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. | |
11 | * | |
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 | ||
24 | #include <linux/config.h> | |
25 | #include <linux/module.h> | |
26 | ||
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 AM |
57 | #include <linux/audit.h> |
58 | ||
1da177e4 LT |
59 | #include <net/sock.h> |
60 | #include <net/scm.h> | |
61 | ||
62 | #define Nprintk(a...) | |
63 | ||
64 | struct netlink_sock { | |
65 | /* struct sock has to be the first member of netlink_sock */ | |
66 | struct sock sk; | |
67 | u32 pid; | |
68 | unsigned int groups; | |
69 | u32 dst_pid; | |
d629b836 | 70 | u32 dst_group; |
1da177e4 LT |
71 | unsigned long state; |
72 | wait_queue_head_t wait; | |
73 | struct netlink_callback *cb; | |
74 | spinlock_t cb_lock; | |
75 | void (*data_ready)(struct sock *sk, int bytes); | |
77247bbb PM |
76 | struct module *module; |
77 | u32 flags; | |
1da177e4 LT |
78 | }; |
79 | ||
77247bbb PM |
80 | #define NETLINK_KERNEL_SOCKET 0x1 |
81 | ||
1da177e4 LT |
82 | static inline struct netlink_sock *nlk_sk(struct sock *sk) |
83 | { | |
84 | return (struct netlink_sock *)sk; | |
85 | } | |
86 | ||
87 | struct nl_pid_hash { | |
88 | struct hlist_head *table; | |
89 | unsigned long rehash_time; | |
90 | ||
91 | unsigned int mask; | |
92 | unsigned int shift; | |
93 | ||
94 | unsigned int entries; | |
95 | unsigned int max_shift; | |
96 | ||
97 | u32 rnd; | |
98 | }; | |
99 | ||
100 | struct netlink_table { | |
101 | struct nl_pid_hash hash; | |
102 | struct hlist_head mc_list; | |
103 | unsigned int nl_nonroot; | |
77247bbb | 104 | struct module *module; |
ab33a171 | 105 | int registered; |
1da177e4 LT |
106 | }; |
107 | ||
108 | static struct netlink_table *nl_table; | |
109 | ||
110 | static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait); | |
111 | ||
112 | static int netlink_dump(struct sock *sk); | |
113 | static void netlink_destroy_callback(struct netlink_callback *cb); | |
114 | ||
115 | static DEFINE_RWLOCK(nl_table_lock); | |
116 | static atomic_t nl_table_users = ATOMIC_INIT(0); | |
117 | ||
118 | static struct notifier_block *netlink_chain; | |
119 | ||
d629b836 PM |
120 | static u32 netlink_group_mask(u32 group) |
121 | { | |
122 | return group ? 1 << (group - 1) : 0; | |
123 | } | |
124 | ||
1da177e4 LT |
125 | static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid) |
126 | { | |
127 | return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask]; | |
128 | } | |
129 | ||
130 | static void netlink_sock_destruct(struct sock *sk) | |
131 | { | |
132 | skb_queue_purge(&sk->sk_receive_queue); | |
133 | ||
134 | if (!sock_flag(sk, SOCK_DEAD)) { | |
135 | printk("Freeing alive netlink socket %p\n", sk); | |
136 | return; | |
137 | } | |
138 | BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc)); | |
139 | BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc)); | |
140 | BUG_TRAP(!nlk_sk(sk)->cb); | |
141 | } | |
142 | ||
143 | /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP. | |
144 | * Look, when several writers sleep and reader wakes them up, all but one | |
145 | * immediately hit write lock and grab all the cpus. Exclusive sleep solves | |
146 | * this, _but_ remember, it adds useless work on UP machines. | |
147 | */ | |
148 | ||
149 | static void netlink_table_grab(void) | |
150 | { | |
151 | write_lock_bh(&nl_table_lock); | |
152 | ||
153 | if (atomic_read(&nl_table_users)) { | |
154 | DECLARE_WAITQUEUE(wait, current); | |
155 | ||
156 | add_wait_queue_exclusive(&nl_table_wait, &wait); | |
157 | for(;;) { | |
158 | set_current_state(TASK_UNINTERRUPTIBLE); | |
159 | if (atomic_read(&nl_table_users) == 0) | |
160 | break; | |
161 | write_unlock_bh(&nl_table_lock); | |
162 | schedule(); | |
163 | write_lock_bh(&nl_table_lock); | |
164 | } | |
165 | ||
166 | __set_current_state(TASK_RUNNING); | |
167 | remove_wait_queue(&nl_table_wait, &wait); | |
168 | } | |
169 | } | |
170 | ||
171 | static __inline__ void netlink_table_ungrab(void) | |
172 | { | |
173 | write_unlock_bh(&nl_table_lock); | |
174 | wake_up(&nl_table_wait); | |
175 | } | |
176 | ||
177 | static __inline__ void | |
178 | netlink_lock_table(void) | |
179 | { | |
180 | /* read_lock() synchronizes us to netlink_table_grab */ | |
181 | ||
182 | read_lock(&nl_table_lock); | |
183 | atomic_inc(&nl_table_users); | |
184 | read_unlock(&nl_table_lock); | |
185 | } | |
186 | ||
187 | static __inline__ void | |
188 | netlink_unlock_table(void) | |
189 | { | |
190 | if (atomic_dec_and_test(&nl_table_users)) | |
191 | wake_up(&nl_table_wait); | |
192 | } | |
193 | ||
194 | static __inline__ struct sock *netlink_lookup(int protocol, u32 pid) | |
195 | { | |
196 | struct nl_pid_hash *hash = &nl_table[protocol].hash; | |
197 | struct hlist_head *head; | |
198 | struct sock *sk; | |
199 | struct hlist_node *node; | |
200 | ||
201 | read_lock(&nl_table_lock); | |
202 | head = nl_pid_hashfn(hash, pid); | |
203 | sk_for_each(sk, node, head) { | |
204 | if (nlk_sk(sk)->pid == pid) { | |
205 | sock_hold(sk); | |
206 | goto found; | |
207 | } | |
208 | } | |
209 | sk = NULL; | |
210 | found: | |
211 | read_unlock(&nl_table_lock); | |
212 | return sk; | |
213 | } | |
214 | ||
215 | static inline struct hlist_head *nl_pid_hash_alloc(size_t size) | |
216 | { | |
217 | if (size <= PAGE_SIZE) | |
218 | return kmalloc(size, GFP_ATOMIC); | |
219 | else | |
220 | return (struct hlist_head *) | |
221 | __get_free_pages(GFP_ATOMIC, get_order(size)); | |
222 | } | |
223 | ||
224 | static inline void nl_pid_hash_free(struct hlist_head *table, size_t size) | |
225 | { | |
226 | if (size <= PAGE_SIZE) | |
227 | kfree(table); | |
228 | else | |
229 | free_pages((unsigned long)table, get_order(size)); | |
