK_PKTINFO: | |
1064 | if (val) | |
1065 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
1066 | else | |
1067 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
1068 | err = 0; | |
1069 | break; | |
1070 | case NETLINK_ADD_MEMBERSHIP: | |
1071 | case NETLINK_DROP_MEMBERSHIP: { | |
9a4595bc PM |
1072 | if (!netlink_capable(sock, NL_NONROOT_RECV)) |
1073 | return -EPERM; | |
b4ff4f04 JB |
1074 | err = netlink_realloc_groups(sk); |
1075 | if (err) | |
1076 | return err; | |
9a4595bc PM |
1077 | if (!val || val - 1 >= nlk->ngroups) |
1078 | return -EINVAL; | |
1079 | netlink_table_grab(); | |
84659eb5 JB |
1080 | netlink_update_socket_mc(nlk, val, |
1081 | optname == NETLINK_ADD_MEMBERSHIP); | |
9a4595bc PM |
1082 | netlink_table_ungrab(); |
1083 | err = 0; | |
1084 | break; | |
1085 | } | |
1086 | default: | |
1087 | err = -ENOPROTOOPT; | |
1088 | } | |
1089 | return err; | |
1090 | } | |
1091 | ||
1092 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1093 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1094 | { |
1095 | struct sock *sk = sock->sk; | |
1096 | struct netlink_sock *nlk = nlk_sk(sk); | |
1097 | int len, val, err; | |
1098 | ||
1099 | if (level != SOL_NETLINK) | |
1100 | return -ENOPROTOOPT; | |
1101 | ||
1102 | if (get_user(len, optlen)) | |
1103 | return -EFAULT; | |
1104 | if (len < 0) | |
1105 | return -EINVAL; | |
1106 | ||
1107 | switch (optname) { | |
1108 | case NETLINK_PKTINFO: | |
1109 | if (len < sizeof(int)) | |
1110 | return -EINVAL; | |
1111 | len = sizeof(int); | |
1112 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
1113 | if (put_user(len, optlen) || |
1114 | put_user(val, optval)) | |
1115 | return -EFAULT; | |
9a4595bc PM |
1116 | err = 0; |
1117 | break; | |
1118 | default: | |
1119 | err = -ENOPROTOOPT; | |
1120 | } | |
1121 | return err; | |
1122 | } | |
1123 | ||
1124 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1125 | { | |
1126 | struct nl_pktinfo info; | |
1127 | ||
1128 | info.group = NETLINK_CB(skb).dst_group; | |
1129 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1130 | } | |
1131 | ||
1da177e4 LT |
1132 | static inline void netlink_rcv_wake(struct sock *sk) |
1133 | { | |
1134 | struct netlink_sock *nlk = nlk_sk(sk); | |
1135 | ||
b03efcfb | 1136 | if (skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
1137 | clear_bit(0, &nlk->state); |
1138 | if (!test_bit(0, &nlk->state)) | |
1139 | wake_up_interruptible(&nlk->wait); | |
1140 | } | |
1141 | ||
1142 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1143 | struct msghdr *msg, size_t len) | |
1144 | { | |
1145 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1146 | struct sock *sk = sock->sk; | |
1147 | struct netlink_sock *nlk = nlk_sk(sk); | |
1148 | struct sockaddr_nl *addr=msg->msg_name; | |
1149 | u32 dst_pid; | |
d629b836 | 1150 | u32 dst_group; |
1da177e4 LT |
1151 | struct sk_buff *skb; |
1152 | int err; | |
1153 | struct scm_cookie scm; | |
1154 | ||
1155 | if (msg->msg_flags&MSG_OOB) | |
1156 | return -EOPNOTSUPP; | |
1157 | ||
1158 | if (NULL == siocb->scm) | |
1159 | siocb->scm = &scm; | |
1160 | err = scm_send(sock, msg, siocb->scm); | |
1161 | if (err < 0) | |
1162 | return err; | |
1163 | ||
1164 | if (msg->msg_namelen) { | |
1165 | if (addr->nl_family != AF_NETLINK) | |
1166 | return -EINVAL; | |
1167 | dst_pid = addr->nl_pid; | |
d629b836 PM |
1168 | dst_group = ffs(addr->nl_groups); |
1169 | if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND)) | |
1da177e4 LT |
1170 | return -EPERM; |
1171 | } else { | |
1172 | dst_pid = nlk->dst_pid; | |
d629b836 | 1173 | dst_group = nlk->dst_group; |
1da177e4 LT |
1174 | } |
1175 | ||
1176 | if (!nlk->pid) { | |
1177 | err = netlink_autobind(sock); | |
1178 | if (err) | |
1179 | goto out; | |
1180 | } | |
