?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / netlink / af_netlink.c
blob6b9772d95872245b71cd305eda34f57335ac406c
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)
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
22 */
23
24 #include <linux/config.h>
25 #include <linux/module.h>
26
27 #include <linux/capability.h>
28 #include <linux/kernel.h>
29 #include <linux/init.h>
30 #include <linux/signal.h>
31 #include <linux/sched.h>
32 #include <linux/errno.h>
33 #include <linux/string.h>
34 #include <linux/stat.h>
35 #include <linux/socket.h>
36 #include <linux/un.h>
37 #include <linux/fcntl.h>
38 #include <linux/termios.h>
39 #include <linux/sockios.h>
40 #include <linux/net.h>
41 #include <linux/fs.h>
42 #include <linux/slab.h>
43 #include <asm/uaccess.h>
44 #include <linux/skbuff.h>
45 #include <linux/netdevice.h>
46 #include <linux/rtnetlink.h>
47 #include <linux/proc_fs.h>
48 #include <linux/seq_file.h>
49 #include <linux/smp_lock.h>
50 #include <linux/notifier.h>
51 #include <linux/security.h>
52 #include <linux/jhash.h>
53 #include <linux/jiffies.h>
54 #include <linux/random.h>
55 #include <linux/bitops.h>
56 #include <linux/mm.h>
57 #include <linux/types.h>
58 #include <linux/audit.h>
59
60 #include <net/sock.h>
61 #include <net/scm.h>
62 #include <net/netlink.h>
63
64 #define Nprintk(a...)
65 #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
66
67 struct netlink_sock {
68 /* struct sock has to be the first member of netlink_sock */
69 struct sock sk;
70 u32 pid;
71 u32 dst_pid;
72 u32 dst_group;
73 u32 flags;
74 u32 subscriptions;
75 u32 ngroups;
76 unsigned long *groups;
77 unsigned long state;
78 wait_queue_head_t wait;
79 struct netlink_callback *cb;
80 spinlock_t cb_lock;
81 void (*data_ready)(struct sock *sk, int bytes);
82 struct module *module;
83 };
84
85 #define NETLINK_KERNEL_SOCKET 0x1
86 #define NETLINK_RECV_PKTINFO 0x2
87
88 static inline struct netlink_sock *nlk_sk(struct sock *sk)
89 {
90 return (struct netlink_sock *)sk;
91 }
92
93 struct nl_pid_hash {
94 struct hlist_head *table;
95 unsigned long rehash_time;
96
97 unsigned int mask;
98 unsigned int shift;
99
100 unsigned int entries;
101 unsigned int max_shift;
102
103 u32 rnd;
104 };
105
106 struct netlink_table {
107 struct nl_pid_hash hash;
108 struct hlist_head mc_list;
109 unsigned int nl_nonroot;
110 unsigned int groups;
111 struct module *module;
112 int registered;
113 };
114
115 static struct netlink_table *nl_table;
116
117 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
118
119 static int netlink_dump(struct sock *sk);
120 static void netlink_destroy_callback(struct netlink_callback *cb);
121
122 static DEFINE_RWLOCK(nl_table_lock);
123 static atomic_t nl_table_users = ATOMIC_INIT(0);
124
125 static struct notifier_block *netlink_chain;
126
127 static u32 netlink_group_mask(u32 group)
128 {
129 return group ? 1 << (group - 1) : 0;
130 }
131
132 static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
133 {
134 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
135 }
136
137 static void netlink_sock_destruct(struct sock *sk)
138 {
139 skb_queue_purge(&sk->sk_receive_queue);
140
141 if (!sock_flag(sk, SOCK_DEAD)) {
142 printk("Freeing alive netlink socket %p\n", sk);
143 return;
144 }
145 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
146 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
147 BUG_TRAP(!nlk_sk(sk)->cb);
148 BUG_TRAP(!nlk_sk(sk)->groups);
149 }
150
151 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
152 * Look, when several writers sleep and reader wakes them up, all but one
153 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
154 * this, _but_ remember, it adds useless work on UP machines.