230 | } | |
231 | ||
232 | static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow) | |
233 | { | |
234 | unsigned int omask, mask, shift; | |
235 | size_t osize, size; | |
236 | struct hlist_head *otable, *table; | |
237 | int i; | |
238 | ||
239 | omask = mask = hash->mask; | |
240 | osize = size = (mask + 1) * sizeof(*table); | |
241 | shift = hash->shift; | |
242 | ||
243 | if (grow) { | |
244 | if (++shift > hash->max_shift) | |
245 | return 0; | |
246 | mask = mask * 2 + 1; | |
247 | size *= 2; | |
248 | } | |
249 | ||
250 | table = nl_pid_hash_alloc(size); | |
251 | if (!table) | |
252 | return 0; | |
253 | ||
254 | memset(table, 0, size); | |
255 | otable = hash->table; | |
256 | hash->table = table; | |
257 | hash->mask = mask; | |
258 | hash->shift = shift; | |
259 | get_random_bytes(&hash->rnd, sizeof(hash->rnd)); | |
260 | ||
261 | for (i = 0; i <= omask; i++) { | |
262 | struct sock *sk; | |
263 | struct hlist_node *node, *tmp; | |
264 | ||
265 | sk_for_each_safe(sk, node, tmp, &otable[i]) | |
266 | __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid)); | |
267 | } | |
268 | ||
269 | nl_pid_hash_free(otable, osize); | |
270 | hash->rehash_time = jiffies + 10 * 60 * HZ; | |
271 | return 1; | |
272 | } | |
273 | ||
274 | static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len) | |
275 | { | |
276 | int avg = hash->entries >> hash->shift; | |
277 | ||
278 | if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1)) | |
279 | return 1; | |
280 | ||
281 | if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) { | |
282 | nl_pid_hash_rehash(hash, 0); | |
283 | return 1; | |
284 | } | |
285 | ||
286 | return 0; | |
287 | } | |
288 | ||
289 | static struct proto_ops netlink_ops; | |
290 | ||
291 | static int netlink_insert(struct sock *sk, u32 pid) | |
292 | { | |
293 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
294 | struct hlist_head *head; | |
295 | int err = -EADDRINUSE; | |
296 | struct sock *osk; | |
297 | struct hlist_node *node; | |
298 | int len; | |
299 | ||
300 | netlink_table_grab(); | |
301 | head = nl_pid_hashfn(hash, pid); | |
302 | len = 0; | |
303 | sk_for_each(osk, node, head) { | |
304 | if (nlk_sk(osk)->pid == pid) | |
305 | break; | |
306 | len++; | |
307 | } | |
308 | if (node) | |
309 | goto err; | |
310 | ||
311 | err = -EBUSY; | |
312 | if (nlk_sk(sk)->pid) | |
313 | goto err; | |
314 | ||
315 | err = -ENOMEM; | |
316 | if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX)) | |
317 | goto err; | |
318 | ||
319 | if (len && nl_pid_hash_dilute(hash, len)) | |
320 | head = nl_pid_hashfn(hash, pid); | |
321 | hash->entries++; | |
322 | nlk_sk(sk)->pid = pid; | |
323 | sk_add_node(sk, head); | |
324 | err = 0; | |
325 | ||
326 | err: | |
327 | netlink_table_ungrab(); | |
328 | return err; | |
329 | } | |
330 | ||
331 | static void netlink_remove(struct sock *sk) | |
332 | { | |
333 | netlink_table_grab(); | |
d470e3b4 DM |
334 | if (sk_del_node_init(sk)) |
335 | nl_table[sk->sk_protocol].hash.entries--; | |
1da177e4 LT |
336 | if (nlk_sk(sk)->groups) |
337 | __sk_del_bind_node(sk); | |
338 | netlink_table_ungrab(); | |
339 | } | |
340 | ||
341 | static struct proto netlink_proto = { | |
342 | .name = "NETLINK", | |
343 | .owner = THIS_MODULE, | |
344 | .obj_size = sizeof(struct netlink_sock), | |
345 | }; | |
346 | ||
ab33a171 | 347 | static int __netlink_create(struct socket *sock, int protocol) |
1da177e4 LT |
348 | { |
349 | struct sock *sk; | |
350 | struct netlink_sock *nlk; | |
ab33a171 PM |
351 | |
352 | sock->ops = &netlink_ops; | |
353 | ||
354 | sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1); | |
355 | if (!sk) | |
356 | return -ENOMEM; | |
357 | ||
358 | sock_init_data(sock, sk); | |
359 | ||
360 | nlk = nlk_sk(sk); | |
361 | spin_lock_init(&nlk->cb_lock); | |
362 | init_waitqueue_head(&nlk->wait); | |
363 | ||
364 | sk->sk_destruct = netlink_sock_destruct; | |
365 | sk->sk_protocol = protocol; | |
366 | return 0; | |
367 | } | |
368 | ||
369 | static int netlink_create(struct socket *sock, int protocol) | |
370 | { | |
371 | struct module *module = NULL; | |
372 | int err = 0; | |
1da177e4 LT |
373 | |
374 | sock->state = SS_UNCONNECTED; | |
375 | ||
376 | if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM) | |
377 | return -ESOCKTNOSUPPORT; | |
378 | ||
379 | if (protocol<0 || protocol >= MAX_LINKS) | |
380 | return -EPROTONOSUPPORT; | |
381 | ||
77247bbb | 382 | netlink_lock_table(); |
4fdb3bb7 | 383 | #ifdef CONFIG_KMOD |
ab33a171 | 384 | if (!nl_table[protocol].registered) { |
77247bbb | 385 | netlink_unlock_table(); |
4fdb3bb7 | 386 | request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol); |
77247bbb | 387 | netlink_lock_table(); |
4fdb3bb7 | 388 | } |
ab33a171 PM |
389 | #endif |
390 | if (nl_table[protocol].registered && | |
391 | try_module_get(nl_table[protocol].module)) | |
392 | module = nl_table[protocol].module; | |
393 | else | |
394 | err = -EPROTONOSUPPORT; | |
77247bbb | 395 | netlink_unlock_table(); |
4fdb3bb7 | 396 | |
ab33a171 PM |
397 | if (err) |
398 | goto out; | |
1da177e4 | 399 | |
ab33a171 PM |
400 | if ((err = __netlink_create(sock, protocol) < 0)) |
401 | goto out_module; | |
1da177e4 | 402 | |
ab33a171 PM |
403 | nlk_sk(sock->sk)->module = module; |
404 | out: | |
405 | return err; | |
1da177e4 | 406 | |
ab33a171 PM |
407 | out_module: |
408 | module_put(module); | |
409 | goto out; | |
1da177e4 LT |
410 | } |
411 | ||
412 | static int netlink_release(struct socket *sock) | |
413 | { | |
414 | struct sock *sk = sock->sk; | |
415 | struct netlink_sock *nlk; | |
416 | ||
417 | if (!sk) | |
418 | return 0; | |
419 | ||
420 | netlink_remove(sk); | |
421 | nlk = nlk_sk(sk); | |