1181 | ||
1182 | err = -EMSGSIZE; | |
1183 | if (len > sk->sk_sndbuf - 32) | |
1184 | goto out; | |
1185 | err = -ENOBUFS; | |
339bf98f | 1186 | skb = alloc_skb(len, GFP_KERNEL); |
1da177e4 LT |
1187 | if (skb==NULL) |
1188 | goto out; | |
1189 | ||
1190 | NETLINK_CB(skb).pid = nlk->pid; | |
d629b836 | 1191 | NETLINK_CB(skb).dst_group = dst_group; |
c94c257c | 1192 | NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context); |
e7c34970 | 1193 | selinux_get_task_sid(current, &(NETLINK_CB(skb).sid)); |
1da177e4 LT |
1194 | memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred)); |
1195 | ||
1196 | /* What can I do? Netlink is asynchronous, so that | |
1197 | we will have to save current capabilities to | |
1198 | check them, when this message will be delivered | |
1199 | to corresponding kernel module. --ANK (980802) | |
1200 | */ | |
1201 | ||
1202 | err = -EFAULT; | |
1203 | if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) { | |
1204 | kfree_skb(skb); | |
1205 | goto out; | |
1206 | } | |
1207 | ||
1208 | err = security_netlink_send(sk, skb); | |
1209 | if (err) { | |
1210 | kfree_skb(skb); | |
1211 | goto out; | |
1212 | } | |
1213 | ||
d629b836 | 1214 | if (dst_group) { |
1da177e4 | 1215 | atomic_inc(&skb->users); |
d629b836 | 1216 | netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL); |
1da177e4 LT |
1217 | } |
1218 | err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT); | |
1219 | ||
1220 | out: | |
1221 | return err; | |
1222 | } | |
1223 | ||
1224 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
1225 | struct msghdr *msg, size_t len, | |
1226 | int flags) | |
1227 | { | |
1228 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1229 | struct scm_cookie scm; | |
1230 | struct sock *sk = sock->sk; | |
1231 | struct netlink_sock *nlk = nlk_sk(sk); | |
1232 | int noblock = flags&MSG_DONTWAIT; | |
1233 | size_t copied; | |
1234 | struct sk_buff *skb; | |
1235 | int err; | |
1236 | ||
1237 | if (flags&MSG_OOB) | |
1238 | return -EOPNOTSUPP; | |
1239 | ||
1240 | copied = 0; | |
1241 | ||
1242 | skb = skb_recv_datagram(sk,flags,noblock,&err); | |
1243 | if (skb==NULL) | |
1244 | goto out; | |
1245 | ||
1246 | msg->msg_namelen = 0; | |
1247 | ||
1248 | copied = skb->len; | |
1249 | if (len < copied) { | |
1250 | msg->msg_flags |= MSG_TRUNC; | |
1251 | copied = len; | |
1252 | } | |
1253 | ||
badff6d0 | 1254 | skb_reset_transport_header(skb); |
1da177e4 LT |
1255 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); |
1256 | ||
1257 | if (msg->msg_name) { | |
1258 | struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name; | |
1259 | addr->nl_family = AF_NETLINK; | |
1260 | addr->nl_pad = 0; | |
1261 | addr->nl_pid = NETLINK_CB(skb).pid; | |
d629b836 | 1262 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1263 | msg->msg_namelen = sizeof(*addr); |
1264 | } | |
1265 | ||
cc9a06cd PM |
1266 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
1267 | netlink_cmsg_recv_pktinfo(msg, skb); | |
1268 | ||
1da177e4 LT |
1269 | if (NULL == siocb->scm) { |
1270 | memset(&scm, 0, sizeof(scm)); | |
1271 | siocb->scm = &scm; | |
1272 | } | |
1273 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
188ccb55 PM |
1274 | if (flags & MSG_TRUNC) |
1275 | copied = skb->len; | |
1da177e4 LT |
1276 | skb_free_datagram(sk, skb); |
1277 | ||
1278 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) | |
1279 | netlink_dump(sk); | |
1280 | ||
1281 | scm_recv(sock, msg, siocb->scm, flags); | |
1da177e4 LT |
1282 | out: |
1283 | netlink_rcv_wake(sk); | |
1284 | return err ? : copied; | |
1285 | } | |
1286 | ||
1287 | static void netlink_data_ready(struct sock *sk, int len) | |
1288 | { | |
1289 | struct netlink_sock *nlk = nlk_sk(sk); | |
1290 | ||
1291 | if (nlk->data_ready) | |
