]>
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> | |
48 | #include <linux/smp_lock.h> | |
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> |
e7c34970 | 58 | #include <linux/selinux.h> |
af65bdfc | 59 | #include <linux/mutex.h> |
54e0f520 | 60 | |
1da177e4 LT |
61 | #include <net/sock.h> |
62 | #include <net/scm.h> | |
82ace47a | 63 | #include <net/netlink.h> |
1da177e4 | 64 | |
f7fa9b10 | 65 | #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8) |
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); | |
124 | ||
125 | static DEFINE_RWLOCK(nl_table_lock); | |
126 | static atomic_t nl_table_users = ATOMIC_INIT(0); | |
127 | ||
e041c683 | 128 | static ATOMIC_NOTIFIER_HEAD(netlink_chain); |
1da177e4 | 129 | |
d629b836 PM |
130 | static u32 netlink_group_mask(u32 group) |
131 | { | |
132 | return group ? 1 << (group - 1) : 0; | |
133 | } | |
134 | ||
1da177e4 LT |
135 | static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid) |
136 | { | |
137 | return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask]; | |
138 | } | |
139 | ||
140 | static void netlink_sock_destruct(struct sock *sk) | |
141 | { | |
142 | skb_queue_purge(&sk->sk_receive_queue); | |
143 | ||
144 | if (!sock_flag(sk, SOCK_DEAD)) { | |
145 | printk("Freeing alive netlink socket %p\n", sk); | |
146 | return; | |
147 | } | |
148 | BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); | |
149 | BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); | |
150 | BUG_TRAP(!nlk_sk(sk)->cb); | |
f7fa9b10 | 151 | BUG_TRAP(!nlk_sk(sk)->groups); |
1da177e4 LT |
152 | } |
153 | ||
154 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP. | |
155 | * Look, when several writers sleep and reader wakes them up, all but one | |
156 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves | |
157 | * this, _but_ remember, it adds useless work on UP machines. | |
158 | */ | |
159 | ||
160 | static void netlink_table_grab(void) | |
161 | { | |
6abd219c | 162 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
163 | |
164 | if (atomic_read(&nl_table_users)) { | |
165 | DECLARE_WAITQUEUE(wait, current); | |
166 | ||
167 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
168 | for(;;) { | |
169 | set_current_state(TASK_UNINTERRUPTIBLE); | |
170 | if (atomic_read(&nl_table_users) == 0) | |
171 | break; | |
6abd219c | 172 | write_unlock_irq(&nl_table_lock); |
1da177e4 | 173 | schedule(); |
6abd219c | 174 | write_lock_irq(&nl_table_lock); |
1da177e4 LT |
175 | } |
176 | ||
177 | __set_current_state(TASK_RUNNING); | |
178 | remove_wait_queue(&nl_table_wait, &wait); | |
179 | } | |
180 | } | |
181 | ||
182 | static __inline__ void netlink_table_ungrab(void) | |
183 | { | |
6abd219c | 184 | write_unlock_irq(&nl_table_lock); |
1da177e4 LT |
185 | wake_up(&nl_table_wait); |
186 | } | |
187 | ||
188 | static __inline__ void | |
189 | netlink_lock_table(void) | |
190 | { | |
191 | /* read_lock() synchronizes us to netlink_table_grab */ | |
192 | ||
193 | read_lock(&nl_table_lock); | |
194 | atomic_inc(&nl_table_users); | |
195 | read_unlock(&nl_table_lock); | |
196 | } | |
197 | ||
198 | static __inline__ void | |
199 | netlink_unlock_table(void) | |
200 | { | |
201 | if (atomic_dec_and_test(&nl_table_users)) | |
202 | wake_up(&nl_table_wait); | |
203 | } | |
204 | ||
205 | static __inline__ struct sock *netlink_lookup(int protocol, u32 pid) | |
206 | { | |
207 | struct nl_pid_hash *hash = &nl_table[protocol].hash; | |
208 | struct hlist_head *head; | |
209 | struct sock *sk; | |
210 | struct hlist_node *node; | |
211 | ||
212 | read_lock(&nl_table_lock); | |
213 | head = nl_pid_hashfn(hash, pid); | |
214 | sk_for_each(sk, node, head) { | |
215 | if (nlk_sk(sk)->pid == pid) { | |
216 | sock_hold(sk); | |
217 | goto found; | |
218 | } | |
219 | } | |
220 | sk = NULL; | |
221 | found: | |
222 | read_unlock(&nl_table_lock); | |
223 | return sk; | |
224 | } | |
225 | ||
226 | static inline struct hlist_head *nl_pid_hash_alloc(size_t size) | |
227 | { | |
228 | if (size <= PAGE_SIZE) | |
229 | return kmalloc(size, GFP_ATOMIC); | |
230 | else | |
231 | return (struct hlist_head *) | |
232 | __get_free_pages(GFP_ATOMIC, get_order(size)); | |
233 | } | |
234 | ||
235 | static inline void nl_pid_hash_free(struct hlist_head *table, size_t size) | |
236 | { | |
237 | if (size <= PAGE_SIZE) | |
238 | kfree(table); | |
239 | else | |
240 | free_pages((unsigned long)table, get_order(size)); | |
241 | } | |
242 | ||
243 | static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow) | |
244 | { | |
245 | unsigned int omask, mask, shift; | |
246 | size_t osize, size; | |
247 | struct hlist_head *otable, *table; | |
248 | int i; | |
249 | ||
250 | omask = mask = hash->mask; | |
251 | osize = size = (mask + 1) * sizeof(*table); | |
252 | shift = hash->shift; | |
253 | ||
254 | if (grow) { | |
255 | if (++shift > hash->max_shift) | |
256 | return 0; | |
257 | mask = mask * 2 + 1; | |
258 | size *= 2; | |
259 | } | |
260 | ||
261 | table = nl_pid_hash_alloc(size); | |
262 | if (!table) | |
263 | return 0; | |
264 | ||
265 | memset(table, 0, size); | |
266 | otable = hash->table; | |
267 | hash->table = table; | |
268 | hash->mask = mask; | |
269 | hash->shift = shift; | |
270 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
271 | ||
272 | for (i = 0; i <= omask; i++) { | |
273 | struct sock *sk; | |
274 | struct hlist_node *node, *tmp; | |
275 | ||
276 | sk_for_each_safe(sk, node, tmp, &otable[i]) | |
277 | __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid)); | |
278 | } | |
279 | ||
280 | nl_pid_hash_free(otable, osize); | |
281 | hash->rehash_time = jiffies + 10 * 60 * HZ; | |
282 | return 1; | |
283 | } | |
284 | ||
285 | static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len) | |
286 | { | |
287 | int avg = hash->entries >> hash->shift; | |
288 | ||
289 | if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1)) | |
290 | return 1; | |
291 | ||
292 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
293 | nl_pid_hash_rehash(hash, 0); | |
294 | return 1; | |
295 | } | |
296 | ||
297 | return 0; | |
298 | } | |
299 | ||
90ddc4f0 | 300 | static const struct proto_ops netlink_ops; |
1da177e4 | 301 | |
4277a083 PM |
302 | static void |
303 | netlink_update_listeners(struct sock *sk) | |
304 | { | |
305 | struct netlink_table *tbl = &nl_table[sk->sk_protocol]; | |
306 | struct hlist_node *node; | |
307 | unsigned long mask; | |
308 | unsigned int i; | |
309 | ||
310 | for (i = 0; i < NLGRPSZ(tbl->groups)/sizeof(unsigned long); i++) { | |
311 | mask = 0; | |
312 | sk_for_each_bound(sk, node, &tbl->mc_list) | |
313 | mask |= nlk_sk(sk)->groups[i]; | |
314 | tbl->listeners[i] = mask; | |
315 | } | |
316 | /* this function is only called with the netlink table "grabbed", which | |
317 | * makes sure updates are visible before bind or setsockopt return. */ | |
318 | } | |
319 | ||
1da177e4 LT |
320 | static int netlink_insert(struct sock *sk, u32 pid) |
321 | { | |
322 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
323 | struct hlist_head *head; | |
324 | int err = -EADDRINUSE; | |
325 | struct sock *osk; | |
326 | struct hlist_node *node; | |
327 | int len; | |
328 | ||