155 */
156
157 static void netlink_table_grab(void)
158 {
159 write_lock_bh(&nl_table_lock);
160
161 if (atomic_read(&nl_table_users)) {
162 DECLARE_WAITQUEUE(wait, current);
163
164 add_wait_queue_exclusive(&nl_table_wait, &wait);
165 for(;;) {
166 set_current_state(TASK_UNINTERRUPTIBLE);
167 if (atomic_read(&nl_table_users) == 0)
168 break;
169 write_unlock_bh(&nl_table_lock);
170 schedule();
171 write_lock_bh(&nl_table_lock);
172 }
173
174 __set_current_state(TASK_RUNNING);
175 remove_wait_queue(&nl_table_wait, &wait);
176 }
177 }
178
179 static __inline__ void netlink_table_ungrab(void)
180 {
181 write_unlock_bh(&nl_table_lock);
182 wake_up(&nl_table_wait);
183 }
184
185 static __inline__ void
186 netlink_lock_table(void)
187 {
188 /* read_lock() synchronizes us to netlink_table_grab */
189
190 read_lock(&nl_table_lock);
191 atomic_inc(&nl_table_users);
192 read_unlock(&nl_table_lock);
193 }
194
195 static __inline__ void
196 netlink_unlock_table(void)
197 {
198 if (atomic_dec_and_test(&nl_table_users))
199 wake_up(&nl_table_wait);
200 }
201
202 static __inline__ struct sock *netlink_lookup(int protocol, u32 pid)
203 {
204 struct nl_pid_hash *hash = &nl_table[protocol].hash;
205 struct hlist_head *head;
206 struct sock *sk;
207 struct hlist_node *node;
208
209 read_lock(&nl_table_lock);
210 head = nl_pid_hashfn(hash, pid);
211 sk_for_each(sk, node, head) {
212 if (nlk_sk(sk)->pid == pid) {
213 sock_hold(sk);
214 goto found;
215 }
216 }
217 sk = NULL;
218 found:
219 read_unlock(&nl_table_lock);
220 return sk;
221 }
222
223 static inline struct hlist_head *nl_pid_hash_alloc(size_t size)
224 {
225 if (size <= PAGE_SIZE)
226 return kmalloc(size, GFP_ATOMIC);
227 else
228 return (struct hlist_head *)
229 __get_free_pages(GFP_ATOMIC, get_order(size));
230 }
231
232 static inline void nl_pid_hash_free(struct hlist_head *table, size_t size)
233 {
234 if (size <= PAGE_SIZE)
235 kfree(table);
236 else
237 free_pages((unsigned long)table, get_order(size));
238 }
239
240 static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
241 {
242 unsigned int omask, mask, shift;
243 size_t osize, size;
244 struct hlist_head *otable, *table;
245 int i;
246
247 omask = mask = hash->mask;
248 osize = size = (mask + 1) * sizeof(*table);
249 shift = hash->shift;
250
251 if (grow) {
252 if (++shift > hash->max_shift)
253 return 0;
254 mask = mask * 2 + 1;
255 size *= 2;
256 }
257
258 table = nl_pid_hash_alloc(size);
259 if (!table)
260 return 0;
261
262 memset(table, 0, size);
263 otable = hash->table;
264 hash->table = table;
265 hash->mask = mask;
266 hash->shift = shift;
267 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
268
269 for (i = 0; i <= omask; i++) {
270 struct sock *sk;
271 struct hlist_node *node, *tmp;
272
273 sk_for_each_safe(sk, node, tmp, &otable[i])
274 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
275 }
276
277 nl_pid_hash_free(otable, osize);
278 hash->rehash_time = jiffies + 10 * 60 * HZ;
279 return 1;
280 }
281
282 static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
283 {
284 int avg = hash->entries >> hash->shift;
285
286 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
287 return 1;
288
289 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
290 nl_pid_hash_rehash(hash, 0);
291 return 1;
292 }
293
294 return 0;
295 }
296
297 static const struct proto_ops netlink_ops;
298
299 static int netlink_insert(struct sock *sk, u32 pid)
300 {
301 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
302 struct hlist_head *head;
303 int err = -EADDRINUSE;
304 struct sock *osk;
305 struct hlist_node *node;
306 int len;
307
308 netlink_table_grab();
309 head = nl_pid_hashfn(hash, pid);
310 len = 0;
311 sk_for_each(osk, node, head) {
312 if (nlk_sk(osk)->pid == pid)
313 break;
314 len++;
315 }
316 if (node)
317 goto err;
318
319 err = -EBUSY;
320 if (nlk_sk(sk)->pid)
321 goto err;
322
323 err = -ENOMEM;
324 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
325 goto err;
326
327 if (len && nl_pid_hash_dilute(hash, len))
328 head = nl_pid_hashfn(hash, pid);
329 hash->entries++;
330 nlk_sk(sk)->pid = pid;
331 sk_add_node(sk, head);
332 err = 0;
333
334 err:
335 netlink_table_ungrab();
336 return err;
337 }
338
339 static void netlink_remove(struct sock *sk)
340 {
341 netlink_table_grab();
342 if (sk_del_node_init(sk))