1292 | nlk->data_ready(sk, len); | |
1293 | netlink_rcv_wake(sk); | |
1294 | } | |
1295 | ||
1296 | /* | |
746fac4d | 1297 | * We export these functions to other modules. They provide a |
1da177e4 LT |
1298 | * complete set of kernel non-blocking support for message |
1299 | * queueing. | |
1300 | */ | |
1301 | ||
1302 | struct sock * | |
06628607 | 1303 | netlink_kernel_create(int unit, unsigned int groups, |
746fac4d | 1304 | void (*input)(struct sock *sk, int len), |
af65bdfc | 1305 | struct mutex *cb_mutex, struct module *module) |
1da177e4 LT |
1306 | { |
1307 | struct socket *sock; | |
1308 | struct sock *sk; | |
77247bbb | 1309 | struct netlink_sock *nlk; |
4277a083 | 1310 | unsigned long *listeners = NULL; |
1da177e4 | 1311 | |
fab2caf6 | 1312 | BUG_ON(!nl_table); |
1da177e4 LT |
1313 | |
1314 | if (unit<0 || unit>=MAX_LINKS) | |
1315 | return NULL; | |
1316 | ||
1317 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1318 | return NULL; | |
1319 | ||
af65bdfc | 1320 | if (__netlink_create(sock, cb_mutex, unit) < 0) |
77247bbb | 1321 | goto out_sock_release; |
4fdb3bb7 | 1322 | |
4277a083 PM |
1323 | if (groups < 32) |
1324 | groups = 32; | |
1325 | ||
1326 | listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL); | |
1327 | if (!listeners) | |
1328 | goto out_sock_release; | |
1329 | ||
1da177e4 LT |
1330 | sk = sock->sk; |
1331 | sk->sk_data_ready = netlink_data_ready; | |
1332 | if (input) | |
1333 | nlk_sk(sk)->data_ready = input; | |
1334 | ||
77247bbb PM |
1335 | if (netlink_insert(sk, 0)) |
1336 | goto out_sock_release; | |
4fdb3bb7 | 1337 | |
77247bbb PM |
1338 | nlk = nlk_sk(sk); |
1339 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 1340 | |
4fdb3bb7 | 1341 | netlink_table_grab(); |
4277a083 PM |
1342 | nl_table[unit].groups = groups; |
1343 | nl_table[unit].listeners = listeners; | |
af65bdfc | 1344 | nl_table[unit].cb_mutex = cb_mutex; |
77247bbb | 1345 | nl_table[unit].module = module; |
ab33a171 | 1346 | nl_table[unit].registered = 1; |
4fdb3bb7 | 1347 | netlink_table_ungrab(); |
77247bbb PM |
1348 | |
1349 | return sk; | |
1350 | ||
4fdb3bb7 | 1351 | out_sock_release: |
4277a083 | 1352 | kfree(listeners); |
4fdb3bb7 | 1353 | sock_release(sock); |
77247bbb | 1354 | return NULL; |
1da177e4 LT |
1355 | } |
1356 | ||
b4ff4f04 JB |
1357 | /** |
1358 | * netlink_change_ngroups - change number of multicast groups | |
1359 | * | |
1360 | * This changes the number of multicast groups that are available | |
1361 | * on a certain netlink family. Note that it is not possible to | |
84659eb5 JB |
1362 | * change the number of groups to below 32. Also note that it does |
1363 | * not implicitly call netlink_clear_multicast_users() when the | |
1364 | * number of groups is reduced. | |
b4ff4f04 JB |
1365 | * |
1366 | * @sk: The kernel netlink socket, as returned by netlink_kernel_create(). | |
1367 | * @groups: The new number of groups. | |
1368 | */ | |
1369 | int netlink_change_ngroups(struct sock *sk, unsigned int groups) | |
1370 | { | |
1371 | unsigned long *listeners, *old = NULL; | |
1372 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
1373 | int err = 0; | |
1374 | ||
1375 | if (groups < 32) | |
1376 | groups = 32; | |
1377 | ||
1378 | netlink_table_grab(); | |
1379 | if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) { | |
1380 | listeners = kzalloc(NLGRPSZ(groups), GFP_ATOMIC); | |
1381 | if (!listeners) { | |
1382 | err = -ENOMEM; | |
1383 | goto out_ungrab; | |
1384 | } | |
1385 | old = tbl->listeners; | |
1386 | memcpy(listeners, old, NLGRPSZ(tbl->groups)); | |
1387 | rcu_assign_pointer(tbl->listeners, listeners); | |
1388 | } | |
1389 | tbl->groups = groups; | |
1390 | ||
1391 | out_ungrab: | |
1392 | netlink_table_ungrab(); | |
1393 | synchronize_rcu(); | |