329 | netlink_table_grab(); | |
330 | head = nl_pid_hashfn(hash, pid); | |
331 | len = 0; | |
332 | sk_for_each(osk, node, head) { | |
333 | if (nlk_sk(osk)->pid == pid) | |
334 | break; | |
335 | len++; | |
336 | } | |
337 | if (node) | |
338 | goto err; | |
339 | ||
340 | err = -EBUSY; | |
341 | if (nlk_sk(sk)->pid) | |
342 | goto err; | |
343 | ||
344 | err = -ENOMEM; | |
345 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
346 | goto err; | |
347 | ||
348 | if (len && nl_pid_hash_dilute(hash, len)) | |
349 | head = nl_pid_hashfn(hash, pid); | |
350 | hash->entries++; | |
351 | nlk_sk(sk)->pid = pid; | |
352 | sk_add_node(sk, head); | |
353 | err = 0; | |
354 | ||
355 | err: | |
356 | netlink_table_ungrab(); | |
357 | return err; | |
358 | } | |
359 | ||
360 | static void netlink_remove(struct sock *sk) | |
361 | { | |
362 | netlink_table_grab(); | |
d470e3b4 DM |
363 | if (sk_del_node_init(sk)) |
364 | nl_table[sk->sk_protocol].hash.entries--; | |
f7fa9b10 | 365 | if (nlk_sk(sk)->subscriptions) |
1da177e4 LT |
366 | __sk_del_bind_node(sk); |
367 | netlink_table_ungrab(); | |
368 | } | |
369 | ||
370 | static struct proto netlink_proto = { | |
371 | .name = "NETLINK", | |
372 | .owner = THIS_MODULE, | |
373 | .obj_size = sizeof(struct netlink_sock), | |
374 | }; | |
375 | ||
af65bdfc PM |
376 | static int __netlink_create(struct socket *sock, struct mutex *cb_mutex, |
377 | int protocol) | |
1da177e4 LT |
378 | { |
379 | struct sock *sk; | |
380 | struct netlink_sock *nlk; | |
ab33a171 PM |
381 | |
382 | sock->ops = &netlink_ops; | |
383 | ||
384 | sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1); | |
385 | if (!sk) | |
386 | return -ENOMEM; | |
387 | ||
388 | sock_init_data(sock, sk); | |
389 | ||
390 | nlk = nlk_sk(sk); | |
af65bdfc PM |
391 | nlk->cb_mutex = cb_mutex ? : &nlk->cb_def_mutex; |
392 | mutex_init(nlk->cb_mutex); | |
ab33a171 PM |
393 | init_waitqueue_head(&nlk->wait); |
394 | ||
395 | sk->sk_destruct = netlink_sock_destruct; | |
396 | sk->sk_protocol = protocol; | |
397 | return 0; | |
398 | } | |
399 | ||
400 | static int netlink_create(struct socket *sock, int protocol) | |
401 | { | |
402 | struct module *module = NULL; | |
af65bdfc | 403 | struct mutex *cb_mutex; |
f7fa9b10 | 404 | struct netlink_sock *nlk; |
ab33a171 | 405 | int err = 0; |
1da177e4 LT |
406 | |
407 | sock->state = SS_UNCONNECTED; | |
408 | ||
409 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
410 | return -ESOCKTNOSUPPORT; | |
411 | ||
412 | if (protocol<0 || protocol >= MAX_LINKS) | |
413 | return -EPROTONOSUPPORT; | |
414 | ||
77247bbb | 415 | netlink_lock_table(); |
4fdb3bb7 | 416 | #ifdef CONFIG_KMOD |
ab33a171 | 417 | if (!nl_table[protocol].registered) { |
77247bbb | 418 | netlink_unlock_table(); |
4fdb3bb7 | 419 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 420 | netlink_lock_table(); |
4fdb3bb7 | 421 | } |
ab33a171 PM |
422 | #endif |
423 | if (nl_table[protocol].registered && | |
424 | try_module_get(nl_table[protocol].module)) | |
425 | module = nl_table[protocol].module; | |
af65bdfc | 426 | cb_mutex = nl_table[protocol].cb_mutex; |
77247bbb | 427 | netlink_unlock_table(); |
4fdb3bb7 | 428 | |
af65bdfc | 429 | if ((err = __netlink_create(sock, cb_mutex, protocol)) < 0) |
f7fa9b10 PM |
430 | goto out_module; |
431 | ||
432 | nlk = nlk_sk(sock->sk); | |
f7fa9b10 | 433 | nlk->module = module; |
ab33a171 PM |
434 | out: |
435 | return err; | |
1da177e4 | 436 | |
ab33a171 PM |
437 | out_module: |
438 | module_put(module); | |
439 | goto out; | |
1da177e4 LT |
440 | } |
441 | ||
442 | static int netlink_release(struct socket *sock) | |
443 | { | |
444 | struct sock *sk = sock->sk; | |
445 | struct netlink_sock *nlk; | |
446 | ||
447 | if (!sk) | |
448 | return 0; | |
449 | ||
450 | netlink_remove(sk); | |
ac57b3a9 | 451 | sock_orphan(sk); |
1da177e4 LT |
452 | nlk = nlk_sk(sk); |
453 | ||
af65bdfc | 454 | mutex_lock(nlk->cb_mutex); |
1da177e4 | 455 | if (nlk->cb) { |
a8f74b22 TG |
456 | if (nlk->cb->done) |
457 | nlk->cb->done(nlk->cb); | |
1da177e4 LT |
458 | netlink_destroy_callback(nlk->cb); |
459 | nlk->cb = NULL; | |
1da177e4 | 460 | } |
af65bdfc | 461 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
462 | |
463 | /* OK. Socket is unlinked, and, therefore, | |
464 | no new packets will arrive */ | |
465 | ||
1da177e4 LT |
466 | sock->sk = NULL; |
467 | wake_up_interruptible_all(&nlk->wait); | |
468 | ||
469 | skb_queue_purge(&sk->sk_write_queue); | |
470 | ||
f7fa9b10 | 471 | if (nlk->pid && !nlk->subscriptions) { |
1da177e4 LT |
472 | struct netlink_notify n = { |
473 | .protocol = sk->sk_protocol, | |
474 | .pid = nlk->pid, | |
475 | }; | |
e041c683 AS |
476 | atomic_notifier_call_chain(&netlink_chain, |
477 | NETLINK_URELEASE, &n); | |
746fac4d | 478 | } |
4fdb3bb7 | 479 | |
5e7c001c | 480 | module_put(nlk->module); |
4fdb3bb7 | 481 | |
4277a083 | 482 | netlink_table_grab(); |
77247bbb | 483 | if (nlk->flags & NETLINK_KERNEL_SOCKET) { |
4277a083 | 484 | kfree(nl_table[sk->sk_protocol].listeners); |
77247bbb | 485 | nl_table[sk->sk_protocol].module = NULL; |
ab33a171 | 486 | nl_table[sk->sk_protocol].registered = 0; |
4277a083 PM |
487 | } else if (nlk->subscriptions) |
488 | netlink_update_listeners(sk); | |
489 | netlink_table_ungrab(); | |
77247bbb | 490 | |
f7fa9b10 PM |
491 | kfree(nlk->groups); |
492 | nlk->groups = NULL; | |
493 | ||
1da177e4 LT |
494 | sock_put(sk); |
495 | return 0; | |
496 | } | |
497 | ||
498 | static int netlink_autobind(struct socket *sock) | |
499 | { | |
500 | struct sock *sk = sock->sk; | |
501 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
502 | struct hlist_head *head; | |
503 | struct sock *osk; | |
504 | struct hlist_node *node; | |
c27bd492 | 505 | s32 pid = current->tgid; |
1da177e4 LT |
506 | int err; |
507 | static s32 rover = -4097; | |
508 | ||
509 | retry: | |
510 | cond_resched(); | |
511 | netlink_table_grab(); | |
512 | head = nl_pid_hashfn(hash, pid); | |
513 | sk_for_each(osk, node, head) { | |
514 | if (nlk_sk(osk)->pid == pid) { | |
515 | /* Bind collision, search negative pid values. */ | |
516 | pid = rover--; | |
517 | if (rover > -4097) | |
518 | rover = -4097; | |
519 | netlink_table_ungrab(); | |
520 | goto retry; | |
521 | } | |
522 | } | |
523 | netlink_table_ungrab(); | |
524 | ||
525 | err = netlink_insert(sk, pid); | |
526 | if (err == -EADDRINUSE) | |
527 | goto retry; | |
d470e3b4 DM |
528 | |
529 | /* If 2 threads race to autobind, that is fine. */ | |
530 | if (err == -EBUSY) | |
531 | err = 0; | |
532 | ||
533 | return err; | |
1da177e4 LT |
534 | } |
535 | ||
746fac4d YH |
536 | static inline int netlink_capable(struct socket *sock, unsigned int flag) |
537 | { | |
1da177e4 LT |
538 | return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) || |
539 | capable(CAP_NET_ADMIN); | |
746fac4d | 540 | } |
1da177e4 | 541 | |
f7fa9b10 PM |
542 | static void |
543 | netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions) | |
544 | { | |
545 | struct netlink_sock *nlk = nlk_sk(sk); | |
546 | ||
547 | if (nlk->subscriptions && !subscriptions) | |
548 | __sk_del_bind_node(sk); | |
549 | else if (!nlk->subscriptions && subscriptions) | |