343 nl_table[sk->sk_protocol].hash.entries--;
344 if (nlk_sk(sk)->subscriptions)
345 __sk_del_bind_node(sk);
346 netlink_table_ungrab();
347 }
348
349 static struct proto netlink_proto = {
350 .name = "NETLINK",
351 .owner = THIS_MODULE,
352 .obj_size = sizeof(struct netlink_sock),
353 };
354
355 static int __netlink_create(struct socket *sock, int protocol)
356 {
357 struct sock *sk;
358 struct netlink_sock *nlk;
359
360 sock->ops = &netlink_ops;
361
362 sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1);
363 if (!sk)
364 return -ENOMEM;
365
366 sock_init_data(sock, sk);
367
368 nlk = nlk_sk(sk);
369 spin_lock_init(&nlk->cb_lock);
370 init_waitqueue_head(&nlk->wait);
371
372 sk->sk_destruct = netlink_sock_destruct;
373 sk->sk_protocol = protocol;
374 return 0;
375 }
376
377 static int netlink_create(struct socket *sock, int protocol)
378 {
379 struct module *module = NULL;
380 struct netlink_sock *nlk;
381 unsigned int groups;
382 int err = 0;
383
384 sock->state = SS_UNCONNECTED;
385
386 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
387 return -ESOCKTNOSUPPORT;
388
389 if (protocol<0 || protocol >= MAX_LINKS)
390 return -EPROTONOSUPPORT;
391
392 netlink_lock_table();
393 #ifdef CONFIG_KMOD
394 if (!nl_table[protocol].registered) {
395 netlink_unlock_table();
396 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
397 netlink_lock_table();
398 }
399 #endif
400 if (nl_table[protocol].registered &&
401 try_module_get(nl_table[protocol].module))
402 module = nl_table[protocol].module;
403 groups = nl_table[protocol].groups;
404 netlink_unlock_table();
405
406 if ((err = __netlink_create(sock, protocol)) < 0)
407 goto out_module;
408
409 nlk = nlk_sk(sock->sk);
410 nlk->module = module;
411 out:
412 return err;
413
414 out_module:
415 module_put(module);
416 goto out;
417 }
418
419 static int netlink_release(struct socket *sock)
420 {
421 struct sock *sk = sock->sk;
422 struct netlink_sock *nlk;
423
424 if (!sk)
425 return 0;
426
427 netlink_remove(sk);
428 nlk = nlk_sk(sk);
429
430 spin_lock(&nlk->cb_lock);
431 if (nlk->cb) {
432 if (nlk->cb->done)
433 nlk->cb->done(nlk->cb);
434 netlink_destroy_callback(nlk->cb);
435 nlk->cb = NULL;
436 }
437 spin_unlock(&nlk->cb_lock);
438
439 /* OK. Socket is unlinked, and, therefore,
440 no new packets will arrive */
441
442 sock_orphan(sk);
443 sock->sk = NULL;
444 wake_up_interruptible_all(&nlk->wait);
445
446 skb_queue_purge(&sk->sk_write_queue);
447
448 if (nlk->pid && !nlk->subscriptions) {
449 struct netlink_notify n = {
450 .protocol = sk->sk_protocol,
451 .pid = nlk->pid,
452 };
453 notifier_call_chain(&netlink_chain, NETLINK_URELEASE, &n);
454 }
455
456 if (nlk->module)
457 module_put(nlk->module);
458
459 if (nlk->flags & NETLINK_KERNEL_SOCKET) {
460 netlink_table_grab();
461 nl_table[sk->sk_protocol].module = NULL;
462 nl_table[sk->sk_protocol].registered = 0;
463 netlink_table_ungrab();
464 }
465
466 kfree(nlk->groups);
467 nlk->groups = NULL;
468
469 sock_put(sk);
470 return 0;
471 }
472
473 static int netlink_autobind(struct socket *sock)
474 {
475 struct sock *sk = sock->sk;
476 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
477 struct hlist_head *head;
478 struct sock *osk;
479 struct hlist_node *node;
480 s32 pid = current->tgid;
481 int err;
482 static s32 rover = -4097;
483
484 retry:
485 cond_resched();
486 netlink_table_grab();
487 head = nl_pid_hashfn(hash, pid);
488 sk_for_each(osk, node, head) {
489 if (nlk_sk(osk)->pid == pid) {
490 /* Bind collision, search negative pid values. */
491 pid = rover--;
492 if (rover > -4097)
493 rover = -4097;
494 netlink_table_ungrab();
495 goto retry;
496 }
497 }
498 netlink_table_ungrab();
499
500 err = netlink_insert(sk, pid);
501 if (err == -EADDRINUSE)
502 goto retry;
503
504 /* If 2 threads race to autobind, that is fine. */
505 if (err == -EBUSY)
506 err = 0;
507
508 return err;
509 }
510
511 static inline int netlink_capable(struct socket *sock, unsigned int flag)
512 {
513 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
514 capable(CAP_NET_ADMIN);
515 }
516
517 static void
518 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
519 {
520 struct netlink_sock *nlk = nlk_sk(sk);
521
522 if (nlk->subscriptions && !subscriptions)
523 __sk_del_bind_node(sk);
524 else if (!nlk->subscriptions && subscriptions)