422 | ||
423 | spin_lock(&nlk->cb_lock); | |
424 | if (nlk->cb) { | |
425 | nlk->cb->done(nlk->cb); | |
426 | netlink_destroy_callback(nlk->cb); | |
427 | nlk->cb = NULL; | |
1da177e4 LT |
428 | } |
429 | spin_unlock(&nlk->cb_lock); | |
430 | ||
431 | /* OK. Socket is unlinked, and, therefore, | |
432 | no new packets will arrive */ | |
433 | ||
434 | sock_orphan(sk); | |
435 | sock->sk = NULL; | |
436 | wake_up_interruptible_all(&nlk->wait); | |
437 | ||
438 | skb_queue_purge(&sk->sk_write_queue); | |
439 | ||
440 | if (nlk->pid && !nlk->groups) { | |
441 | struct netlink_notify n = { | |
442 | .protocol = sk->sk_protocol, | |
443 | .pid = nlk->pid, | |
444 | }; | |
445 | notifier_call_chain(&netlink_chain, NETLINK_URELEASE, &n); | |
446 | } | |
4fdb3bb7 | 447 | |
77247bbb PM |
448 | if (nlk->module) |
449 | module_put(nlk->module); | |
4fdb3bb7 | 450 | |
77247bbb | 451 | if (nlk->flags & NETLINK_KERNEL_SOCKET) { |
4fdb3bb7 | 452 | netlink_table_grab(); |
77247bbb | 453 | nl_table[sk->sk_protocol].module = NULL; |
ab33a171 | 454 | nl_table[sk->sk_protocol].registered = 0; |
4fdb3bb7 HW |
455 | netlink_table_ungrab(); |
456 | } | |
77247bbb | 457 | |
1da177e4 LT |
458 | sock_put(sk); |
459 | return 0; | |
460 | } | |
461 | ||
462 | static int netlink_autobind(struct socket *sock) | |
463 | { | |
464 | struct sock *sk = sock->sk; | |
465 | struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash; | |
466 | struct hlist_head *head; | |
467 | struct sock *osk; | |
468 | struct hlist_node *node; | |
469 | s32 pid = current->pid; | |
470 | int err; | |
471 | static s32 rover = -4097; | |
472 | ||
473 | retry: | |
474 | cond_resched(); | |
475 | netlink_table_grab(); | |
476 | head = nl_pid_hashfn(hash, pid); | |
477 | sk_for_each(osk, node, head) { | |
478 | if (nlk_sk(osk)->pid == pid) { | |
479 | /* Bind collision, search negative pid values. */ | |
480 | pid = rover--; | |
481 | if (rover > -4097) | |
482 | rover = -4097; | |
483 | netlink_table_ungrab(); | |
484 | goto retry; | |
485 | } | |
486 | } | |
487 | netlink_table_ungrab(); | |
488 | ||
489 | err = netlink_insert(sk, pid); | |
490 | if (err == -EADDRINUSE) | |
491 | goto retry; | |
d470e3b4 DM |
492 | |
493 | /* If 2 threads race to autobind, that is fine. */ | |
494 | if (err == -EBUSY) | |
495 | err = 0; | |
496 | ||
497 | return err; | |
1da177e4 LT |
498 | } |
499 | ||
500 | static inline int netlink_capable(struct socket *sock, unsigned int flag) | |
501 | { | |
502 | return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) || | |
503 | capable(CAP_NET_ADMIN); | |
504 | } | |
505 | ||
506 | static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len) | |
507 | { | |
508 | struct sock *sk = sock->sk; | |
509 | struct netlink_sock *nlk = nlk_sk(sk); | |
510 | struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr; | |
511 | int err; | |
512 | ||
513 | if (nladdr->nl_family != AF_NETLINK) | |
514 | return -EINVAL; | |
515 | ||
516 | /* Only superuser is allowed to listen multicasts */ | |
517 | if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_RECV)) | |
518 | return -EPERM; | |
519 | ||
520 | if (nlk->pid) { | |
521 | if (nladdr->nl_pid != nlk->pid) | |
522 | return -EINVAL; | |
523 | } else { | |
524 | err = nladdr->nl_pid ? | |
525 | netlink_insert(sk, nladdr->nl_pid) : | |
526 | netlink_autobind(sock); | |
527 | if (err) | |
528 | return err; | |
529 | } | |
530 | ||
531 | if (!nladdr->nl_groups && !nlk->groups) | |
532 | return 0; | |
533 | ||
534 | netlink_table_grab(); | |
535 | if (nlk->groups && !nladdr->nl_groups) | |
536 | __sk_del_bind_node(sk); | |
537 | else if (!nlk->groups && nladdr->nl_groups) | |
538 | sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list); | |
539 | nlk->groups = nladdr->nl_groups; | |
540 | netlink_table_ungrab(); | |
541 | ||
542 | return 0; | |
543 | } | |
544 | ||
545 | static int netlink_connect(struct socket *sock, struct sockaddr *addr, | |
546 | int alen, int flags) | |
547 | { | |
548 | int err = 0; | |
549 | struct sock *sk = sock->sk; | |
550 | struct netlink_sock *nlk = nlk_sk(sk); | |
551 | struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr; | |
552 | ||
553 | if (addr->sa_family == AF_UNSPEC) { | |
554 | sk->sk_state = NETLINK_UNCONNECTED; | |
555 | nlk->dst_pid = 0; | |
d629b836 | 556 | nlk->dst_group = 0; |
1da177e4 LT |
557 | return 0; |
558 | } | |
559 | if (addr->sa_family != AF_NETLINK) | |
560 | return -EINVAL; | |
561 | ||
562 | /* Only superuser is allowed to send multicasts */ | |
563 | if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND)) | |
564 | return -EPERM; | |
565 | ||
566 | if (!nlk->pid) | |
567 | err = netlink_autobind(sock); | |
568 | ||
569 | if (err == 0) { | |
570 | sk->sk_state = NETLINK_CONNECTED; | |
571 | nlk->dst_pid = nladdr->nl_pid; | |
d629b836 | 572 | nlk->dst_group = ffs(nladdr->nl_groups); |
1da177e4 LT |
573 | } |
574 | ||
575 | return err; | |
576 | } | |
577 | ||
578 | static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer) | |
579 | { | |
580 | struct sock *sk = sock->sk; | |
581 | struct netlink_sock *nlk = nlk_sk(sk); | |
582 | struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr; | |
583 | ||
584 | nladdr->nl_family = AF_NETLINK; | |
585 | nladdr->nl_pad = 0; | |
586 | *addr_len = sizeof(*nladdr); | |
587 | ||
588 | if (peer) { | |
589 | nladdr->nl_pid = nlk->dst_pid; | |
d629b836 | 590 | nladdr->nl_groups = netlink_group_mask(nlk->dst_group); |
1da177e4 LT |
591 | } else { |
592 | nladdr->nl_pid = nlk->pid; | |
d629b836 | 593 | nladdr->nl_groups = nlk->groups; |
1da177e4 LT |
594 | } |
595 | return 0; | |
596 | } | |
597 | ||
598 | static void netlink_overrun(struct sock *sk) | |
599 | { | |