550 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
551 | nlk->subscriptions = subscriptions; | |
552 | } | |
553 | ||
513c2500 PM |
554 | static int netlink_alloc_groups(struct sock *sk) |
555 | { | |
556 | struct netlink_sock *nlk = nlk_sk(sk); | |
557 | unsigned int groups; | |
558 | int err = 0; | |
559 | ||
560 | netlink_lock_table(); | |
561 | groups = nl_table[sk->sk_protocol].groups; | |
562 | if (!nl_table[sk->sk_protocol].registered) | |
563 | err = -ENOENT; | |
564 | netlink_unlock_table(); | |
565 | ||
566 | if (err) | |
567 | return err; | |
568 | ||
0da974f4 | 569 | nlk->groups = kzalloc(NLGRPSZ(groups), GFP_KERNEL); |
513c2500 PM |
570 | if (nlk->groups == NULL) |
571 | return -ENOMEM; | |
513c2500 PM |
572 | nlk->ngroups = groups; |
573 | return 0; | |
574 | } | |
575 | ||
1da177e4 LT |
576 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len) |
577 | { | |
578 | struct sock *sk = sock->sk; | |
579 | struct netlink_sock *nlk = nlk_sk(sk); | |
580 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
581 | int err; | |
746fac4d | 582 | |
1da177e4 LT |
583 | if (nladdr->nl_family != AF_NETLINK) |
584 | return -EINVAL; | |
585 | ||
586 | /* Only superuser is allowed to listen multicasts */ | |
513c2500 PM |
587 | if (nladdr->nl_groups) { |
588 | if (!netlink_capable(sock, NL_NONROOT_RECV)) | |
589 | return -EPERM; | |
590 | if (nlk->groups == NULL) { | |
591 | err = netlink_alloc_groups(sk); | |
592 | if (err) | |
593 | return err; | |
594 | } | |
595 | } | |
1da177e4 LT |
596 | |
597 | if (nlk->pid) { | |
598 | if (nladdr->nl_pid != nlk->pid) | |
599 | return -EINVAL; | |
600 | } else { | |
601 | err = nladdr->nl_pid ? | |
602 | netlink_insert(sk, nladdr->nl_pid) : | |
603 | netlink_autobind(sock); | |
604 | if (err) | |
605 | return err; | |
606 | } | |
607 | ||
513c2500 | 608 | if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0])) |
1da177e4 LT |
609 | return 0; |
610 | ||
611 | netlink_table_grab(); | |
f7fa9b10 | 612 | netlink_update_subscriptions(sk, nlk->subscriptions + |
746fac4d YH |
613 | hweight32(nladdr->nl_groups) - |
614 | hweight32(nlk->groups[0])); | |
615 | nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; | |
4277a083 | 616 | netlink_update_listeners(sk); |
1da177e4 LT |
617 | netlink_table_ungrab(); |
618 | ||
619 | return 0; | |
620 | } | |
621 | ||
622 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
623 | int alen, int flags) | |
624 | { | |
625 | int err = 0; | |
626 | struct sock *sk = sock->sk; | |
627 | struct netlink_sock *nlk = nlk_sk(sk); | |
628 | struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr; | |
629 | ||
630 | if (addr->sa_family == AF_UNSPEC) { | |
631 | sk->sk_state = NETLINK_UNCONNECTED; | |
632 | nlk->dst_pid = 0; | |
d629b836 | 633 | nlk->dst_group = 0; |
1da177e4 LT |
634 | return 0; |
635 | } | |
636 | if (addr->sa_family != AF_NETLINK) | |
637 | return -EINVAL; | |
638 | ||
639 | /* Only superuser is allowed to send multicasts */ | |
640 | if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND)) | |
641 | return -EPERM; | |
642 | ||
643 | if (!nlk->pid) | |
644 | err = netlink_autobind(sock); | |
645 | ||
646 | if (err == 0) { | |
647 | sk->sk_state = NETLINK_CONNECTED; | |
648 | nlk->dst_pid = nladdr->nl_pid; | |
d629b836 | 649 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
650 | } |
651 | ||
652 | return err; | |
653 | } | |
654 | ||
655 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer) | |
656 | { | |
657 | struct sock *sk = sock->sk; | |
658 | struct netlink_sock *nlk = nlk_sk(sk); | |
659 | struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr; | |
746fac4d | 660 | |
1da177e4 LT |
661 | nladdr->nl_family = AF_NETLINK; |
662 | nladdr->nl_pad = 0; | |
663 | *addr_len = sizeof(*nladdr); | |
664 | ||
665 | if (peer) { | |
666 | nladdr->nl_pid = nlk->dst_pid; | |
d629b836 | 667 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 LT |
668 | } else { |
669 | nladdr->nl_pid = nlk->pid; | |
513c2500 | 670 | nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0; |
1da177e4 LT |
671 | } |
672 | return 0; | |
673 | } | |
674 | ||
675 | static void netlink_overrun(struct sock *sk) | |
676 | { | |
677 | if (!test_and_set_bit(0, &nlk_sk(sk)->state)) { | |
678 | sk->sk_err = ENOBUFS; | |
679 | sk->sk_error_report(sk); | |
680 | } | |
681 | } | |
682 | ||
683 | static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid) | |
684 | { | |
685 | int protocol = ssk->sk_protocol; | |
686 | struct sock *sock; | |
687 | struct netlink_sock *nlk; | |
688 | ||
689 | sock = netlink_lookup(protocol, pid); | |
690 | if (!sock) | |
691 | return ERR_PTR(-ECONNREFUSED); | |
692 | ||
693 | /* Don't bother queuing skb if kernel socket has no input function */ | |
694 | nlk = nlk_sk(sock); | |
695 | if ((nlk->pid == 0 && !nlk->data_ready) || | |
696 | (sock->sk_state == NETLINK_CONNECTED && | |
697 | nlk->dst_pid != nlk_sk(ssk)->pid)) { | |
698 | sock_put(sock); | |
699 | return ERR_PTR(-ECONNREFUSED); | |
700 | } | |
701 | return sock; | |
702 | } | |
703 | ||
704 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
705 | { | |
6db5fc5d | 706 | struct inode *inode = filp->f_path.dentry->d_inode; |
1da177e4 LT |
707 | struct sock *sock; |
708 | ||
709 | if (!S_ISSOCK(inode->i_mode)) | |
710 | return ERR_PTR(-ENOTSOCK); | |
711 | ||
712 | sock = SOCKET_I(inode)->sk; | |
713 | if (sock->sk_family != AF_NETLINK) | |
714 | return ERR_PTR(-EINVAL); | |
715 | ||
716 | sock_hold(sock); | |
717 | return sock; | |
718 | } | |
719 | ||
720 | /* | |
721 | * Attach a skb to a netlink socket. | |
722 | * The caller must hold a reference to the destination socket. On error, the | |
723 | * reference is dropped. The skb is not send to the destination, just all | |
724 | * all error checks are performed and memory in the queue is reserved. | |
725 | * Return values: | |
726 | * < 0: error. skb freed, reference to sock dropped. | |
727 | * 0: continue | |
728 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
729 | */ | |
a70ea994 AK |
730 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock, |
731 | long timeo, struct sock *ssk) | |
1da177e4 LT |
732 | { |
733 | struct netlink_sock *nlk; | |
734 | ||
735 | nlk = nlk_sk(sk); | |
736 | ||
737 | if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
738 | test_bit(0, &nlk->state)) { | |
739 | DECLARE_WAITQUEUE(wait, current); | |
740 | if (!timeo) { | |
a70ea994 | 741 | if (!ssk || nlk_sk(ssk)->pid == 0) |
1da177e4 LT |
742 | netlink_overrun(sk); |
743 | sock_put(sk); | |
744 | kfree_skb(skb); | |
745 | return -EAGAIN; | |
746 | } | |
747 | ||
748 | __set_current_state(TASK_INTERRUPTIBLE); | |
749 | add_wait_queue(&nlk->wait, &wait); | |
750 | ||
751 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
752 | test_bit(0, &nlk->state)) && | |
753 | !sock_flag(sk, SOCK_DEAD)) | |
754 | timeo = schedule_timeout(timeo); | |
755 | ||
756 | __set_current_state(TASK_RUNNING); | |
757 | remove_wait_queue(&nlk->wait, &wait); | |
758 | sock_put(sk); | |
759 | ||
760 | if (signal_pending(current)) { | |
761 | kfree_skb(skb); | |
762 | return sock_intr_errno(timeo); | |
763 | } | |
764 | return 1; | |
765 | } | |