1394 | kfree(old); | |
1395 | return err; | |
1396 | } | |
1397 | EXPORT_SYMBOL(netlink_change_ngroups); | |
1398 | ||
84659eb5 JB |
1399 | /** |
1400 | * netlink_clear_multicast_users - kick off multicast listeners | |
1401 | * | |
1402 | * This function removes all listeners from the given group. | |
1403 | * @ksk: The kernel netlink socket, as returned by | |
1404 | * netlink_kernel_create(). | |
1405 | * @group: The multicast group to clear. | |
1406 | */ | |
1407 | void netlink_clear_multicast_users(struct sock *ksk, unsigned int group) | |
1408 | { | |
1409 | struct sock *sk; | |
1410 | struct hlist_node *node; | |
1411 | struct netlink_table *tbl = &nl_table[ksk->sk_protocol]; | |
1412 | ||
1413 | netlink_table_grab(); | |
1414 | ||
1415 | sk_for_each_bound(sk, node, &tbl->mc_list) | |
1416 | netlink_update_socket_mc(nlk_sk(sk), group, 0); | |
1417 | ||
1418 | netlink_table_ungrab(); | |
1419 | } | |
1420 | EXPORT_SYMBOL(netlink_clear_multicast_users); | |
1421 | ||
1da177e4 | 1422 | void netlink_set_nonroot(int protocol, unsigned int flags) |
746fac4d YH |
1423 | { |
1424 | if ((unsigned int)protocol < MAX_LINKS) | |
1da177e4 | 1425 | nl_table[protocol].nl_nonroot = flags; |
746fac4d | 1426 | } |
1da177e4 LT |
1427 | |
1428 | static void netlink_destroy_callback(struct netlink_callback *cb) | |
1429 | { | |
1430 | if (cb->skb) | |
1431 | kfree_skb(cb->skb); | |
1432 | kfree(cb); | |
1433 | } | |
1434 | ||
1435 | /* | |
1436 | * It looks a bit ugly. | |
1437 | * It would be better to create kernel thread. | |
1438 | */ | |
1439 | ||
1440 | static int netlink_dump(struct sock *sk) | |
1441 | { | |
1442 | struct netlink_sock *nlk = nlk_sk(sk); | |
1443 | struct netlink_callback *cb; | |
1444 | struct sk_buff *skb; | |
1445 | struct nlmsghdr *nlh; | |
bf8b79e4 | 1446 | int len, err = -ENOBUFS; |
746fac4d | 1447 | |
1da177e4 LT |
1448 | skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL); |
1449 | if (!skb) | |
bf8b79e4 | 1450 | goto errout; |
1da177e4 | 1451 | |
af65bdfc | 1452 | mutex_lock(nlk->cb_mutex); |
1da177e4 LT |
1453 | |
1454 | cb = nlk->cb; | |
1455 | if (cb == NULL) { | |
bf8b79e4 TG |
1456 | err = -EINVAL; |
1457 | goto errout_skb; | |
1da177e4 LT |
1458 | } |
1459 | ||
1460 | len = cb->dump(skb, cb); | |
1461 | ||
1462 | if (len > 0) { | |
af65bdfc | 1463 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1464 | skb_queue_tail(&sk->sk_receive_queue, skb); |
1465 | sk->sk_data_ready(sk, len); | |
1466 | return 0; | |
1467 | } | |
1468 | ||
bf8b79e4 TG |
1469 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
1470 | if (!nlh) | |
1471 | goto errout_skb; | |
1472 | ||
1473 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); | |
1474 | ||
1da177e4 LT |
1475 | skb_queue_tail(&sk->sk_receive_queue, skb); |
1476 | sk->sk_data_ready(sk, skb->len); | |
1477 | ||
a8f74b22 TG |
1478 | if (cb->done) |
1479 | cb->done(cb); | |
1da177e4 | 1480 | nlk->cb = NULL; |
af65bdfc | 1481 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1482 | |
1483 | netlink_destroy_callback(cb); | |
1da177e4 | 1484 | return 0; |
1797754e | 1485 | |
bf8b79e4 | 1486 | errout_skb: |
af65bdfc | 1487 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 TG |
1488 | kfree_skb(skb); |
1489 | errout: | |
1490 | return err; | |
1da177e4 LT |
1491 | } |
1492 | ||
1493 | int netlink_dump_start(struct sock *ssk, struct sk_buff *skb, | |
1494 | struct nlmsghdr *nlh, | |
1495 | int (*dump)(struct sk_buff *skb, struct netlink_callback*), | |
1496 | int (*done)(struct netlink_callback*)) | |
1497 | { | |
1498 | struct netlink_callback *cb; | |
1499 | struct sock *sk; | |
1500 | struct netlink_sock *nlk; | |
1501 | ||
0da974f4 | 1502 | cb = kzalloc(sizeof(*cb), GFP_KERNEL); |
1da177e4 LT |
1503 | if (cb == NULL) |