600 | if (!test_and_set_bit(0, &nlk_sk(sk)->state)) { | |
601 | sk->sk_err = ENOBUFS; | |
602 | sk->sk_error_report(sk); | |
603 | } | |
604 | } | |
605 | ||
606 | static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid) | |
607 | { | |
608 | int protocol = ssk->sk_protocol; | |
609 | struct sock *sock; | |
610 | struct netlink_sock *nlk; | |
611 | ||
612 | sock = netlink_lookup(protocol, pid); | |
613 | if (!sock) | |
614 | return ERR_PTR(-ECONNREFUSED); | |
615 | ||
616 | /* Don't bother queuing skb if kernel socket has no input function */ | |
617 | nlk = nlk_sk(sock); | |
618 | if ((nlk->pid == 0 && !nlk->data_ready) || | |
619 | (sock->sk_state == NETLINK_CONNECTED && | |
620 | nlk->dst_pid != nlk_sk(ssk)->pid)) { | |
621 | sock_put(sock); | |
622 | return ERR_PTR(-ECONNREFUSED); | |
623 | } | |
624 | return sock; | |
625 | } | |
626 | ||
627 | struct sock *netlink_getsockbyfilp(struct file *filp) | |
628 | { | |
629 | struct inode *inode = filp->f_dentry->d_inode; | |
630 | struct sock *sock; | |
631 | ||
632 | if (!S_ISSOCK(inode->i_mode)) | |
633 | return ERR_PTR(-ENOTSOCK); | |
634 | ||
635 | sock = SOCKET_I(inode)->sk; | |
636 | if (sock->sk_family != AF_NETLINK) | |
637 | return ERR_PTR(-EINVAL); | |
638 | ||
639 | sock_hold(sock); | |
640 | return sock; | |
641 | } | |
642 | ||
643 | /* | |
644 | * Attach a skb to a netlink socket. | |
645 | * The caller must hold a reference to the destination socket. On error, the | |
646 | * reference is dropped. The skb is not send to the destination, just all | |
647 | * all error checks are performed and memory in the queue is reserved. | |
648 | * Return values: | |
649 | * < 0: error. skb freed, reference to sock dropped. | |
650 | * 0: continue | |
651 | * 1: repeat lookup - reference dropped while waiting for socket memory. | |
652 | */ | |
653 | int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock, long timeo) | |
654 | { | |
655 | struct netlink_sock *nlk; | |
656 | ||
657 | nlk = nlk_sk(sk); | |
658 | ||
659 | if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
660 | test_bit(0, &nlk->state)) { | |
661 | DECLARE_WAITQUEUE(wait, current); | |
662 | if (!timeo) { | |
663 | if (!nlk->pid) | |
664 | netlink_overrun(sk); | |
665 | sock_put(sk); | |
666 | kfree_skb(skb); | |
667 | return -EAGAIN; | |
668 | } | |
669 | ||
670 | __set_current_state(TASK_INTERRUPTIBLE); | |
671 | add_wait_queue(&nlk->wait, &wait); | |
672 | ||
673 | if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || | |
674 | test_bit(0, &nlk->state)) && | |
675 | !sock_flag(sk, SOCK_DEAD)) | |
676 | timeo = schedule_timeout(timeo); | |
677 | ||
678 | __set_current_state(TASK_RUNNING); | |
679 | remove_wait_queue(&nlk->wait, &wait); | |
680 | sock_put(sk); | |
681 | ||
682 | if (signal_pending(current)) { | |
683 | kfree_skb(skb); | |
684 | return sock_intr_errno(timeo); | |
685 | } | |
686 | return 1; | |
687 | } | |
688 | skb_set_owner_r(skb, sk); | |
689 | return 0; | |
690 | } | |
691 | ||
692 | int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol) | |
693 | { | |
694 | struct netlink_sock *nlk; | |
695 | int len = skb->len; | |
696 | ||
697 | nlk = nlk_sk(sk); | |
698 | ||
699 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
700 | sk->sk_data_ready(sk, len); | |
701 | sock_put(sk); | |
702 | return len; | |
703 | } | |
704 | ||
705 | void netlink_detachskb(struct sock *sk, struct sk_buff *skb) | |
706 | { | |
707 | kfree_skb(skb); | |
708 | sock_put(sk); | |
709 | } | |
710 | ||
37da647d VF |
711 | static inline struct sk_buff *netlink_trim(struct sk_buff *skb, |
712 | unsigned int __nocast allocation) | |
1da177e4 LT |
713 | { |
714 | int delta; | |
715 | ||
716 | skb_orphan(skb); | |
717 | ||
718 | delta = skb->end - skb->tail; | |
719 | if (delta * 2 < skb->truesize) | |
720 | return skb; | |
721 | ||
722 | if (skb_shared(skb)) { | |
723 | struct sk_buff *nskb = skb_clone(skb, allocation); | |
724 | if (!nskb) | |
725 | return skb; | |
726 | kfree_skb(skb); | |
727 | skb = nskb; | |
728 | } | |
729 | ||
730 | if (!pskb_expand_head(skb, 0, -delta, allocation)) | |
731 | skb->truesize -= delta; | |
732 | ||
733 | return skb; | |
734 | } | |
735 | ||
736 | int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock) | |
737 | { | |
738 | struct sock *sk; | |
739 | int err; | |
740 | long timeo; | |
741 | ||
742 | skb = netlink_trim(skb, gfp_any()); | |
743 | ||
744 | timeo = sock_sndtimeo(ssk, nonblock); | |
745 | retry: | |
746 | sk = netlink_getsockbypid(ssk, pid); | |
747 | if (IS_ERR(sk)) { | |
748 | kfree_skb(skb); | |
749 | return PTR_ERR(sk); | |
750 | } | |
751 | err = netlink_attachskb(sk, skb, nonblock, timeo); | |
752 | if (err == 1) | |
753 | goto retry; | |
754 | if (err) | |
755 | return err; | |
756 | ||
757 | return netlink_sendskb(sk, skb, ssk->sk_protocol); | |
758 | } | |
759 | ||
760 | static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) | |
761 | { | |
762 | struct netlink_sock *nlk = nlk_sk(sk); | |
763 | ||
764 | if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && | |
765 | !test_bit(0, &nlk->state)) { | |
766 | skb_set_owner_r(skb, sk); | |
767 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
768 | sk->sk_data_ready(sk, skb->len); | |
769 | return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf; | |
770 | } | |
771 | return -1; | |
772 | } | |
773 | ||
774 | struct netlink_broadcast_data { | |
775 | struct sock *exclude_sk; | |
776 | u32 pid; | |
777 | u32 group; | |
778 | int failure; | |
779 | int congested; | |
780 | int delivered; | |
37da647d | 781 | unsigned int allocation; |
1da177e4 LT |
782 | struct sk_buff *skb, *skb2; |
783 | }; | |
784 | ||
785 | static inline int do_one_broadcast(struct sock *sk, | |