525 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
526 nlk->subscriptions = subscriptions;
527 }
528
529 static int netlink_alloc_groups(struct sock *sk)
530 {
531 struct netlink_sock *nlk = nlk_sk(sk);
532 unsigned int groups;
533 int err = 0;
534
535 netlink_lock_table();
536 groups = nl_table[sk->sk_protocol].groups;
537 if (!nl_table[sk->sk_protocol].registered)
538 err = -ENOENT;
539 netlink_unlock_table();
540
541 if (err)
542 return err;
543
544 nlk->groups = kmalloc(NLGRPSZ(groups), GFP_KERNEL);
545 if (nlk->groups == NULL)
546 return -ENOMEM;
547 memset(nlk->groups, 0, NLGRPSZ(groups));
548 nlk->ngroups = groups;
549 return 0;
550 }
551
552 static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
553 {
554 struct sock *sk = sock->sk;
555 struct netlink_sock *nlk = nlk_sk(sk);
556 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
557 int err;
558
559 if (nladdr->nl_family != AF_NETLINK)
560 return -EINVAL;
561
562 /* Only superuser is allowed to listen multicasts */
563 if (nladdr->nl_groups) {
564 if (!netlink_capable(sock, NL_NONROOT_RECV))
565 return -EPERM;
566 if (nlk->groups == NULL) {
567 err = netlink_alloc_groups(sk);
568 if (err)
569 return err;
570 }
571 }
572
573 if (nlk->pid) {
574 if (nladdr->nl_pid != nlk->pid)
575 return -EINVAL;
576 } else {
577 err = nladdr->nl_pid ?
578 netlink_insert(sk, nladdr->nl_pid) :
579 netlink_autobind(sock);
580 if (err)
581 return err;
582 }
583
584 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
585 return 0;
586
587 netlink_table_grab();
588 netlink_update_subscriptions(sk, nlk->subscriptions +
589 hweight32(nladdr->nl_groups) -
590 hweight32(nlk->groups[0]));
591 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
592 netlink_table_ungrab();
593
594 return 0;
595 }
596
597 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
598 int alen, int flags)
599 {
600 int err = 0;
601 struct sock *sk = sock->sk;
602 struct netlink_sock *nlk = nlk_sk(sk);
603 struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr;
604
605 if (addr->sa_family == AF_UNSPEC) {
606 sk->sk_state = NETLINK_UNCONNECTED;
607 nlk->dst_pid = 0;
608 nlk->dst_group = 0;
609 return 0;
610 }
611 if (addr->sa_family != AF_NETLINK)
612 return -EINVAL;
613
614 /* Only superuser is allowed to send multicasts */
615 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
616 return -EPERM;
617
618 if (!nlk->pid)
619 err = netlink_autobind(sock);
620
621 if (err == 0) {
622 sk->sk_state = NETLINK_CONNECTED;
623 nlk->dst_pid = nladdr->nl_pid;
624 nlk->dst_group = ffs(nladdr->nl_groups);
625 }
626
627 return err;
628 }
629
630 static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer)
631 {
632 struct sock *sk = sock->sk;
633 struct netlink_sock *nlk = nlk_sk(sk);
634 struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr;
635
636 nladdr->nl_family = AF_NETLINK;
637 nladdr->nl_pad = 0;
638 *addr_len = sizeof(*nladdr);
639
640 if (peer) {
641 nladdr->nl_pid = nlk->dst_pid;
642 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
643 } else {
644 nladdr->nl_pid = nlk->pid;
645 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
646 }
647 return 0;
648 }
649
650 static void netlink_overrun(struct sock *sk)
651 {
652 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
653 sk->sk_err = ENOBUFS;
654 sk->sk_error_report(sk);
655 }
656 }
657
658 static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
659 {
660 int protocol = ssk->sk_protocol;
661 struct sock *sock;
662 struct netlink_sock *nlk;
663
664 sock = netlink_lookup(protocol, pid);
665 if (!sock)
666 return ERR_PTR(-ECONNREFUSED);
667
668 /* Don't bother queuing skb if kernel socket has no input function */
669 nlk = nlk_sk(sock);
670 if ((nlk->pid == 0 && !nlk->data_ready) ||
671 (sock->sk_state == NETLINK_CONNECTED &&
672 nlk->dst_pid != nlk_sk(ssk)->pid)) {
673 sock_put(sock);
674 return ERR_PTR(-ECONNREFUSED);
675 }
676 return sock;
677 }
678
679 struct sock *netlink_getsockbyfilp(struct file *filp)
680 {
681 struct inode *inode = filp->f_dentry->d_inode;
682 struct sock *sock;
683
684 if (!S_ISSOCK(inode->i_mode))
685 return ERR_PTR(-ENOTSOCK);
686
687 sock = SOCKET_I(inode)->sk;
688 if (sock->sk_family != AF_NETLINK)
689 return ERR_PTR(-EINVAL);
690
691 sock_hold(sock);
692 return sock;
693 }
694
695 /*
696 * Attach a skb to a netlink socket.