1504 | return -ENOBUFS; | |
1505 | ||
1da177e4 LT |
1506 | cb->dump = dump; |
1507 | cb->done = done; | |
1508 | cb->nlh = nlh; | |
1509 | atomic_inc(&skb->users); | |
1510 | cb->skb = skb; | |
1511 | ||
1512 | sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid); | |
1513 | if (sk == NULL) { | |
1514 | netlink_destroy_callback(cb); | |
1515 | return -ECONNREFUSED; | |
1516 | } | |
1517 | nlk = nlk_sk(sk); | |
3f660d66 | 1518 | /* A dump is in progress... */ |
af65bdfc | 1519 | mutex_lock(nlk->cb_mutex); |
3f660d66 | 1520 | if (nlk->cb) { |
af65bdfc | 1521 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1522 | netlink_destroy_callback(cb); |
1523 | sock_put(sk); | |
1524 | return -EBUSY; | |
1525 | } | |
1526 | nlk->cb = cb; | |
af65bdfc | 1527 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1528 | |
1529 | netlink_dump(sk); | |
1530 | sock_put(sk); | |
c702e804 TG |
1531 | |
1532 | /* We successfully started a dump, by returning -EINTR we | |
1533 | * signal the queue mangement to interrupt processing of | |
1534 | * any netlink messages so userspace gets a chance to read | |
1535 | * the results. */ | |
1536 | return -EINTR; | |
1da177e4 LT |
1537 | } |
1538 | ||
1539 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
1540 | { | |
1541 | struct sk_buff *skb; | |
1542 | struct nlmsghdr *rep; | |
1543 | struct nlmsgerr *errmsg; | |
339bf98f | 1544 | size_t payload = sizeof(*errmsg); |
1da177e4 | 1545 | |
339bf98f TG |
1546 | /* error messages get the original request appened */ |
1547 | if (err) | |
1548 | payload += nlmsg_len(nlh); | |
1da177e4 | 1549 | |
339bf98f | 1550 | skb = nlmsg_new(payload, GFP_KERNEL); |
1da177e4 LT |
1551 | if (!skb) { |
1552 | struct sock *sk; | |
1553 | ||
1554 | sk = netlink_lookup(in_skb->sk->sk_protocol, | |
1555 | NETLINK_CB(in_skb).pid); | |
1556 | if (sk) { | |
1557 | sk->sk_err = ENOBUFS; | |
1558 | sk->sk_error_report(sk); | |
1559 | sock_put(sk); | |
1560 | } | |
1561 | return; | |
1562 | } | |
1563 | ||
1564 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, | |
1797754e | 1565 | NLMSG_ERROR, sizeof(struct nlmsgerr), 0); |
bf8b79e4 | 1566 | errmsg = nlmsg_data(rep); |
1da177e4 | 1567 | errmsg->error = err; |
bf8b79e4 | 1568 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
1da177e4 LT |
1569 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); |
1570 | } | |
1571 | ||
82ace47a | 1572 | static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 1573 | struct nlmsghdr *)) |
82ace47a | 1574 | { |
82ace47a TG |
1575 | struct nlmsghdr *nlh; |
1576 | int err; | |
1577 | ||
1578 | while (skb->len >= nlmsg_total_size(0)) { | |
b529ccf2 | 1579 | nlh = nlmsg_hdr(skb); |
d35b6856 | 1580 | err = 0; |
82ace47a | 1581 | |
ad8e4b75 | 1582 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
1583 | return 0; |
1584 | ||
d35b6856 TG |
1585 | /* Only requests are handled by the kernel */ |
1586 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
1587 | goto skip; | |
45e7ae7f TG |
1588 | |
1589 | /* Skip control messages */ | |
1590 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
1591 | goto skip; | |
d35b6856 | 1592 | |
1d00a4eb TG |
1593 | err = cb(skb, nlh); |
1594 | if (err == -EINTR) { | |
1595 | /* Not an error, but we interrupt processing */ | |
1596 | netlink_queue_skip(nlh, skb); | |
1597 | return err; | |
d35b6856 TG |
1598 | } |
1599 | skip: | |
1600 | if (nlh->nlmsg_flags & NLM_F_ACK || err) | |
82ace47a | 1601 | netlink_ack(skb, nlh, err); |
82ace47a | 1602 | |
bf8b79e4 | 1603 | netlink_queue_skip(nlh, skb); |
82ace47a TG |
1604 | } |
1605 | ||
1606 | return 0; | |
1607 | } | |
1608 | ||
1609 | /** | |
1610 | * nelink_run_queue - Process netlink receive queue. | |
1611 | * @sk: Netlink socket containing the queue | |