786 | struct netlink_broadcast_data *p) | |
787 | { | |
788 | struct netlink_sock *nlk = nlk_sk(sk); | |
789 | int val; | |
790 | ||
791 | if (p->exclude_sk == sk) | |
792 | goto out; | |
793 | ||
d629b836 | 794 | if (nlk->pid == p->pid || !(nlk->groups & netlink_group_mask(p->group))) |
1da177e4 LT |
795 | goto out; |
796 | ||
797 | if (p->failure) { | |
798 | netlink_overrun(sk); | |
799 | goto out; | |
800 | } | |
801 | ||
802 | sock_hold(sk); | |
803 | if (p->skb2 == NULL) { | |
68acc024 | 804 | if (skb_shared(p->skb)) { |
1da177e4 LT |
805 | p->skb2 = skb_clone(p->skb, p->allocation); |
806 | } else { | |
68acc024 TC |
807 | p->skb2 = skb_get(p->skb); |
808 | /* | |
809 | * skb ownership may have been set when | |
810 | * delivered to a previous socket. | |
811 | */ | |
812 | skb_orphan(p->skb2); | |
1da177e4 LT |
813 | } |
814 | } | |
815 | if (p->skb2 == NULL) { | |
816 | netlink_overrun(sk); | |
817 | /* Clone failed. Notify ALL listeners. */ | |
818 | p->failure = 1; | |
819 | } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) { | |
820 | netlink_overrun(sk); | |
821 | } else { | |
822 | p->congested |= val; | |
823 | p->delivered = 1; | |
824 | p->skb2 = NULL; | |
825 | } | |
826 | sock_put(sk); | |
827 | ||
828 | out: | |
829 | return 0; | |
830 | } | |
831 | ||
832 | int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid, | |
833 | u32 group, int allocation) | |
834 | { | |
835 | struct netlink_broadcast_data info; | |
836 | struct hlist_node *node; | |
837 | struct sock *sk; | |
838 | ||
839 | skb = netlink_trim(skb, allocation); | |
840 | ||
841 | info.exclude_sk = ssk; | |
842 | info.pid = pid; | |
843 | info.group = group; | |
844 | info.failure = 0; | |
845 | info.congested = 0; | |
846 | info.delivered = 0; | |
847 | info.allocation = allocation; | |
848 | info.skb = skb; | |
849 | info.skb2 = NULL; | |
850 | ||
851 | /* While we sleep in clone, do not allow to change socket list */ | |
852 | ||
853 | netlink_lock_table(); | |
854 | ||
855 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
856 | do_one_broadcast(sk, &info); | |
857 | ||
aa1c6a6f TC |
858 | kfree_skb(skb); |
859 | ||
1da177e4 LT |
860 | netlink_unlock_table(); |
861 | ||
862 | if (info.skb2) | |
863 | kfree_skb(info.skb2); | |
1da177e4 LT |
864 | |
865 | if (info.delivered) { | |
866 | if (info.congested && (allocation & __GFP_WAIT)) | |
867 | yield(); | |
868 | return 0; | |
869 | } | |
870 | if (info.failure) | |
871 | return -ENOBUFS; | |
872 | return -ESRCH; | |
873 | } | |
874 | ||
875 | struct netlink_set_err_data { | |
876 | struct sock *exclude_sk; | |
877 | u32 pid; | |
878 | u32 group; | |
879 | int code; | |
880 | }; | |
881 | ||
882 | static inline int do_one_set_err(struct sock *sk, | |
883 | struct netlink_set_err_data *p) | |
884 | { | |
885 | struct netlink_sock *nlk = nlk_sk(sk); | |
886 | ||
887 | if (sk == p->exclude_sk) | |
888 | goto out; | |
889 | ||
d629b836 | 890 | if (nlk->pid == p->pid || !(nlk->groups & netlink_group_mask(p->group))) |
1da177e4 LT |
891 | goto out; |
892 | ||
893 | sk->sk_err = p->code; | |
894 | sk->sk_error_report(sk); | |
895 | out: | |
896 | return 0; | |
897 | } | |
898 | ||
899 | void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code) | |
900 | { | |
901 | struct netlink_set_err_data info; | |
902 | struct hlist_node *node; | |
903 | struct sock *sk; | |
904 | ||
905 | info.exclude_sk = ssk; | |
906 | info.pid = pid; | |
907 | info.group = group; | |
908 | info.code = code; | |
909 | ||
910 | read_lock(&nl_table_lock); | |
911 | ||
912 | sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list) | |
913 | do_one_set_err(sk, &info); | |
914 | ||
915 | read_unlock(&nl_table_lock); | |
916 | } | |
917 | ||
918 | static inline void netlink_rcv_wake(struct sock *sk) | |
919 | { | |
920 | struct netlink_sock *nlk = nlk_sk(sk); | |
921 | ||
b03efcfb | 922 | if (skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
923 | clear_bit(0, &nlk->state); |
924 | if (!test_bit(0, &nlk->state)) | |
925 | wake_up_interruptible(&nlk->wait); | |
926 | } | |
927 | ||
928 | static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock, | |
929 | struct msghdr *msg, size_t len) | |
930 | { | |
931 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
932 | struct sock *sk = sock->sk; | |
933 | struct netlink_sock *nlk = nlk_sk(sk); | |
934 | struct sockaddr_nl *addr=msg->msg_name; | |
935 | u32 dst_pid; | |
d629b836 | 936 | u32 dst_group; |
1da177e4 LT |
937 | struct sk_buff *skb; |
938 | int err; | |
939 | struct scm_cookie scm; | |
940 | ||
941 | if (msg->msg_flags&MSG_OOB) | |
942 | return -EOPNOTSUPP; | |
943 | ||
944 | if (NULL == siocb->scm) | |
945 | siocb->scm = &scm; | |
946 | err = scm_send(sock, msg, siocb->scm); | |
947 | if (err < 0) | |
948 | return err; | |
949 | ||
950 | if (msg->msg_namelen) { | |
951 | if (addr->nl_family != AF_NETLINK) | |
952 | return -EINVAL; | |
953 | dst_pid = addr->nl_pid; | |
d629b836 PM |
954 | dst_group = ffs(addr->nl_groups); |
955 | if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND)) | |
1da177e4 LT |
956 | return -EPERM; |
957 | } else { | |
958 | dst_pid = nlk->dst_pid; | |
d629b836 | 959 | dst_group = nlk->dst_group; |
1da177e4 LT |
960 | } |
961 | ||
962 | if (!nlk->pid) { | |
963 | err = netlink_autobind(sock); | |
964 | if (err) | |
965 | goto out; | |
966 | } | |
967 | ||
968 | err = -EMSGSIZE; | |
969 | if (len > sk->sk_sndbuf - 32) | |
970 | goto out; | |
971 | err = -ENOBUFS; | |
972 | skb = alloc_skb(len, GFP_KERNEL); | |
973 | if (skb==NULL) | |
974 | goto out; | |
975 | ||
976 | NETLINK_CB(skb).pid = nlk->pid; | |
1da177e4 | 977 | NETLINK_CB(skb).dst_pid = dst_pid; |
d629b836 | 978 | NETLINK_CB(skb).dst_group = dst_group; |