697 * The caller must hold a reference to the destination socket. On error, the
698 * reference is dropped. The skb is not send to the destination, just all
699 * all error checks are performed and memory in the queue is reserved.
700 * Return values:
701 * < 0: error. skb freed, reference to sock dropped.
702 * 0: continue
703 * 1: repeat lookup - reference dropped while waiting for socket memory.
704 */
705 int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock,
706 long timeo, struct sock *ssk)
707 {
708 struct netlink_sock *nlk;
709
710 nlk = nlk_sk(sk);
711
712 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
713 test_bit(0, &nlk->state)) {
714 DECLARE_WAITQUEUE(wait, current);
715 if (!timeo) {
716 if (!ssk || nlk_sk(ssk)->pid == 0)
717 netlink_overrun(sk);
718 sock_put(sk);
719 kfree_skb(skb);
720 return -EAGAIN;
721 }
722
723 __set_current_state(TASK_INTERRUPTIBLE);
724 add_wait_queue(&nlk->wait, &wait);
725
726 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
727 test_bit(0, &nlk->state)) &&
728 !sock_flag(sk, SOCK_DEAD))
729 timeo = schedule_timeout(timeo);
730
731 __set_current_state(TASK_RUNNING);
732 remove_wait_queue(&nlk->wait, &wait);
733 sock_put(sk);
734
735 if (signal_pending(current)) {
736 kfree_skb(skb);
737 return sock_intr_errno(timeo);
738 }
739 return 1;
740 }
741 skb_set_owner_r(skb, sk);
742 return 0;
743 }
744
745 int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol)
746 {
747 int len = skb->len;
748
749 skb_queue_tail(&sk->sk_receive_queue, skb);
750 sk->sk_data_ready(sk, len);
751 sock_put(sk);
752 return len;
753 }
754
755 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
756 {
757 kfree_skb(skb);
758 sock_put(sk);
759 }
760
761 static inline struct sk_buff *netlink_trim(struct sk_buff *skb,
762 gfp_t allocation)
763 {
764 int delta;
765
766 skb_orphan(skb);
767
768 delta = skb->end - skb->tail;
769 if (delta * 2 < skb->truesize)
770 return skb;
771
772 if (skb_shared(skb)) {
773 struct sk_buff *nskb = skb_clone(skb, allocation);
774 if (!nskb)
775 return skb;
776 kfree_skb(skb);
777 skb = nskb;
778 }
779
780 if (!pskb_expand_head(skb, 0, -delta, allocation))
781 skb->truesize -= delta;
782
783 return skb;
784 }
785
786 int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock)
787 {
788 struct sock *sk;
789 int err;
790 long timeo;
791
792 skb = netlink_trim(skb, gfp_any());
793
794 timeo = sock_sndtimeo(ssk, nonblock);
795 retry:
796 sk = netlink_getsockbypid(ssk, pid);
797 if (IS_ERR(sk)) {
798 kfree_skb(skb);
799 return PTR_ERR(sk);
800 }
801 err = netlink_attachskb(sk, skb, nonblock, timeo, ssk);
802 if (err == 1)
803 goto retry;
804 if (err)
805 return err;
806
807 return netlink_sendskb(sk, skb, ssk->sk_protocol);
808 }
809
810 static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
811 {
812 struct netlink_sock *nlk = nlk_sk(sk);
813
814 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
815 !test_bit(0, &nlk->state)) {
816 skb_set_owner_r(skb, sk);
817 skb_queue_tail(&sk->sk_receive_queue, skb);
818 sk->sk_data_ready(sk, skb->len);
819 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf;
820 }
821 return -1;
822 }
823
824 struct netlink_broadcast_data {
825 struct sock *exclude_sk;
826 u32 pid;
827 u32 group;
828 int failure;
829 int congested;
830 int delivered;
831 gfp_t allocation;
832 struct sk_buff *skb, *skb2;
833 };
834
835 static inline int do_one_broadcast(struct sock *sk,
836 struct netlink_broadcast_data *p)
837 {
838 struct netlink_sock *nlk = nlk_sk(sk);
839 int val;
840
841 if (p->exclude_sk == sk)
842 goto out;
843
844 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
845 !test_bit(p->group - 1, nlk->groups))
846 goto out;
847
848 if (p->failure) {
849 netlink_overrun(sk);
850 goto out;
851 }
852
853 sock_hold(sk);
854 if (p->skb2 == NULL) {
855 if (skb_shared(p->skb)) {
856 p->skb2 = skb_clone(p->skb, p->allocation);
857 } else {
858 p->skb2 = skb_get(p->skb);
859 /*
860 * skb ownership may have been set when
861 * delivered to a previous socket.