1612 | * @qlen: Place to store queue length upon entry | |
1613 | * @cb: Callback function invoked for each netlink message found | |
1614 | * | |
1615 | * Processes as much as there was in the queue upon entry and invokes | |
1616 | * a callback function for each netlink message found. The callback | |
1617 | * function may refuse a message by returning a negative error code | |
1618 | * but setting the error pointer to 0 in which case this function | |
1619 | * returns with a qlen != 0. | |
1620 | * | |
1621 | * qlen must be initialized to 0 before the initial entry, afterwards | |
1622 | * the function may be called repeatedly until qlen reaches 0. | |
1d00a4eb TG |
1623 | * |
1624 | * The callback function may return -EINTR to signal that processing | |
1625 | * of netlink messages shall be interrupted. In this case the message | |
1626 | * currently being processed will NOT be requeued onto the receive | |
1627 | * queue. | |
82ace47a TG |
1628 | */ |
1629 | void netlink_run_queue(struct sock *sk, unsigned int *qlen, | |
1d00a4eb | 1630 | int (*cb)(struct sk_buff *, struct nlmsghdr *)) |
82ace47a TG |
1631 | { |
1632 | struct sk_buff *skb; | |
1633 | ||
1634 | if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue)) | |
1635 | *qlen = skb_queue_len(&sk->sk_receive_queue); | |
1636 | ||
1637 | for (; *qlen; (*qlen)--) { | |
1638 | skb = skb_dequeue(&sk->sk_receive_queue); | |
1639 | if (netlink_rcv_skb(skb, cb)) { | |
1640 | if (skb->len) | |
1641 | skb_queue_head(&sk->sk_receive_queue, skb); | |
1642 | else { | |
1643 | kfree_skb(skb); | |
1644 | (*qlen)--; | |
1645 | } | |
1646 | break; | |
1647 | } | |
1648 | ||
1649 | kfree_skb(skb); | |
1650 | } | |
1651 | } | |
1652 | ||
1653 | /** | |
1654 | * netlink_queue_skip - Skip netlink message while processing queue. | |
1655 | * @nlh: Netlink message to be skipped | |
1656 | * @skb: Socket buffer containing the netlink messages. | |
1657 | * | |
1658 | * Pulls the given netlink message off the socket buffer so the next | |
1659 | * call to netlink_queue_run() will not reconsider the message. | |
1660 | */ | |
42bad1da | 1661 | static void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb) |
82ace47a TG |
1662 | { |
1663 | int msglen = NLMSG_ALIGN(nlh->nlmsg_len); | |
1664 | ||
1665 | if (msglen > skb->len) | |
1666 | msglen = skb->len; | |
1667 | ||
1668 | skb_pull(skb, msglen); | |
1669 | } | |
1da177e4 | 1670 | |
d387f6ad TG |
1671 | /** |
1672 | * nlmsg_notify - send a notification netlink message | |
1673 | * @sk: netlink socket to use | |
1674 | * @skb: notification message | |
1675 | * @pid: destination netlink pid for reports or 0 | |
1676 | * @group: destination multicast group or 0 | |
1677 | * @report: 1 to report back, 0 to disable | |
1678 | * @flags: allocation flags | |
1679 | */ | |
1680 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid, | |
1681 | unsigned int group, int report, gfp_t flags) | |
1682 | { | |
1683 | int err = 0; | |
1684 | ||
1685 | if (group) { | |
1686 | int exclude_pid = 0; | |
1687 | ||
1688 | if (report) { | |
1689 | atomic_inc(&skb->users); | |
1690 | exclude_pid = pid; | |
1691 | } | |
1692 | ||
1693 | /* errors reported via destination sk->sk_err */ | |
1694 | nlmsg_multicast(sk, skb, exclude_pid, group, flags); | |
1695 | } | |
1696 | ||
1697 | if (report) | |
1698 | err = nlmsg_unicast(sk, skb, pid); | |
1699 | ||
1700 | return err; | |
1701 | } | |
1702 | ||
1da177e4 LT |
1703 | #ifdef CONFIG_PROC_FS |
1704 | struct nl_seq_iter { | |
1705 | int link; | |
1706 | int hash_idx; | |
1707 | }; | |
1708 | ||
1709 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
1710 | { | |
1711 | struct nl_seq_iter *iter = seq->private; | |
1712 | int i, j; | |
1713 | struct sock *s; | |
1714 | struct hlist_node *node; | |