766 | skb_set_owner_r(skb, sk); | |
767 | return 0; | |
768 | } | |
769 | ||
770 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol) | |
771 | { | |
1da177e4 LT |
772 | int len = skb->len; |
773 | ||
1da177e4 LT |
774 | skb_queue_tail(&sk->sk_receive_queue, skb); |
775 | sk->sk_data_ready(sk, len); | |
776 | sock_put(sk); | |
777 | return len; | |
778 | } | |
779 | ||
780 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
781 | { | |
782 | kfree_skb(skb); | |
783 | sock_put(sk); | |
784 | } | |
785 | ||
37da647d | 786 | static inline struct sk_buff *netlink_trim(struct sk_buff *skb, |
dd0fc66f | 787 | gfp_t allocation) |
1da177e4 LT |
788 | { |
789 | int delta; | |
790 | ||
791 | skb_orphan(skb); | |
792 | ||
4305b541 | 793 | delta = skb->end - skb->tail; |
1da177e4 LT |
794 | if (delta * 2 < skb->truesize) |
795 | return skb; | |
796 | ||
797 | if (skb_shared(skb)) { | |
798 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
799 | if (!nskb) | |
800 | return skb; | |
801 | kfree_skb(skb); | |
802 | skb = nskb; | |
803 | } | |
804 | ||
805 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
806 | skb->truesize -= delta; | |
807 | ||
808 | return skb; | |
809 | } | |
810 | ||
811 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock) | |
812 | { | |
813 | struct sock *sk; | |
814 | int err; | |
815 | long timeo; | |
816 | ||
817 | skb = netlink_trim(skb, gfp_any()); | |
818 | ||
819 | timeo = sock_sndtimeo(ssk, nonblock); | |
820 | retry: | |
821 | sk = netlink_getsockbypid(ssk, pid); | |
822 | if (IS_ERR(sk)) { | |
823 | kfree_skb(skb); | |
824 | return PTR_ERR(sk); | |
825 | } | |
a70ea994 | 826 | err = netlink_attachskb(sk, skb, nonblock, timeo, ssk); |
1da177e4 LT |
827 | if (err == 1) |
828 | goto retry; | |
829 | if (err) | |
830 | return err; | |
831 | ||
832 | return netlink_sendskb(sk, skb, ssk->sk_protocol); | |
833 | } | |
834 | ||
4277a083 PM |
835 | int netlink_has_listeners(struct sock *sk, unsigned int group) |
836 | { | |
837 | int res = 0; | |
838 | ||
839 | BUG_ON(!(nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET)); | |
840 | if (group - 1 < nl_table[sk->sk_protocol].groups) | |
841 | res = test_bit(group - 1, nl_table[sk->sk_protocol].listeners); | |
842 | return res; | |
843 | } | |
844 | EXPORT_SYMBOL_GPL(netlink_has_listeners); | |
845 | ||
1da177e4 LT |
846 | static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) |
847 | { | |
848 | struct netlink_sock *nlk = nlk_sk(sk); | |
849 | ||
850 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
851 | !test_bit(0, &nlk->state)) { | |
852 | skb_set_owner_r(skb, sk); | |
853 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
854 | sk->sk_data_ready(sk, skb->len); | |
855 | return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf; | |
856 | } | |
857 | return -1; | |
858 | } | |
859 | ||
860 | struct netlink_broadcast_data { | |
861 | struct sock *exclude_sk; | |
862 | u32 pid; | |
863 | u32 group; | |
864 | int failure; | |
865 | int congested; | |
866 | int delivered; | |
7d877f3b | 867 | gfp_t allocation; |
1da177e4 LT |
868 | struct sk_buff *skb, *skb2; |
869 | }; | |
870 | ||
871 | static inline int do_one_broadcast(struct sock *sk, | |
872 | struct netlink_broadcast_data *p) | |
873 | { | |
874 | struct netlink_sock *nlk = nlk_sk(sk); | |
875 | int val; | |
876 | ||
877 | if (p->exclude_sk == sk) | |
878 | goto out; | |
879 | ||
f7fa9b10 PM |
880 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
881 | !test_bit(p->group - 1, nlk->groups)) | |
1da177e4 LT |
882 | goto out; |
883 | ||
884 | if (p->failure) { | |
885 | netlink_overrun(sk); | |
886 | goto out; | |
887 | } | |
888 | ||
889 | sock_hold(sk); | |
890 | if (p->skb2 == NULL) { | |
68acc024 | 891 | if (skb_shared(p->skb)) { |
1da177e4 LT |
892 | p->skb2 = skb_clone(p->skb, p->allocation); |
893 | } else { | |
68acc024 TC |
894 | p->skb2 = skb_get(p->skb); |
895 | /* | |
896 | * skb ownership may have been set when | |
897 | * delivered to a previous socket. | |
898 | */ | |
899 | skb_orphan(p->skb2); | |
1da177e4 LT |
900 | } |
901 | } | |
902 | if (p->skb2 == NULL) { | |
903 | netlink_overrun(sk); | |
904 | /* Clone failed. Notify ALL listeners. */ | |
905 | p->failure = 1; | |
906 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { | |
907 | netlink_overrun(sk); | |
908 | } else { | |
909 | p->congested |= val; | |
910 | p->delivered = 1; | |
911 | p->skb2 = NULL; | |
912 | } | |
913 | sock_put(sk); | |
914 | ||
915 | out: | |
916 | return 0; | |
917 | } | |
918 | ||
919 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid, | |
dd0fc66f | 920 | u32 group, gfp_t allocation) |
1da177e4 LT |
921 | { |
922 | struct netlink_broadcast_data info; | |
923 | struct hlist_node *node; | |
924 | struct sock *sk; | |
925 | ||
926 | skb = netlink_trim(skb, allocation); | |
927 | ||
928 | info.exclude_sk = ssk; | |
929 | info.pid = pid; | |
930 | info.group = group; | |
931 | info.failure = 0; | |
932 | info.congested = 0; | |
933 | info.delivered = 0; | |
934 | info.allocation = allocation; | |
935 | info.skb = skb; | |
936 | info.skb2 = NULL; | |
937 | ||
938 | /* While we sleep in clone, do not allow to change socket list */ | |
939 | ||
940 | netlink_lock_table(); | |
941 | ||
942 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
943 | do_one_broadcast(sk, &info); | |
944 | ||
aa1c6a6f TC |
945 | kfree_skb(skb); |
946 | ||
1da177e4 LT |
947 | netlink_unlock_table(); |
948 | ||
949 | if (info.skb2) | |
950 | kfree_skb(info.skb2); | |
1da177e4 LT |
951 | |
952 | if (info.delivered) { | |
953 | if (info.congested && (allocation & __GFP_WAIT)) | |
954 | yield(); | |
955 | return 0; | |
956 | } | |
957 | if (info.failure) | |
958 | return -ENOBUFS; | |
959 | return -ESRCH; | |
960 | } | |
961 | ||
962 | struct netlink_set_err_data { | |
963 | struct sock *exclude_sk; | |
964 | u32 pid; | |
965 | u32 group; | |
966 | int code; | |
967 | }; | |
968 | ||
969 | static inline int do_one_set_err(struct sock *sk, | |
970 | struct netlink_set_err_data *p) | |
971 | { | |
972 | struct netlink_sock *nlk = nlk_sk(sk); | |
973 | ||
974 | if (sk == p->exclude_sk) | |
975 | goto out; | |
976 | ||
f7fa9b10 PM |
977 | if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups || |
978 | !test_bit(p->group - 1, nlk->groups)) | |
1da177e4 LT |
979 | goto out; |
980 | ||
981 | sk->sk_err = p->code; | |
982 | sk->sk_error_report(sk); | |
983 | out: | |
984 | return 0; | |
985 | } | |
986 | ||
987 | void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code) | |
988 | { | |
989 | struct netlink_set_err_data info; | |
990 | struct hlist_node *node; | |
991 | struct sock *sk; | |
992 | ||
993 | info.exclude_sk = ssk; | |
994 | info.pid = pid; | |
995 | info.group = group; | |
996 | info.code = code; | |
997 | ||
998 | read_lock(&nl_table_lock); | |
999 | ||
1000 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
1001 | do_one_set_err(sk, &info); | |
1002 | ||
1003 | read_unlock(&nl_table_lock); | |
1004 | } | |
1005 | ||
9a4595bc | 1006 | static int netlink_setsockopt(struct socket *sock, int level, int optname, |
746fac4d | 1007 | char __user *optval, int optlen) |
9a4595bc PM |
1008 | { |