862 */
863 skb_orphan(p->skb2);
864 }
865 }
866 if (p->skb2 == NULL) {
867 netlink_overrun(sk);
868 /* Clone failed. Notify ALL listeners. */
869 p->failure = 1;
870 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
871 netlink_overrun(sk);
872 } else {
873 p->congested |= val;
874 p->delivered = 1;
875 p->skb2 = NULL;
876 }
877 sock_put(sk);
878
879 out:
880 return 0;
881 }
882
883 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
884 u32 group, gfp_t allocation)
885 {
886 struct netlink_broadcast_data info;
887 struct hlist_node *node;
888 struct sock *sk;
889
890 skb = netlink_trim(skb, allocation);
891
892 info.exclude_sk = ssk;
893 info.pid = pid;
894 info.group = group;
895 info.failure = 0;
896 info.congested = 0;
897 info.delivered = 0;
898 info.allocation = allocation;
899 info.skb = skb;
900 info.skb2 = NULL;
901
902 /* While we sleep in clone, do not allow to change socket list */
903
904 netlink_lock_table();
905
906 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
907 do_one_broadcast(sk, &info);
908
909 kfree_skb(skb);
910
911 netlink_unlock_table();
912
913 if (info.skb2)
914 kfree_skb(info.skb2);
915
916 if (info.delivered) {
917 if (info.congested && (allocation & __GFP_WAIT))
918 yield();
919 return 0;
920 }
921 if (info.failure)
922 return -ENOBUFS;
923 return -ESRCH;
924 }
925
926 struct netlink_set_err_data {
927 struct sock *exclude_sk;
928 u32 pid;
929 u32 group;
930 int code;
931 };
932
933 static inline int do_one_set_err(struct sock *sk,
934 struct netlink_set_err_data *p)
935 {
936 struct netlink_sock *nlk = nlk_sk(sk);
937
938 if (sk == p->exclude_sk)
939 goto out;
940
941 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
942 !test_bit(p->group - 1, nlk->groups))
943 goto out;
944
945 sk->sk_err = p->code;
946 sk->sk_error_report(sk);
947 out:
948 return 0;
949 }
950
951 void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
952 {
953 struct netlink_set_err_data info;
954 struct hlist_node *node;
955 struct sock *sk;
956
957 info.exclude_sk = ssk;
958 info.pid = pid;
959 info.group = group;
960 info.code = code;
961
962 read_lock(&nl_table_lock);
963
964 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
965 do_one_set_err(sk, &info);
966
967 read_unlock(&nl_table_lock);
968 }
969
970 static int netlink_setsockopt(struct socket *sock, int level, int optname,
971 char __user *optval, int optlen)
972 {
973 struct sock *sk = sock->sk;
974 struct netlink_sock *nlk = nlk_sk(sk);
975 int val = 0, err;
976
977 if (level != SOL_NETLINK)
978 return -ENOPROTOOPT;
979
980 if (optlen >= sizeof(int) &&
981 get_user(val, (int __user *)optval))
982 return -EFAULT;
983
984 switch (optname) {
985 case NETLINK_PKTINFO:
986 if (val)
987 nlk->flags |= NETLINK_RECV_PKTINFO;
988 else
989 nlk->flags &= ~NETLINK_RECV_PKTINFO;
990 err = 0;
991 break;
992 case NETLINK_ADD_MEMBERSHIP:
993 case NETLINK_DROP_MEMBERSHIP: {
994 unsigned int subscriptions;
995 int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0;
996
997 if (!netlink_capable(sock, NL_NONROOT_RECV))
998 return -EPERM;
999 if (nlk->groups == NULL) {
1000 err = netlink_alloc_groups(sk);
1001 if (err)
1002 return err;
1003 }
1004 if (!val || val - 1 >= nlk->ngroups)
1005 return -EINVAL;
1006 netlink_table_grab();
1007 old = test_bit(val - 1, nlk->groups);
1008 subscriptions = nlk->subscriptions - old + new;
1009 if (new)
1010 __set_bit(val - 1, nlk->groups);
1011 else
1012 __clear_bit(val - 1, nlk->groups);
1013 netlink_update_subscriptions(sk, subscriptions);
1014 netlink_table_ungrab();
1015 err = 0;
1016 break;
1017 }
1018 default:
1019 err = -ENOPROTOOPT;
1020 }
1021 return err;
1022 }
1023
1024 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1025 char __user *optval, int __user *optlen)
1026 {
1027 struct sock *sk = sock->sk;
1028 struct netlink_sock *nlk = nlk_sk(sk);
1029 int len, val, err;
1030
1031 if (level != SOL_NETLINK)
1032 return -ENOPROTOOPT;
1033
1034 if (get_user(len, optlen))
1035 return -EFAULT;
1036 if (len < 0)
1037 return -EINVAL;
1038
1039 switch (optname) {
1040 case NETLINK_PKTINFO:
1041 if (len < sizeof(int))
1042 return -EINVAL;
1043 len = sizeof(int);
1044 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
1045 put_user(len, optlen);
1046 put_user(val, optval);
1047 err = 0;
1048 break;
1049 default:
1050 err = -ENOPROTOOPT;
1051 }
1052 return err;
1053 }
1054
1055 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1056 {
1057 struct nl_pktinfo info;
1058