1715 | loff_t off = 0; | |
1716 | ||
1717 | for (i=0; i<MAX_LINKS; i++) { | |
1718 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1719 | ||
1720 | for (j = 0; j <= hash->mask; j++) { | |
1721 | sk_for_each(s, node, &hash->table[j]) { | |
1722 | if (off == pos) { | |
1723 | iter->link = i; | |
1724 | iter->hash_idx = j; | |
1725 | return s; | |
1726 | } | |
1727 | ++off; | |
1728 | } | |
1729 | } | |
1730 | } | |
1731 | return NULL; | |
1732 | } | |
1733 | ||
1734 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
1735 | { | |
1736 | read_lock(&nl_table_lock); | |
1737 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
1738 | } | |
1739 | ||
1740 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1741 | { | |
1742 | struct sock *s; | |
1743 | struct nl_seq_iter *iter; | |
1744 | int i, j; | |
1745 | ||
1746 | ++*pos; | |
1747 | ||
1748 | if (v == SEQ_START_TOKEN) | |
1749 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 1750 | |
1da177e4 LT |
1751 | s = sk_next(v); |
1752 | if (s) | |
1753 | return s; | |
1754 | ||
1755 | iter = seq->private; | |
1756 | i = iter->link; | |
1757 | j = iter->hash_idx + 1; | |
1758 | ||
1759 | do { | |
1760 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1761 | ||
1762 | for (; j <= hash->mask; j++) { | |
1763 | s = sk_head(&hash->table[j]); | |
1764 | if (s) { | |
1765 | iter->link = i; | |
1766 | iter->hash_idx = j; | |
1767 | return s; | |
1768 | } | |
1769 | } | |
1770 | ||
1771 | j = 0; | |
1772 | } while (++i < MAX_LINKS); | |
1773 | ||
1774 | return NULL; | |
1775 | } | |
1776 | ||
1777 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
1778 | { | |
1779 | read_unlock(&nl_table_lock); | |
1780 | } | |
1781 | ||
1782 | ||
1783 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
1784 | { | |
1785 | if (v == SEQ_START_TOKEN) | |
1786 | seq_puts(seq, | |
1787 | "sk Eth Pid Groups " | |
1788 | "Rmem Wmem Dump Locks\n"); | |
1789 | else { | |
1790 | struct sock *s = v; | |
1791 | struct netlink_sock *nlk = nlk_sk(s); | |
1792 | ||
1793 | seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n", | |
1794 | s, | |
1795 | s->sk_protocol, | |
1796 | nlk->pid, | |
513c2500 | 1797 | nlk->groups ? (u32)nlk->groups[0] : 0, |
1da177e4 LT |
1798 | atomic_read(&s->sk_rmem_alloc), |
1799 | atomic_read(&s->sk_wmem_alloc), | |
1800 | nlk->cb, | |
1801 | atomic_read(&s->sk_refcnt) | |
1802 | ); | |
1803 | ||
1804 | } | |
1805 | return 0; | |
1806 | } | |
1807 | ||
56b3d975 | 1808 | static const struct seq_operations netlink_seq_ops = { |
1da177e4 LT |
1809 | .start = netlink_seq_start, |
1810 | .next = netlink_seq_next, | |
1811 | .stop = netlink_seq_stop, | |
1812 | .show = netlink_seq_show, | |
1813 | }; | |
1814 | ||
1815 | ||
1816 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
1817 | { | |
1818 | struct seq_file *seq; | |
1819 | struct nl_seq_iter *iter; | |
1820 | int err; | |
1821 | ||
0da974f4 | 1822 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); |
1da177e4 LT |
1823 | if (!iter) |
1824 | return -ENOMEM; | |
1825 | ||
1826 | err = seq_open(file, &netlink_seq_ops); | |
1827 | if (err) { | |
1828 | kfree(iter); | |
1829 | return err; | |
1830 | } | |
1831 | ||
1da177e4 LT |
1832 | seq = file->private_data; |
1833 | seq->private = iter; | |
1834 | return 0; | |
1835 | } | |
1836 | ||
da7071d7 | 1837 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
1838 | .owner = THIS_MODULE, |
1839 | .open = netlink_seq_open, | |
1840 | .read = seq_read, | |
1841 | .llseek = seq_lseek, | |
1842 | .release = seq_release_private, | |
1843 | }; | |
1844 | ||
1845 | #endif | |
1846 | ||
1847 | int netlink_register_notifier(struct notifier_block *nb) | |
1848 | { | |
e041c683 | 1849 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 LT |
1850 | } |
1851 | ||