1009 | struct sock *sk = sock->sk; | |
1010 | struct netlink_sock *nlk = nlk_sk(sk); | |
1011 | int val = 0, err; | |
1012 | ||
1013 | if (level != SOL_NETLINK) | |
1014 | return -ENOPROTOOPT; | |
1015 | ||
1016 | if (optlen >= sizeof(int) && | |
1017 | get_user(val, (int __user *)optval)) | |
1018 | return -EFAULT; | |
1019 | ||
1020 | switch (optname) { | |
1021 | case NETLINK_PKTINFO: | |
1022 | if (val) | |
1023 | nlk->flags |= NETLINK_RECV_PKTINFO; | |
1024 | else | |
1025 | nlk->flags &= ~NETLINK_RECV_PKTINFO; | |
1026 | err = 0; | |
1027 | break; | |
1028 | case NETLINK_ADD_MEMBERSHIP: | |
1029 | case NETLINK_DROP_MEMBERSHIP: { | |
1030 | unsigned int subscriptions; | |
1031 | int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0; | |
1032 | ||
1033 | if (!netlink_capable(sock, NL_NONROOT_RECV)) | |
1034 | return -EPERM; | |
513c2500 PM |
1035 | if (nlk->groups == NULL) { |
1036 | err = netlink_alloc_groups(sk); | |
1037 | if (err) | |
1038 | return err; | |
1039 | } | |
9a4595bc PM |
1040 | if (!val || val - 1 >= nlk->ngroups) |
1041 | return -EINVAL; | |
1042 | netlink_table_grab(); | |
1043 | old = test_bit(val - 1, nlk->groups); | |
1044 | subscriptions = nlk->subscriptions - old + new; | |
1045 | if (new) | |
1046 | __set_bit(val - 1, nlk->groups); | |
1047 | else | |
1048 | __clear_bit(val - 1, nlk->groups); | |
1049 | netlink_update_subscriptions(sk, subscriptions); | |
4277a083 | 1050 | netlink_update_listeners(sk); |
9a4595bc PM |
1051 | netlink_table_ungrab(); |
1052 | err = 0; | |
1053 | break; | |
1054 | } | |
1055 | default: | |
1056 | err = -ENOPROTOOPT; | |
1057 | } | |
1058 | return err; | |
1059 | } | |
1060 | ||
1061 | static int netlink_getsockopt(struct socket *sock, int level, int optname, | |
746fac4d | 1062 | char __user *optval, int __user *optlen) |
9a4595bc PM |
1063 | { |
1064 | struct sock *sk = sock->sk; | |
1065 | struct netlink_sock *nlk = nlk_sk(sk); | |
1066 | int len, val, err; | |
1067 | ||
1068 | if (level != SOL_NETLINK) | |
1069 | return -ENOPROTOOPT; | |
1070 | ||
1071 | if (get_user(len, optlen)) | |
1072 | return -EFAULT; | |
1073 | if (len < 0) | |
1074 | return -EINVAL; | |
1075 | ||
1076 | switch (optname) { | |
1077 | case NETLINK_PKTINFO: | |
1078 | if (len < sizeof(int)) | |
1079 | return -EINVAL; | |
1080 | len = sizeof(int); | |
1081 | val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0; | |
a27b58fe HC |
1082 | if (put_user(len, optlen) || |
1083 | put_user(val, optval)) | |
1084 | return -EFAULT; | |
9a4595bc PM |
1085 | err = 0; |
1086 | break; | |
1087 | default: | |
1088 | err = -ENOPROTOOPT; | |
1089 | } | |
1090 | return err; | |
1091 | } | |
1092 | ||
1093 | static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) | |
1094 | { | |
1095 | struct nl_pktinfo info; | |
1096 | ||
1097 | info.group = NETLINK_CB(skb).dst_group; | |
1098 | put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info); | |
1099 | } | |
1100 | ||
1da177e4 LT |
1101 | static inline void netlink_rcv_wake(struct sock *sk) |
1102 | { | |
1103 | struct netlink_sock *nlk = nlk_sk(sk); | |
1104 | ||
b03efcfb | 1105 | if (skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
1106 | clear_bit(0, &nlk->state); |
1107 | if (!test_bit(0, &nlk->state)) | |
1108 | wake_up_interruptible(&nlk->wait); | |
1109 | } | |
1110 | ||
1111 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
1112 | struct msghdr *msg, size_t len) | |
1113 | { | |
1114 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1115 | struct sock *sk = sock->sk; | |
1116 | struct netlink_sock *nlk = nlk_sk(sk); | |
1117 | struct sockaddr_nl *addr=msg->msg_name; | |
1118 | u32 dst_pid; | |
d629b836 | 1119 | u32 dst_group; |
1da177e4 LT |
1120 | struct sk_buff *skb; |
1121 | int err; | |
1122 | struct scm_cookie scm; | |
1123 | ||
1124 | if (msg->msg_flags&MSG_OOB) | |
1125 | return -EOPNOTSUPP; | |
1126 | ||
1127 | if (NULL == siocb->scm) | |
1128 | siocb->scm = &scm; | |
1129 | err = scm_send(sock, msg, siocb->scm); | |
1130 | if (err < 0) | |
1131 | return err; | |
1132 | ||
1133 | if (msg->msg_namelen) { | |
1134 | if (addr->nl_family != AF_NETLINK) | |
1135 | return -EINVAL; | |
1136 | dst_pid = addr->nl_pid; | |
d629b836 PM |
1137 | dst_group = ffs(addr->nl_groups); |
1138 | if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND)) | |
1da177e4 LT |
1139 | return -EPERM; |
1140 | } else { | |
1141 | dst_pid = nlk->dst_pid; | |
d629b836 | 1142 | dst_group = nlk->dst_group; |
1da177e4 LT |
1143 | } |
1144 | ||
1145 | if (!nlk->pid) { | |
1146 | err = netlink_autobind(sock); | |
1147 | if (err) | |
1148 | goto out; | |
1149 | } | |
1150 | ||
1151 | err = -EMSGSIZE; | |
1152 | if (len > sk->sk_sndbuf - 32) | |
1153 | goto out; | |
1154 | err = -ENOBUFS; | |
339bf98f | 1155 | skb = alloc_skb(len, GFP_KERNEL); |
1da177e4 LT |
1156 | if (skb==NULL) |
1157 | goto out; | |
1158 | ||
1159 | NETLINK_CB(skb).pid = nlk->pid; | |
d629b836 | 1160 | NETLINK_CB(skb).dst_group = dst_group; |
c94c257c | 1161 | NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context); |
e7c34970 | 1162 | selinux_get_task_sid(current, &(NETLINK_CB(skb).sid)); |
1da177e4 LT |
1163 | memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred)); |
1164 | ||
1165 | /* What can I do? Netlink is asynchronous, so that | |
1166 | we will have to save current capabilities to | |
1167 | check them, when this message will be delivered | |
1168 | to corresponding kernel module. --ANK (980802) | |
1169 | */ | |
1170 | ||
1171 | err = -EFAULT; | |
1172 | if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) { | |
1173 | kfree_skb(skb); | |
1174 | goto out; | |
1175 | } | |
1176 | ||
1177 | err = security_netlink_send(sk, skb); | |
1178 | if (err) { | |
1179 | kfree_skb(skb); | |
1180 | goto out; | |
1181 | } | |
1182 | ||
d629b836 | 1183 | if (dst_group) { |
1da177e4 | 1184 | atomic_inc(&skb->users); |
d629b836 | 1185 | netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL); |
1da177e4 LT |
1186 | } |
1187 | err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT); | |
1188 | ||
1189 | out: | |
1190 | return err; | |
1191 | } | |
1192 | ||
1193 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
1194 | struct msghdr *msg, size_t len, | |
1195 | int flags) | |
1196 | { | |
1197 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1198 | struct scm_cookie scm; | |
1199 | struct sock *sk = sock->sk; | |
1200 | struct netlink_sock *nlk = nlk_sk(sk); | |
1201 | int noblock = flags&MSG_DONTWAIT; | |
1202 | size_t copied; | |
1203 | struct sk_buff *skb; | |
1204 | int err; | |
1205 | ||
1206 | if (flags&MSG_OOB) | |
1207 | return -EOPNOTSUPP; | |
1208 | ||
1209 | copied = 0; | |
1210 | ||
1211 | skb = skb_recv_datagram(sk,flags,noblock,&err); | |
1212 | if (skb==NULL) | |
1213 | goto out; | |
1214 | ||
1215 | msg->msg_namelen = 0; | |
1216 | ||
1217 | copied = skb->len; | |
1218 | if (len < copied) { | |
1219 | msg->msg_flags |= MSG_TRUNC; | |
1220 | copied = len; | |
1221 | } | |
1222 | ||
badff6d0 | 1223 | skb_reset_transport_header(skb); |
1da177e4 LT |
1224 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); |
1225 | ||
1226 | if (msg->msg_name) { | |