1059 info.group = NETLINK_CB(skb).dst_group;
1060 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1061 }
1062
1063 static inline void netlink_rcv_wake(struct sock *sk)
1064 {
1065 struct netlink_sock *nlk = nlk_sk(sk);
1066
1067 if (skb_queue_empty(&sk->sk_receive_queue))
1068 clear_bit(0, &nlk->state);
1069 if (!test_bit(0, &nlk->state))
1070 wake_up_interruptible(&nlk->wait);
1071 }
1072
1073 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1074 struct msghdr *msg, size_t len)
1075 {
1076 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1077 struct sock *sk = sock->sk;
1078 struct netlink_sock *nlk = nlk_sk(sk);
1079 struct sockaddr_nl *addr=msg->msg_name;
1080 u32 dst_pid;
1081 u32 dst_group;
1082 struct sk_buff *skb;
1083 int err;
1084 struct scm_cookie scm;
1085
1086 if (msg->msg_flags&MSG_OOB)
1087 return -EOPNOTSUPP;
1088
1089 if (NULL == siocb->scm)
1090 siocb->scm = &scm;
1091 err = scm_send(sock, msg, siocb->scm);
1092 if (err < 0)
1093 return err;
1094
1095 if (msg->msg_namelen) {
1096 if (addr->nl_family != AF_NETLINK)
1097 return -EINVAL;
1098 dst_pid = addr->nl_pid;
1099 dst_group = ffs(addr->nl_groups);
1100 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
1101 return -EPERM;
1102 } else {
1103 dst_pid = nlk->dst_pid;
1104 dst_group = nlk->dst_group;
1105 }
1106
1107 if (!nlk->pid) {
1108 err = netlink_autobind(sock);
1109 if (err)
1110 goto out;
1111 }
1112
1113 err = -EMSGSIZE;
1114 if (len > sk->sk_sndbuf - 32)
1115 goto out;
1116 err = -ENOBUFS;
1117 skb = alloc_skb(len, GFP_KERNEL);
1118 if (skb==NULL)
1119 goto out;
1120
1121 NETLINK_CB(skb).pid = nlk->pid;
1122 NETLINK_CB(skb).dst_pid = dst_pid;
1123 NETLINK_CB(skb).dst_group = dst_group;
1124 NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context);
1125 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1126
1127 /* What can I do? Netlink is asynchronous, so that
1128 we will have to save current capabilities to
1129 check them, when this message will be delivered
1130 to corresponding kernel module. --ANK (980802)
1131 */
1132
1133 err = -EFAULT;
1134 if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) {
1135 kfree_skb(skb);
1136 goto out;
1137 }
1138
1139 err = security_netlink_send(sk, skb);
1140 if (err) {
1141 kfree_skb(skb);
1142 goto out;
1143 }
1144
1145 if (dst_group) {
1146 atomic_inc(&skb->users);
1147 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
1148 }
1149 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1150
1151 out:
1152 return err;
1153 }
1154
1155 static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1156 struct msghdr *msg, size_t len,
1157 int flags)
1158 {
1159 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1160 struct scm_cookie scm;
1161 struct sock *sk = sock->sk;
1162 struct netlink_sock *nlk = nlk_sk(sk);
1163 int noblock = flags&MSG_DONTWAIT;
1164 size_t copied;
1165 struct sk_buff *skb;
1166 int err;
1167
1168 if (flags&MSG_OOB)
1169 return -EOPNOTSUPP;
1170
1171 copied = 0;
1172
1173 skb = skb_recv_datagram(sk,flags,noblock,&err);
1174 if (skb==NULL)
1175 goto out;
1176
1177 msg->msg_namelen = 0;
1178
1179 copied = skb->len;
1180 if (len < copied) {
1181 msg->msg_flags |= MSG_TRUNC;
1182 copied = len;
1183 }
1184
1185 skb->h.raw = skb->data;
1186 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1187
1188 if (msg->msg_name) {
1189 struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name;
1190 addr->nl_family = AF_NETLINK;
1191 addr->nl_pad = 0;
1192 addr->nl_pid = NETLINK_CB(skb).pid;
1193 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1194 msg->msg_namelen = sizeof(*addr);
1195 }
1196
1197 if (NULL == siocb->scm) {
1198 memset(&scm, 0, sizeof(scm));
1199 siocb->scm = &scm;
1200 }
1201 siocb->scm->creds = *NETLINK_CREDS(skb);
1202 skb_free_datagram(sk, skb);
1203
1204 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2)
1205 netlink_dump(sk);
1206
1207 scm_recv(sock, msg, siocb->scm, flags);
1208 if (nlk->flags & NETLINK_RECV_PKTINFO)
1209 netlink_cmsg_recv_pktinfo(msg, skb);
1210
1211 out:
1212 netlink_rcv_wake(sk);
1213 return err ? : copied;
1214 }
1215
1216 static void netlink_data_ready(struct sock *sk, int len)
1217 {
1218 struct netlink_sock *nlk = nlk_sk(sk);
1219
1220 if (nlk->data_ready)
1221 nlk->data_ready(sk, len);
1222 netlink_rcv_wake(sk);
1223 }
1224
1225 /*
1226 * We export these functions to other modules. They provide a
1227 * complete set of kernel non-blocking support for message
1228 * queueing.