c94c257c | 979 | NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context); |
1da177e4 LT |
980 | memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred)); |
981 | ||
982 | /* What can I do? Netlink is asynchronous, so that | |
983 | we will have to save current capabilities to | |
984 | check them, when this message will be delivered | |
985 | to corresponding kernel module. --ANK (980802) | |
986 | */ | |
987 | ||
988 | err = -EFAULT; | |
989 | if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) { | |
990 | kfree_skb(skb); | |
991 | goto out; | |
992 | } | |
993 | ||
994 | err = security_netlink_send(sk, skb); | |
995 | if (err) { | |
996 | kfree_skb(skb); | |
997 | goto out; | |
998 | } | |
999 | ||
d629b836 | 1000 | if (dst_group) { |
1da177e4 | 1001 | atomic_inc(&skb->users); |
d629b836 | 1002 | netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL); |
1da177e4 LT |
1003 | } |
1004 | err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT); | |
1005 | ||
1006 | out: | |
1007 | return err; | |
1008 | } | |
1009 | ||
1010 | static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock, | |
1011 | struct msghdr *msg, size_t len, | |
1012 | int flags) | |
1013 | { | |
1014 | struct sock_iocb *siocb = kiocb_to_siocb(kiocb); | |
1015 | struct scm_cookie scm; | |
1016 | struct sock *sk = sock->sk; | |
1017 | struct netlink_sock *nlk = nlk_sk(sk); | |
1018 | int noblock = flags&MSG_DONTWAIT; | |
1019 | size_t copied; | |
1020 | struct sk_buff *skb; | |
1021 | int err; | |
1022 | ||
1023 | if (flags&MSG_OOB) | |
1024 | return -EOPNOTSUPP; | |
1025 | ||
1026 | copied = 0; | |
1027 | ||
1028 | skb = skb_recv_datagram(sk,flags,noblock,&err); | |
1029 | if (skb==NULL) | |
1030 | goto out; | |
1031 | ||
1032 | msg->msg_namelen = 0; | |
1033 | ||
1034 | copied = skb->len; | |
1035 | if (len < copied) { | |
1036 | msg->msg_flags |= MSG_TRUNC; | |
1037 | copied = len; | |
1038 | } | |
1039 | ||
1040 | skb->h.raw = skb->data; | |
1041 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); | |
1042 | ||
1043 | if (msg->msg_name) { | |
1044 | struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name; | |
1045 | addr->nl_family = AF_NETLINK; | |
1046 | addr->nl_pad = 0; | |
1047 | addr->nl_pid = NETLINK_CB(skb).pid; | |
d629b836 | 1048 | addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group); |
1da177e4 LT |
1049 | msg->msg_namelen = sizeof(*addr); |
1050 | } | |
1051 | ||
1052 | if (NULL == siocb->scm) { | |
1053 | memset(&scm, 0, sizeof(scm)); | |
1054 | siocb->scm = &scm; | |
1055 | } | |
1056 | siocb->scm->creds = *NETLINK_CREDS(skb); | |
1057 | skb_free_datagram(sk, skb); | |
1058 | ||
1059 | if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) | |
1060 | netlink_dump(sk); | |
1061 | ||
1062 | scm_recv(sock, msg, siocb->scm, flags); | |
1063 | ||
1064 | out: | |
1065 | netlink_rcv_wake(sk); | |
1066 | return err ? : copied; | |
1067 | } | |
1068 | ||
1069 | static void netlink_data_ready(struct sock *sk, int len) | |
1070 | { | |
1071 | struct netlink_sock *nlk = nlk_sk(sk); | |
1072 | ||
1073 | if (nlk->data_ready) | |
1074 | nlk->data_ready(sk, len); | |
1075 | netlink_rcv_wake(sk); | |
1076 | } | |
1077 | ||
1078 | /* | |
1079 | * We export these functions to other modules. They provide a | |
1080 | * complete set of kernel non-blocking support for message | |
1081 | * queueing. | |
1082 | */ | |
1083 | ||
1084 | struct sock * | |
4fdb3bb7 | 1085 | netlink_kernel_create(int unit, void (*input)(struct sock *sk, int len), struct module *module) |
1da177e4 LT |
1086 | { |
1087 | struct socket *sock; | |
1088 | struct sock *sk; | |
77247bbb | 1089 | struct netlink_sock *nlk; |
1da177e4 LT |
1090 | |
1091 | if (!nl_table) | |
1092 | return NULL; | |
1093 | ||
1094 | if (unit<0 || unit>=MAX_LINKS) | |
1095 | return NULL; | |
1096 | ||
1097 | if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock)) | |
1098 | return NULL; | |
1099 | ||
ab33a171 | 1100 | if (__netlink_create(sock, unit) < 0) |
77247bbb | 1101 | goto out_sock_release; |
4fdb3bb7 | 1102 | |
1da177e4 LT |
1103 | sk = sock->sk; |
1104 | sk->sk_data_ready = netlink_data_ready; | |
1105 | if (input) | |
1106 | nlk_sk(sk)->data_ready = input; | |
1107 | ||
77247bbb PM |
1108 | if (netlink_insert(sk, 0)) |
1109 | goto out_sock_release; | |
4fdb3bb7 | 1110 | |
77247bbb PM |
1111 | nlk = nlk_sk(sk); |
1112 | nlk->flags |= NETLINK_KERNEL_SOCKET; | |
4fdb3bb7 | 1113 | |
4fdb3bb7 | 1114 | netlink_table_grab(); |
77247bbb | 1115 | nl_table[unit].module = module; |
ab33a171 | 1116 | nl_table[unit].registered = 1; |
4fdb3bb7 | 1117 | netlink_table_ungrab(); |
77247bbb PM |
1118 | |
1119 | return sk; | |
1120 | ||
4fdb3bb7 HW |
1121 | out_sock_release: |
1122 | sock_release(sock); | |
77247bbb | 1123 | return NULL; |
1da177e4 LT |
1124 | } |
1125 | ||
1126 | void netlink_set_nonroot(int protocol, unsigned int flags) | |
1127 | { | |
1128 | if ((unsigned int)protocol < MAX_LINKS) | |
1129 | nl_table[protocol].nl_nonroot = flags; | |
1130 | } | |
1131 | ||
1132 | static void netlink_destroy_callback(struct netlink_callback *cb) | |
1133 | { | |
1134 | if (cb->skb) | |
1135 | kfree_skb(cb->skb); | |
1136 | kfree(cb); | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * It looks a bit ugly. | |
1141 | * It would be better to create kernel thread. | |
1142 | */ | |
1143 | ||
1144 | static int netlink_dump(struct sock *sk) | |
1145 | { | |
1146 | struct netlink_sock *nlk = nlk_sk(sk); | |
1147 | struct netlink_callback *cb; | |
1148 | struct sk_buff *skb; | |
1149 | struct nlmsghdr *nlh; | |
1150 | int len; | |
1151 | ||
1152 | skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL); | |
1153 | if (!skb) | |