1227 | struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name; | |
1228 | addr->nl_family = AF_NETLINK; | |
1229 | addr->nl_pad = 0; | |
1230 | addr->nl_pid = NETLINK_CB(skb).pid; | |
d629b836 | 1231 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1232 | msg->msg_namelen = sizeof(*addr); |
1233 | } | |
1234 | ||
cc9a06cd PM |
1235 | if (nlk->flags & NETLINK_RECV_PKTINFO) |
1236 | netlink_cmsg_recv_pktinfo(msg, skb); | |
1237 | ||
1da177e4 LT |
1238 | if (NULL == siocb->scm) { |
1239 | memset(&scm, 0, sizeof(scm)); | |
1240 | siocb->scm = &scm; | |
1241 | } | |
1242 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
1243 | skb_free_datagram(sk, skb); | |
1244 | ||
1245 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) | |
1246 | netlink_dump(sk); | |
1247 | ||
1248 | scm_recv(sock, msg, siocb->scm, flags); | |
1249 | ||
b558ff79 DM |
1250 | if (flags & MSG_TRUNC) |
1251 | copied = skb->len; | |
1252 | ||
1da177e4 LT |
1253 | out: |
1254 | netlink_rcv_wake(sk); | |
1255 | return err ? : copied; | |
1256 | } | |
1257 | ||
1258 | static void netlink_data_ready(struct sock *sk, int len) | |
1259 | { | |
1260 | struct netlink_sock *nlk = nlk_sk(sk); | |
1261 | ||
1262 | if (nlk->data_ready) | |
1263 | nlk->data_ready(sk, len); | |
1264 | netlink_rcv_wake(sk); | |
1265 | } | |
1266 | ||
1267 | /* | |
746fac4d | 1268 | * We export these functions to other modules. They provide a |
1da177e4 LT |
1269 | * complete set of kernel non-blocking support for message |
1270 | * queueing. | |
1271 | */ | |
1272 | ||
1273 | struct sock * | |
06628607 | 1274 | netlink_kernel_create(int unit, unsigned int groups, |
746fac4d | 1275 | void (*input)(struct sock *sk, int len), |
af65bdfc | 1276 | struct mutex *cb_mutex, struct module *module) |
1da177e4 LT |
1277 | { |
1278 | struct socket *sock; | |
1279 | struct sock *sk; | |
77247bbb | 1280 | struct netlink_sock *nlk; |
4277a083 | 1281 | unsigned long *listeners = NULL; |
1da177e4 | 1282 | |
fab2caf6 | 1283 | BUG_ON(!nl_table); |
1da177e4 LT |
1284 | |
1285 | if (unit<0 || unit>=MAX_LINKS) | |
1286 | return NULL; | |
1287 | ||
1288 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1289 | return NULL; | |
1290 | ||
af65bdfc | 1291 | if (__netlink_create(sock, cb_mutex, unit) < 0) |
77247bbb | 1292 | goto out_sock_release; |
4fdb3bb7 | 1293 | |
4277a083 PM |
1294 | if (groups < 32) |
1295 | groups = 32; | |
1296 | ||
1297 | listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL); | |
1298 | if (!listeners) | |
1299 | goto out_sock_release; | |
1300 | ||
1da177e4 LT |
1301 | sk = sock->sk; |
1302 | sk->sk_data_ready = netlink_data_ready; | |
1303 | if (input) | |
1304 | nlk_sk(sk)->data_ready = input; | |
1305 | ||
77247bbb PM |
1306 | if (netlink_insert(sk, 0)) |
1307 | goto out_sock_release; | |
4fdb3bb7 | 1308 | |
77247bbb PM |
1309 | nlk = nlk_sk(sk); |
1310 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 1311 | |
4fdb3bb7 | 1312 | netlink_table_grab(); |
4277a083 PM |
1313 | nl_table[unit].groups = groups; |
1314 | nl_table[unit].listeners = listeners; | |
af65bdfc | 1315 | nl_table[unit].cb_mutex = cb_mutex; |
77247bbb | 1316 | nl_table[unit].module = module; |
ab33a171 | 1317 | nl_table[unit].registered = 1; |
4fdb3bb7 | 1318 | netlink_table_ungrab(); |
77247bbb PM |
1319 | |
1320 | return sk; | |
1321 | ||
4fdb3bb7 | 1322 | out_sock_release: |
4277a083 | 1323 | kfree(listeners); |
4fdb3bb7 | 1324 | sock_release(sock); |
77247bbb | 1325 | return NULL; |
1da177e4 LT |
1326 | } |
1327 | ||
1328 | void netlink_set_nonroot(int protocol, unsigned int flags) | |
746fac4d YH |
1329 | { |
1330 | if ((unsigned int)protocol < MAX_LINKS) | |
1da177e4 | 1331 | nl_table[protocol].nl_nonroot = flags; |
746fac4d | 1332 | } |
1da177e4 LT |
1333 | |
1334 | static void netlink_destroy_callback(struct netlink_callback *cb) | |
1335 | { | |
1336 | if (cb->skb) | |
1337 | kfree_skb(cb->skb); | |
1338 | kfree(cb); | |
1339 | } | |
1340 | ||
1341 | /* | |
1342 | * It looks a bit ugly. | |
1343 | * It would be better to create kernel thread. | |
1344 | */ | |
1345 | ||
1346 | static int netlink_dump(struct sock *sk) | |
1347 | { | |
1348 | struct netlink_sock *nlk = nlk_sk(sk); | |
1349 | struct netlink_callback *cb; | |
1350 | struct sk_buff *skb; | |
1351 | struct nlmsghdr *nlh; | |
bf8b79e4 | 1352 | int len, err = -ENOBUFS; |
746fac4d | 1353 | |
1da177e4 LT |
1354 | skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL); |
1355 | if (!skb) | |
bf8b79e4 | 1356 | goto errout; |
1da177e4 | 1357 | |
af65bdfc | 1358 | mutex_lock(nlk->cb_mutex); |
1da177e4 LT |
1359 | |
1360 | cb = nlk->cb; | |
1361 | if (cb == NULL) { | |
bf8b79e4 TG |
1362 | err = -EINVAL; |
1363 | goto errout_skb; | |
1da177e4 LT |
1364 | } |
1365 | ||
1366 | len = cb->dump(skb, cb); | |
1367 | ||
1368 | if (len > 0) { | |
af65bdfc | 1369 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1370 | skb_queue_tail(&sk->sk_receive_queue, skb); |
1371 | sk->sk_data_ready(sk, len); | |
1372 | return 0; | |
1373 | } | |
1374 | ||
bf8b79e4 TG |
1375 | nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
1376 | if (!nlh) | |
1377 | goto errout_skb; | |
1378 | ||
1379 | memcpy(nlmsg_data(nlh), &len, sizeof(len)); | |
1380 | ||
1da177e4 LT |
1381 | skb_queue_tail(&sk->sk_receive_queue, skb); |
1382 | sk->sk_data_ready(sk, skb->len); | |
1383 | ||
a8f74b22 TG |
1384 | if (cb->done) |
1385 | cb->done(cb); | |
1da177e4 | 1386 | nlk->cb = NULL; |
af65bdfc | 1387 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1388 | |
1389 | netlink_destroy_callback(cb); | |
1da177e4 | 1390 | return 0; |
1797754e | 1391 | |
bf8b79e4 | 1392 | errout_skb: |
af65bdfc | 1393 | mutex_unlock(nlk->cb_mutex); |
bf8b79e4 TG |
1394 | kfree_skb(skb); |
1395 | errout: | |
1396 | return err; | |
1da177e4 LT |
1397 | } |
1398 | ||
1399 | int netlink_dump_start(struct sock *ssk, struct sk_buff *skb, | |
1400 | struct nlmsghdr *nlh, | |
1401 | int (*dump)(struct sk_buff *skb, struct netlink_callback*), | |
1402 | int (*done)(struct netlink_callback*)) | |
1403 | { | |
1404 | struct netlink_callback *cb; | |
1405 | struct sock *sk; | |
1406 | struct netlink_sock *nlk; | |
1407 | ||
0da974f4 | 1408 | cb = kzalloc(sizeof(*cb), GFP_KERNEL); |
1da177e4 LT |
1409 | if (cb == NULL) |
1410 | return -ENOBUFS; | |
1411 | ||
1da177e4 LT |
1412 | cb->dump = dump; |
1413 | cb->done = done; | |
1414 | cb->nlh = nlh; | |
1415 | atomic_inc(&skb->users); | |
1416 | cb->skb = skb; | |
1417 | ||
1418 | sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid); | |
1419 | if (sk == NULL) { | |
1420 | netlink_destroy_callback(cb); | |
1421 | return -ECONNREFUSED; | |
1422 | } | |
1423 | nlk = nlk_sk(sk); | |
ac57b3a9 | 1424 | /* A dump or destruction is in progress... */ |
af65bdfc | 1425 | mutex_lock(nlk->cb_mutex); |
ac57b3a9 | 1426 | if (nlk->cb || sock_flag(sk, SOCK_DEAD)) { |
af65bdfc | 1427 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1428 | netlink_destroy_callback(cb); |
1429 | sock_put(sk); | |
1430 | return -EBUSY; | |
1431 | } | |
1432 | nlk->cb = cb; | |
af65bdfc | 1433 | mutex_unlock(nlk->cb_mutex); |
1da177e4 LT |
1434 | |
1435 | netlink_dump(sk); | |
1436 | sock_put(sk); | |
c702e804 TG |