1852 | int netlink_unregister_notifier(struct notifier_block *nb) | |
1853 | { | |
e041c683 | 1854 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 1855 | } |
746fac4d | 1856 | |
90ddc4f0 | 1857 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
1858 | .family = PF_NETLINK, |
1859 | .owner = THIS_MODULE, | |
1860 | .release = netlink_release, | |
1861 | .bind = netlink_bind, | |
1862 | .connect = netlink_connect, | |
1863 | .socketpair = sock_no_socketpair, | |
1864 | .accept = sock_no_accept, | |
1865 | .getname = netlink_getname, | |
1866 | .poll = datagram_poll, | |
1867 | .ioctl = sock_no_ioctl, | |
1868 | .listen = sock_no_listen, | |
1869 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
1870 | .setsockopt = netlink_setsockopt, |
1871 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
1872 | .sendmsg = netlink_sendmsg, |
1873 | .recvmsg = netlink_recvmsg, | |
1874 | .mmap = sock_no_mmap, | |
1875 | .sendpage = sock_no_sendpage, | |
1876 | }; | |
1877 | ||
1878 | static struct net_proto_family netlink_family_ops = { | |
1879 | .family = PF_NETLINK, | |
1880 | .create = netlink_create, | |
1881 | .owner = THIS_MODULE, /* for consistency 8) */ | |
1882 | }; | |
1883 | ||
1da177e4 LT |
1884 | static int __init netlink_proto_init(void) |
1885 | { | |
1886 | struct sk_buff *dummy_skb; | |
1887 | int i; | |
1888 | unsigned long max; | |
1889 | unsigned int order; | |
1890 | int err = proto_register(&netlink_proto, 0); | |
1891 | ||
1892 | if (err != 0) | |
1893 | goto out; | |
1894 | ||
ef047f5e | 1895 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb)); |
1da177e4 | 1896 | |
0da974f4 | 1897 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
1898 | if (!nl_table) |
1899 | goto panic; | |
1da177e4 | 1900 | |
1da177e4 LT |
1901 | if (num_physpages >= (128 * 1024)) |
1902 | max = num_physpages >> (21 - PAGE_SHIFT); | |
1903 | else | |
1904 | max = num_physpages >> (23 - PAGE_SHIFT); | |
1905 | ||
1906 | order = get_bitmask_order(max) - 1 + PAGE_SHIFT; | |
1907 | max = (1UL << order) / sizeof(struct hlist_head); | |
1908 | order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1; | |
1909 | ||
1910 | for (i = 0; i < MAX_LINKS; i++) { | |
1911 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1912 | ||
1913 | hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table)); | |
1914 | if (!hash->table) { | |
1915 | while (i-- > 0) | |
1916 | nl_pid_hash_free(nl_table[i].hash.table, | |
1917 | 1 * sizeof(*hash->table)); | |
1918 | kfree(nl_table); | |
fab2caf6 | 1919 | goto panic; |
1da177e4 LT |
1920 | } |
1921 | memset(hash->table, 0, 1 * sizeof(*hash->table)); | |
1922 | hash->max_shift = order; | |
1923 | hash->shift = 0; | |
1924 | hash->mask = 0; | |
1925 | hash->rehash_time = jiffies; | |
1926 | } | |
1927 | ||
1928 | sock_register(&netlink_family_ops); | |
1929 | #ifdef CONFIG_PROC_FS | |
1930 | proc_net_fops_create("netlink", 0, &netlink_seq_fops); | |
1931 | #endif | |
746fac4d | 1932 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
1933 | rtnetlink_init(); |
1934 | out: | |
1935 | return err; | |
fab2caf6 AM |
1936 | panic: |
1937 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
1938 | } |
1939 | ||
1da177e4 | 1940 | core_initcall(netlink_proto_init); |
1da177e4 LT |
1941 | |
1942 | EXPORT_SYMBOL(netlink_ack); | |
82ace47a | 1943 | EXPORT_SYMBOL(netlink_run_queue); |
1da177e4 LT |
1944 | EXPORT_SYMBOL(netlink_broadcast); |
1945 | EXPORT_SYMBOL(netlink_dump_start); | |
1946 | EXPORT_SYMBOL(netlink_kernel_create); | |
1947 | EXPORT_SYMBOL(netlink_register_notifier); | |
1da177e4 LT |
1948 | EXPORT_SYMBOL(netlink_set_nonroot); |
1949 | EXPORT_SYMBOL(netlink_unicast); | |
1950 | EXPORT_SYMBOL(netlink_unregister_notifier); | |
d387f6ad | 1951 | EXPORT_SYMBOL(nlmsg_notify); |