1154 | return -ENOBUFS; | |
1155 | ||
1156 | spin_lock(&nlk->cb_lock); | |
1157 | ||
1158 | cb = nlk->cb; | |
1159 | if (cb == NULL) { | |
1160 | spin_unlock(&nlk->cb_lock); | |
1161 | kfree_skb(skb); | |
1162 | return -EINVAL; | |
1163 | } | |
1164 | ||
1165 | len = cb->dump(skb, cb); | |
1166 | ||
1167 | if (len > 0) { | |
1168 | spin_unlock(&nlk->cb_lock); | |
1169 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1170 | sk->sk_data_ready(sk, len); | |
1171 | return 0; | |
1172 | } | |
1173 | ||
1797754e | 1174 | nlh = NLMSG_NEW_ANSWER(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI); |
1da177e4 LT |
1175 | memcpy(NLMSG_DATA(nlh), &len, sizeof(len)); |
1176 | skb_queue_tail(&sk->sk_receive_queue, skb); | |
1177 | sk->sk_data_ready(sk, skb->len); | |
1178 | ||
1179 | cb->done(cb); | |
1180 | nlk->cb = NULL; | |
1181 | spin_unlock(&nlk->cb_lock); | |
1182 | ||
1183 | netlink_destroy_callback(cb); | |
1da177e4 | 1184 | return 0; |
1797754e TG |
1185 | |
1186 | nlmsg_failure: | |
1187 | return -ENOBUFS; | |
1da177e4 LT |
1188 | } |
1189 | ||
1190 | int netlink_dump_start(struct sock *ssk, struct sk_buff *skb, | |
1191 | struct nlmsghdr *nlh, | |
1192 | int (*dump)(struct sk_buff *skb, struct netlink_callback*), | |
1193 | int (*done)(struct netlink_callback*)) | |
1194 | { | |
1195 | struct netlink_callback *cb; | |
1196 | struct sock *sk; | |
1197 | struct netlink_sock *nlk; | |
1198 | ||
1199 | cb = kmalloc(sizeof(*cb), GFP_KERNEL); | |
1200 | if (cb == NULL) | |
1201 | return -ENOBUFS; | |
1202 | ||
1203 | memset(cb, 0, sizeof(*cb)); | |
1204 | cb->dump = dump; | |
1205 | cb->done = done; | |
1206 | cb->nlh = nlh; | |
1207 | atomic_inc(&skb->users); | |
1208 | cb->skb = skb; | |
1209 | ||
1210 | sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid); | |
1211 | if (sk == NULL) { | |
1212 | netlink_destroy_callback(cb); | |
1213 | return -ECONNREFUSED; | |
1214 | } | |
1215 | nlk = nlk_sk(sk); | |
1216 | /* A dump is in progress... */ | |
1217 | spin_lock(&nlk->cb_lock); | |
1218 | if (nlk->cb) { | |
1219 | spin_unlock(&nlk->cb_lock); | |
1220 | netlink_destroy_callback(cb); | |
1221 | sock_put(sk); | |
1222 | return -EBUSY; | |
1223 | } | |
1224 | nlk->cb = cb; | |
1da177e4 LT |
1225 | spin_unlock(&nlk->cb_lock); |
1226 | ||
1227 | netlink_dump(sk); | |
1228 | sock_put(sk); | |
1229 | return 0; | |
1230 | } | |
1231 | ||
1232 | void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err) | |
1233 | { | |
1234 | struct sk_buff *skb; | |
1235 | struct nlmsghdr *rep; | |
1236 | struct nlmsgerr *errmsg; | |
1237 | int size; | |
1238 | ||
1239 | if (err == 0) | |
1240 | size = NLMSG_SPACE(sizeof(struct nlmsgerr)); | |
1241 | else | |
1242 | size = NLMSG_SPACE(4 + NLMSG_ALIGN(nlh->nlmsg_len)); | |
1243 | ||
1244 | skb = alloc_skb(size, GFP_KERNEL); | |
1245 | if (!skb) { | |
1246 | struct sock *sk; | |
1247 | ||
1248 | sk = netlink_lookup(in_skb->sk->sk_protocol, | |
1249 | NETLINK_CB(in_skb).pid); | |
1250 | if (sk) { | |
1251 | sk->sk_err = ENOBUFS; | |
1252 | sk->sk_error_report(sk); | |
1253 | sock_put(sk); | |
1254 | } | |
1255 | return; | |
1256 | } | |
1257 | ||
1258 | rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, | |
1797754e | 1259 | NLMSG_ERROR, sizeof(struct nlmsgerr), 0); |
1da177e4 LT |
1260 | errmsg = NLMSG_DATA(rep); |
1261 | errmsg->error = err; | |
1262 | memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(struct nlmsghdr)); | |
1263 | netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT); | |
1264 | } | |
1265 | ||
1266 | ||
1267 | #ifdef CONFIG_PROC_FS | |
1268 | struct nl_seq_iter { | |
1269 | int link; | |
1270 | int hash_idx; | |
1271 | }; | |
1272 | ||
1273 | static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos) | |
1274 | { | |
1275 | struct nl_seq_iter *iter = seq->private; | |
1276 | int i, j; | |
1277 | struct sock *s; | |
1278 | struct hlist_node *node; | |
1279 | loff_t off = 0; | |
1280 | ||
1281 | for (i=0; i<MAX_LINKS; i++) { | |
1282 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1283 | ||
1284 | for (j = 0; j <= hash->mask; j++) { | |
1285 | sk_for_each(s, node, &hash->table[j]) { | |
1286 | if (off == pos) { | |
1287 | iter->link = i; | |
1288 | iter->hash_idx = j; | |
1289 | return s; | |
1290 | } | |
1291 | ++off; | |
1292 | } | |
1293 | } | |
1294 | } | |
1295 | return NULL; | |
1296 | } | |
1297 | ||
1298 | static void *netlink_seq_start(struct seq_file *seq, loff_t *pos) | |
1299 | { | |
1300 | read_lock(&nl_table_lock); | |
1301 | return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
1302 | } | |
1303 | ||
1304 | static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
1305 | { | |
1306 | struct sock *s; | |
1307 | struct nl_seq_iter *iter; | |
1308 | int i, j; | |
1309 | ||
1310 | ++*pos; | |
1311 | ||
1312 | if (v == SEQ_START_TOKEN) | |
1313 | return netlink_seq_socket_idx(seq, 0); | |
1314 | ||
1315 | s = sk_next(v); | |
1316 | if (s) | |
1317 | return s; | |
1318 | ||
1319 | iter = seq->private; | |
1320 | i = iter->link; | |
1321 | j = iter->hash_idx + 1; | |
1322 | ||
1323 | do { | |
1324 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1325 | ||
1326 | for (; j <= hash->mask; j++) { | |
1327 | s = sk_head(&hash->table[j]); | |
1328 | if (s) { | |
1329 | iter->link = i; | |
1330 | iter->hash_idx = j; | |
1331 | return s; | |
1332 | } | |
1333 | } | |
1334 | ||
1335 | j = 0; | |
1336 | } while (++i < MAX_LINKS); | |
1337 | ||
1338 | return NULL; | |
1339 | } | |
1340 | ||
1341 | static void netlink_seq_stop(struct seq_file *seq, void *v) | |
1342 | { | |
1343 | read_unlock(&nl_table_lock); | |
1344 | } | |
1345 | ||
1346 | ||
1347 | static int netlink_seq_show(struct seq_file *seq, void *v) | |
1348 | { | |