1437 | |
1438 | /* We successfully started a dump, by returning -EINTR we | |
1439 | * signal the queue mangement to interrupt processing of | |
1440 | * any netlink messages so userspace gets a chance to read | |
1441 | * the results. */ | |
1442 | return -EINTR; | |
1da177e4 LT |
1443 | } |
1444 | ||
1445 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
1446 | { | |
1447 | struct sk_buff *skb; | |
1448 | struct nlmsghdr *rep; | |
1449 | struct nlmsgerr *errmsg; | |
339bf98f | 1450 | size_t payload = sizeof(*errmsg); |
1da177e4 | 1451 | |
339bf98f TG |
1452 | /* error messages get the original request appened */ |
1453 | if (err) | |
1454 | payload += nlmsg_len(nlh); | |
1da177e4 | 1455 | |
339bf98f | 1456 | skb = nlmsg_new(payload, GFP_KERNEL); |
1da177e4 LT |
1457 | if (!skb) { |
1458 | struct sock *sk; | |
1459 | ||
1460 | sk = netlink_lookup(in_skb->sk->sk_protocol, | |
1461 | NETLINK_CB(in_skb).pid); | |
1462 | if (sk) { | |
1463 | sk->sk_err = ENOBUFS; | |
1464 | sk->sk_error_report(sk); | |
1465 | sock_put(sk); | |
1466 | } | |
1467 | return; | |
1468 | } | |
1469 | ||
1470 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, | |
1797754e | 1471 | NLMSG_ERROR, sizeof(struct nlmsgerr), 0); |
bf8b79e4 | 1472 | errmsg = nlmsg_data(rep); |
1da177e4 | 1473 | errmsg->error = err; |
bf8b79e4 | 1474 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh)); |
1da177e4 LT |
1475 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); |
1476 | } | |
1477 | ||
82ace47a | 1478 | static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, |
1d00a4eb | 1479 | struct nlmsghdr *)) |
82ace47a | 1480 | { |
82ace47a TG |
1481 | struct nlmsghdr *nlh; |
1482 | int err; | |
1483 | ||
1484 | while (skb->len >= nlmsg_total_size(0)) { | |
b529ccf2 | 1485 | nlh = nlmsg_hdr(skb); |
d35b6856 | 1486 | err = 0; |
82ace47a | 1487 | |
ad8e4b75 | 1488 | if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) |
82ace47a TG |
1489 | return 0; |
1490 | ||
d35b6856 TG |
1491 | /* Only requests are handled by the kernel */ |
1492 | if (!(nlh->nlmsg_flags & NLM_F_REQUEST)) | |
1493 | goto skip; | |
45e7ae7f TG |
1494 | |
1495 | /* Skip control messages */ | |
1496 | if (nlh->nlmsg_type < NLMSG_MIN_TYPE) | |
1497 | goto skip; | |
d35b6856 | 1498 | |
1d00a4eb TG |
1499 | err = cb(skb, nlh); |
1500 | if (err == -EINTR) { | |
1501 | /* Not an error, but we interrupt processing */ | |
1502 | netlink_queue_skip(nlh, skb); | |
1503 | return err; | |
d35b6856 TG |
1504 | } |
1505 | skip: | |
1506 | if (nlh->nlmsg_flags & NLM_F_ACK || err) | |
82ace47a | 1507 | netlink_ack(skb, nlh, err); |
82ace47a | 1508 | |
bf8b79e4 | 1509 | netlink_queue_skip(nlh, skb); |
82ace47a TG |
1510 | } |
1511 | ||
1512 | return 0; | |
1513 | } | |
1514 | ||
1515 | /** | |
1516 | * nelink_run_queue - Process netlink receive queue. | |
1517 | * @sk: Netlink socket containing the queue | |
1518 | * @qlen: Place to store queue length upon entry | |
1519 | * @cb: Callback function invoked for each netlink message found | |
1520 | * | |
1521 | * Processes as much as there was in the queue upon entry and invokes | |
1522 | * a callback function for each netlink message found. The callback | |
1523 | * function may refuse a message by returning a negative error code | |
1524 | * but setting the error pointer to 0 in which case this function | |
1525 | * returns with a qlen != 0. | |
1526 | * | |
1527 | * qlen must be initialized to 0 before the initial entry, afterwards | |
1528 | * the function may be called repeatedly until qlen reaches 0. | |
1d00a4eb TG |
1529 | * |
1530 | * The callback function may return -EINTR to signal that processing | |
1531 | * of netlink messages shall be interrupted. In this case the message | |
1532 | * currently being processed will NOT be requeued onto the receive | |
1533 | * queue. | |
82ace47a TG |
1534 | */ |
1535 | void netlink_run_queue(struct sock *sk, unsigned int *qlen, | |
1d00a4eb | 1536 | int (*cb)(struct sk_buff *, struct nlmsghdr *)) |
82ace47a TG |
1537 | { |
1538 | struct sk_buff *skb; | |
1539 | ||
1540 | if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue)) | |
1541 | *qlen = skb_queue_len(&sk->sk_receive_queue); | |
1542 | ||
1543 | for (; *qlen; (*qlen)--) { | |
1544 | skb = skb_dequeue(&sk->sk_receive_queue); | |
1545 | if (netlink_rcv_skb(skb, cb)) { | |
1546 | if (skb->len) | |
1547 | skb_queue_head(&sk->sk_receive_queue, skb); | |
1548 | else { | |
1549 | kfree_skb(skb); | |
1550 | (*qlen)--; | |
1551 | } | |
1552 | break; | |
1553 | } | |
1554 | ||
1555 | kfree_skb(skb); | |
1556 | } | |
1557 | } | |
1558 | ||
1559 | /** | |
1560 | * netlink_queue_skip - Skip netlink message while processing queue. | |
1561 | * @nlh: Netlink message to be skipped | |
1562 | * @skb: Socket buffer containing the netlink messages. | |
1563 | * | |
1564 | * Pulls the given netlink message off the socket buffer so the next | |
1565 | * call to netlink_queue_run() will not reconsider the message. | |
1566 | */ | |
1567 | void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb) | |
1568 | { | |
1569 | int msglen = NLMSG_ALIGN(nlh->nlmsg_len); | |
1570 | ||
1571 | if (msglen > skb->len) | |
1572 | msglen = skb->len; | |
1573 | ||
1574 | skb_pull(skb, msglen); | |
1575 | } | |
1da177e4 | 1576 | |
d387f6ad TG |
1577 | /** |
1578 | * nlmsg_notify - send a notification netlink message | |
1579 | * @sk: netlink socket to use | |
1580 | * @skb: notification message | |
1581 | * @pid: destination netlink pid for reports or 0 | |
1582 | * @group: destination multicast group or 0 | |
1583 | * @report: 1 to report back, 0 to disable | |
1584 | * @flags: allocation flags | |
1585 | */ | |
1586 | int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid, | |
1587 | unsigned int group, int report, gfp_t flags) | |
1588 | { | |
1589 | int err = 0; | |
1590 | ||
1591 | if (group) { | |
1592 | int exclude_pid = 0; | |
1593 | ||
1594 | if (report) { | |
1595 | atomic_inc(&skb->users); | |
1596 | exclude_pid = pid; | |
1597 | } | |
1598 | ||
1599 | /* errors reported via destination sk->sk_err */ | |
1600 | nlmsg_multicast(sk, skb, exclude_pid, group, flags); | |
1601 | } | |
1602 | ||
1603 | if (report) | |
1604 | err = nlmsg_unicast(sk, skb, pid); | |
1605 | ||
1606 | return err; | |
1607 | } | |
1608 | ||
1da177e4 LT |
1609 | #ifdef CONFIG_PROC_FS |
1610 | struct nl_seq_iter { | |
1611 | int link; | |
1612 | int hash_idx; | |
1613 | }; | |
1614 | ||
1615 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
1616 | { | |
1617 | struct nl_seq_iter *iter = seq->private; | |
1618 | int i, j; | |
1619 | struct sock *s; | |
1620 | struct hlist_node *node; | |
1621 | loff_t off = 0; | |
1622 | ||
1623 | for (i=0; i<MAX_LINKS; i++) { | |
1624 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1625 | ||
1626 | for (j = 0; j <= hash->mask; j++) { | |
1627 | sk_for_each(s, node, &hash->table[j]) { | |
1628 | if (off == pos) { | |
1629 | iter->link = i; | |
1630 | iter->hash_idx = j; | |
1631 | return s; | |
1632 | } | |
1633 | ++off; | |
1634 | } | |
1635 | } | |
1636 | } | |
1637 | return NULL; | |
1638 | } | |
1639 | ||
1640 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