1349 | if (v == SEQ_START_TOKEN) | |
1350 | seq_puts(seq, | |
1351 | "sk Eth Pid Groups " | |
1352 | "Rmem Wmem Dump Locks\n"); | |
1353 | else { | |
1354 | struct sock *s = v; | |
1355 | struct netlink_sock *nlk = nlk_sk(s); | |
1356 | ||
1357 | seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n", | |
1358 | s, | |
1359 | s->sk_protocol, | |
1360 | nlk->pid, | |
1361 | nlk->groups, | |
1362 | atomic_read(&s->sk_rmem_alloc), | |
1363 | atomic_read(&s->sk_wmem_alloc), | |
1364 | nlk->cb, | |
1365 | atomic_read(&s->sk_refcnt) | |
1366 | ); | |
1367 | ||
1368 | } | |
1369 | return 0; | |
1370 | } | |
1371 | ||
1372 | static struct seq_operations netlink_seq_ops = { | |
1373 | .start = netlink_seq_start, | |
1374 | .next = netlink_seq_next, | |
1375 | .stop = netlink_seq_stop, | |
1376 | .show = netlink_seq_show, | |
1377 | }; | |
1378 | ||
1379 | ||
1380 | static int netlink_seq_open(struct inode *inode, struct file *file) | |
1381 | { | |
1382 | struct seq_file *seq; | |
1383 | struct nl_seq_iter *iter; | |
1384 | int err; | |
1385 | ||
1386 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); | |
1387 | if (!iter) | |
1388 | return -ENOMEM; | |
1389 | ||
1390 | err = seq_open(file, &netlink_seq_ops); | |
1391 | if (err) { | |
1392 | kfree(iter); | |
1393 | return err; | |
1394 | } | |
1395 | ||
1396 | memset(iter, 0, sizeof(*iter)); | |
1397 | seq = file->private_data; | |
1398 | seq->private = iter; | |
1399 | return 0; | |
1400 | } | |
1401 | ||
1402 | static struct file_operations netlink_seq_fops = { | |
1403 | .owner = THIS_MODULE, | |
1404 | .open = netlink_seq_open, | |
1405 | .read = seq_read, | |
1406 | .llseek = seq_lseek, | |
1407 | .release = seq_release_private, | |
1408 | }; | |
1409 | ||
1410 | #endif | |
1411 | ||
1412 | int netlink_register_notifier(struct notifier_block *nb) | |
1413 | { | |
1414 | return notifier_chain_register(&netlink_chain, nb); | |
1415 | } | |
1416 | ||
1417 | int netlink_unregister_notifier(struct notifier_block *nb) | |
1418 | { | |
1419 | return notifier_chain_unregister(&netlink_chain, nb); | |
1420 | } | |
1421 | ||
1422 | static struct proto_ops netlink_ops = { | |
1423 | .family = PF_NETLINK, | |
1424 | .owner = THIS_MODULE, | |
1425 | .release = netlink_release, | |
1426 | .bind = netlink_bind, | |
1427 | .connect = netlink_connect, | |
1428 | .socketpair = sock_no_socketpair, | |
1429 | .accept = sock_no_accept, | |
1430 | .getname = netlink_getname, | |
1431 | .poll = datagram_poll, | |
1432 | .ioctl = sock_no_ioctl, | |
1433 | .listen = sock_no_listen, | |
1434 | .shutdown = sock_no_shutdown, | |
1435 | .setsockopt = sock_no_setsockopt, | |
1436 | .getsockopt = sock_no_getsockopt, | |
1437 | .sendmsg = netlink_sendmsg, | |
1438 | .recvmsg = netlink_recvmsg, | |
1439 | .mmap = sock_no_mmap, | |
1440 | .sendpage = sock_no_sendpage, | |
1441 | }; | |
1442 | ||
1443 | static struct net_proto_family netlink_family_ops = { | |
1444 | .family = PF_NETLINK, | |
1445 | .create = netlink_create, | |
1446 | .owner = THIS_MODULE, /* for consistency 8) */ | |
1447 | }; | |
1448 | ||
1449 | extern void netlink_skb_parms_too_large(void); | |
1450 | ||
1451 | static int __init netlink_proto_init(void) | |
1452 | { | |
1453 | struct sk_buff *dummy_skb; | |
1454 | int i; | |
1455 | unsigned long max; | |
1456 | unsigned int order; | |
1457 | int err = proto_register(&netlink_proto, 0); | |
1458 | ||
1459 | if (err != 0) | |
1460 | goto out; | |
1461 | ||
1462 | if (sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb)) | |
1463 | netlink_skb_parms_too_large(); | |
1464 | ||
1465 | nl_table = kmalloc(sizeof(*nl_table) * MAX_LINKS, GFP_KERNEL); | |
1466 | if (!nl_table) { | |
1467 | enomem: | |
1468 | printk(KERN_CRIT "netlink_init: Cannot allocate nl_table\n"); | |
1469 | return -ENOMEM; | |
1470 | } | |
1471 | ||
1472 | memset(nl_table, 0, sizeof(*nl_table) * MAX_LINKS); | |
1473 | ||
1474 | if (num_physpages >= (128 * 1024)) | |
1475 | max = num_physpages >> (21 - PAGE_SHIFT); | |
1476 | else | |
1477 | max = num_physpages >> (23 - PAGE_SHIFT); | |
1478 | ||
1479 | order = get_bitmask_order(max) - 1 + PAGE_SHIFT; | |
1480 | max = (1UL << order) / sizeof(struct hlist_head); | |
1481 | order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1; | |
1482 | ||
1483 | for (i = 0; i < MAX_LINKS; i++) { | |
1484 | struct nl_pid_hash *hash = &nl_table[i].hash; | |
1485 | ||
1486 | hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table)); | |
1487 | if (!hash->table) { | |
1488 | while (i-- > 0) | |
1489 | nl_pid_hash_free(nl_table[i].hash.table, | |
1490 | 1 * sizeof(*hash->table)); | |
1491 | kfree(nl_table); | |
1492 | goto enomem; | |
1493 | } | |
1494 | memset(hash->table, 0, 1 * sizeof(*hash->table)); | |
1495 | hash->max_shift = order; | |
1496 | hash->shift = 0; | |
1497 | hash->mask = 0; | |
1498 | hash->rehash_time = jiffies; | |
1499 | } | |
1500 | ||
1501 | sock_register(&netlink_family_ops); | |
1502 | #ifdef CONFIG_PROC_FS | |
1503 | proc_net_fops_create("netlink", 0, &netlink_seq_fops); | |
1504 | #endif | |
1505 | /* The netlink device handler may be needed early. */ | |
1506 | rtnetlink_init(); | |
1507 | out: | |
1508 | return err; | |
1509 | } | |
1510 | ||
1da177e4 | 1511 | core_initcall(netlink_proto_init); |
1da177e4 LT |
1512 | |
1513 | EXPORT_SYMBOL(netlink_ack); | |
1514 | EXPORT_SYMBOL(netlink_broadcast); | |
1515 | EXPORT_SYMBOL(netlink_dump_start); | |
1516 | EXPORT_SYMBOL(netlink_kernel_create); | |
1517 | EXPORT_SYMBOL(netlink_register_notifier); | |
1518 | EXPORT_SYMBOL(netlink_set_err); | |
1519 | EXPORT_SYMBOL(netlink_set_nonroot); | |
1520 | EXPORT_SYMBOL(netlink_unicast); | |
1521 | EXPORT_SYMBOL(netlink_unregister_notifier); | |
1522 |