1641 | { | |
1642 | read_lock(&nl_table_lock); | |
1643 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
1644 | } | |
1645 | ||
1646 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1647 | { | |
1648 | struct sock *s; | |
1649 | struct nl_seq_iter *iter; | |
1650 | int i, j; | |
1651 | ||
1652 | ++*pos; | |
1653 | ||
1654 | if (v == SEQ_START_TOKEN) | |
1655 | return netlink_seq_socket_idx(seq, 0); | |
746fac4d | 1656 | |
1da177e4 LT |
1657 | s = sk_next(v); |
1658 | if (s) | |
1659 | return s; | |
1660 | ||
1661 | iter = seq->private; | |
1662 | i = iter->link; | |
1663 | j = iter->hash_idx + 1; | |
1664 | ||
1665 | do { | |
1666 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1667 | ||
1668 | for (; j <= hash->mask; j++) { | |
1669 | s = sk_head(&hash->table[j]); | |
1670 | if (s) { | |
1671 | iter->link = i; | |
1672 | iter->hash_idx = j; | |
1673 | return s; | |
1674 | } | |
1675 | } | |
1676 | ||
1677 | j = 0; | |
1678 | } while (++i < MAX_LINKS); | |
1679 | ||
1680 | return NULL; | |
1681 | } | |
1682 | ||
1683 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
1684 | { | |
1685 | read_unlock(&nl_table_lock); | |
1686 | } | |
1687 | ||
1688 | ||
1689 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
1690 | { | |
1691 | if (v == SEQ_START_TOKEN) | |
1692 | seq_puts(seq, | |
1693 | "sk Eth Pid Groups " | |
1694 | "Rmem Wmem Dump Locks\n"); | |
1695 | else { | |
1696 | struct sock *s = v; | |
1697 | struct netlink_sock *nlk = nlk_sk(s); | |
1698 | ||
1699 | seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n", | |
1700 | s, | |
1701 | s->sk_protocol, | |
1702 | nlk->pid, | |
513c2500 | 1703 | nlk->groups ? (u32)nlk->groups[0] : 0, |
1da177e4 LT |
1704 | atomic_read(&s->sk_rmem_alloc), |
1705 | atomic_read(&s->sk_wmem_alloc), | |
1706 | nlk->cb, | |
1707 | atomic_read(&s->sk_refcnt) | |
1708 | ); | |
1709 | ||
1710 | } | |
1711 | return 0; | |
1712 | } | |
1713 | ||
1714 | static struct seq_operations netlink_seq_ops = { | |
1715 | .start = netlink_seq_start, | |
1716 | .next = netlink_seq_next, | |
1717 | .stop = netlink_seq_stop, | |
1718 | .show = netlink_seq_show, | |
1719 | }; | |
1720 | ||
1721 | ||
1722 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
1723 | { | |
1724 | struct seq_file *seq; | |
1725 | struct nl_seq_iter *iter; | |
1726 | int err; | |
1727 | ||
0da974f4 | 1728 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); |
1da177e4 LT |
1729 | if (!iter) |
1730 | return -ENOMEM; | |
1731 | ||
1732 | err = seq_open(file, &netlink_seq_ops); | |
1733 | if (err) { | |
1734 | kfree(iter); | |
1735 | return err; | |
1736 | } | |
1737 | ||
1da177e4 LT |
1738 | seq = file->private_data; |
1739 | seq->private = iter; | |
1740 | return 0; | |
1741 | } | |
1742 | ||
da7071d7 | 1743 | static const struct file_operations netlink_seq_fops = { |
1da177e4 LT |
1744 | .owner = THIS_MODULE, |
1745 | .open = netlink_seq_open, | |
1746 | .read = seq_read, | |
1747 | .llseek = seq_lseek, | |
1748 | .release = seq_release_private, | |
1749 | }; | |
1750 | ||
1751 | #endif | |
1752 | ||
1753 | int netlink_register_notifier(struct notifier_block *nb) | |
1754 | { | |
e041c683 | 1755 | return atomic_notifier_chain_register(&netlink_chain, nb); |
1da177e4 LT |
1756 | } |
1757 | ||
1758 | int netlink_unregister_notifier(struct notifier_block *nb) | |
1759 | { | |
e041c683 | 1760 | return atomic_notifier_chain_unregister(&netlink_chain, nb); |
1da177e4 | 1761 | } |
746fac4d | 1762 | |
90ddc4f0 | 1763 | static const struct proto_ops netlink_ops = { |
1da177e4 LT |
1764 | .family = PF_NETLINK, |
1765 | .owner = THIS_MODULE, | |
1766 | .release = netlink_release, | |
1767 | .bind = netlink_bind, | |
1768 | .connect = netlink_connect, | |
1769 | .socketpair = sock_no_socketpair, | |
1770 | .accept = sock_no_accept, | |
1771 | .getname = netlink_getname, | |
1772 | .poll = datagram_poll, | |
1773 | .ioctl = sock_no_ioctl, | |
1774 | .listen = sock_no_listen, | |
1775 | .shutdown = sock_no_shutdown, | |
9a4595bc PM |
1776 | .setsockopt = netlink_setsockopt, |
1777 | .getsockopt = netlink_getsockopt, | |
1da177e4 LT |
1778 | .sendmsg = netlink_sendmsg, |
1779 | .recvmsg = netlink_recvmsg, | |
1780 | .mmap = sock_no_mmap, | |
1781 | .sendpage = sock_no_sendpage, | |
1782 | }; | |
1783 | ||
1784 | static struct net_proto_family netlink_family_ops = { | |
1785 | .family = PF_NETLINK, | |
1786 | .create = netlink_create, | |
1787 | .owner = THIS_MODULE, /* for consistency 8) */ | |
1788 | }; | |
1789 | ||
1da177e4 LT |
1790 | static int __init netlink_proto_init(void) |
1791 | { | |
1792 | struct sk_buff *dummy_skb; | |
1793 | int i; | |
1794 | unsigned long max; | |
1795 | unsigned int order; | |
1796 | int err = proto_register(&netlink_proto, 0); | |
1797 | ||
1798 | if (err != 0) | |
1799 | goto out; | |
1800 | ||
ef047f5e | 1801 | BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb)); |
1da177e4 | 1802 | |
0da974f4 | 1803 | nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL); |
fab2caf6 AM |
1804 | if (!nl_table) |
1805 | goto panic; | |
1da177e4 | 1806 | |
1da177e4 LT |
1807 | if (num_physpages >= (128 * 1024)) |
1808 | max = num_physpages >> (21 - PAGE_SHIFT); | |
1809 | else | |
1810 | max = num_physpages >> (23 - PAGE_SHIFT); | |
1811 | ||
1812 | order = get_bitmask_order(max) - 1 + PAGE_SHIFT; | |
1813 | max = (1UL << order) / sizeof(struct hlist_head); | |
1814 | order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1; | |
1815 | ||
1816 | for (i = 0; i < MAX_LINKS; i++) { | |
1817 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1818 | ||
1819 | hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table)); | |
1820 | if (!hash->table) { | |
1821 | while (i-- > 0) | |
1822 | nl_pid_hash_free(nl_table[i].hash.table, | |
1823 | 1 * sizeof(*hash->table)); | |
1824 | kfree(nl_table); | |
fab2caf6 | 1825 | goto panic; |
1da177e4 LT |
1826 | } |
1827 | memset(hash->table, 0, 1 * sizeof(*hash->table)); | |
1828 | hash->max_shift = order; | |
1829 | hash->shift = 0; | |
1830 | hash->mask = 0; | |
1831 | hash->rehash_time = jiffies; | |
1832 | } | |
1833 | ||
1834 | sock_register(&netlink_family_ops); | |
1835 | #ifdef CONFIG_PROC_FS | |
1836 | proc_net_fops_create("netlink", 0, &netlink_seq_fops); | |
1837 | #endif | |
746fac4d | 1838 | /* The netlink device handler may be needed early. */ |
1da177e4 LT |
1839 | rtnetlink_init(); |
1840 | out: | |
1841 | return err; | |
fab2caf6 AM |
1842 | panic: |
1843 | panic("netlink_init: Cannot allocate nl_table\n"); | |
1da177e4 LT |
1844 | } |
1845 | ||
1da177e4 | 1846 | core_initcall(netlink_proto_init); |
1da177e4 LT |
1847 | |
1848 | EXPORT_SYMBOL(netlink_ack); | |
82ace47a TG |
1849 | EXPORT_SYMBOL(netlink_run_queue); |
1850 | EXPORT_SYMBOL(netlink_queue_skip); | |
1da177e4 LT |
1851 | EXPORT_SYMBOL(netlink_broadcast); |
1852 | EXPORT_SYMBOL(netlink_dump_start); | |
1853 | EXPORT_SYMBOL(netlink_kernel_create); | |
1854 | EXPORT_SYMBOL(netlink_register_notifier); | |
1da177e4 LT |
1855 | EXPORT_SYMBOL(netlink_set_nonroot); |
1856 | EXPORT_SYMBOL(netlink_unicast); | |
1857 | EXPORT_SYMBOL(netlink_unregister_notifier); | |
d387f6ad | 1858 | EXPORT_SYMBOL(nlmsg_notify); |