1229 */
1230
1231 struct sock *
1232 netlink_kernel_create(int unit, unsigned int groups,
1233 void (*input)(struct sock *sk, int len),
1234 struct module *module)
1235 {
1236 struct socket *sock;
1237 struct sock *sk;
1238 struct netlink_sock *nlk;
1239
1240 if (!nl_table)
1241 return NULL;
1242
1243 if (unit<0 || unit>=MAX_LINKS)
1244 return NULL;
1245
1246 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1247 return NULL;
1248
1249 if (__netlink_create(sock, unit) < 0)
1250 goto out_sock_release;
1251
1252 sk = sock->sk;
1253 sk->sk_data_ready = netlink_data_ready;
1254 if (input)
1255 nlk_sk(sk)->data_ready = input;
1256
1257 if (netlink_insert(sk, 0))
1258 goto out_sock_release;
1259
1260 nlk = nlk_sk(sk);
1261 nlk->flags |= NETLINK_KERNEL_SOCKET;
1262
1263 netlink_table_grab();
1264 nl_table[unit].groups = groups < 32 ? 32 : groups;
1265 nl_table[unit].module = module;
1266 nl_table[unit].registered = 1;
1267 netlink_table_ungrab();
1268
1269 return sk;
1270
1271 out_sock_release:
1272 sock_release(sock);
1273 return NULL;
1274 }
1275
1276 void netlink_set_nonroot(int protocol, unsigned int flags)
1277 {
1278 if ((unsigned int)protocol < MAX_LINKS)
1279 nl_table[protocol].nl_nonroot = flags;
1280 }
1281
1282 static void netlink_destroy_callback(struct netlink_callback *cb)
1283 {
1284 if (cb->skb)
1285 kfree_skb(cb->skb);
1286 kfree(cb);
1287 }
1288
1289 /*
1290 * It looks a bit ugly.
1291 * It would be better to create kernel thread.
1292 */
1293
1294 static int netlink_dump(struct sock *sk)
1295 {
1296 struct netlink_sock *nlk = nlk_sk(sk);
1297 struct netlink_callback *cb;
1298 struct sk_buff *skb;
1299 struct nlmsghdr *nlh;
1300 int len;
1301
1302 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL);
1303 if (!skb)
1304 return -ENOBUFS;
1305
1306 spin_lock(&nlk->cb_lock);
1307
1308 cb = nlk->cb;
1309 if (cb == NULL) {
1310 spin_unlock(&nlk->cb_lock);
1311 kfree_skb(skb);
1312 return -EINVAL;
1313 }
1314
1315 len = cb->dump(skb, cb);
1316
1317 if (len > 0) {
1318 spin_unlock(&nlk->cb_lock);
1319 skb_queue_tail(&sk->sk_receive_queue, skb);
1320 sk->sk_data_ready(sk, len);
1321 return 0;
1322 }
1323
1324 nlh = NLMSG_NEW_ANSWER(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1325 memcpy(NLMSG_DATA(nlh), &len, sizeof(len));
1326 skb_queue_tail(&sk->sk_receive_queue, skb);
1327 sk->sk_data_ready(sk, skb->len);
1328
1329 if (cb->done)
1330 cb->done(cb);
1331 nlk->cb = NULL;
1332 spin_unlock(&nlk->cb_lock);
1333
1334 netlink_destroy_callback(cb);
1335 return 0;
1336
1337 nlmsg_failure:
1338 return -ENOBUFS;
1339 }
1340
1341 int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1342 struct nlmsghdr *nlh,
1343 int (*dump)(struct sk_buff *skb, struct netlink_callback*),
1344 int (*done)(struct netlink_callback*))
1345 {
1346 struct netlink_callback *cb;
1347 struct sock *sk;
1348 struct netlink_sock *nlk;
1349
1350 cb = kmalloc(sizeof(*cb), GFP_KERNEL);
1351 if (cb == NULL)
1352 return -ENOBUFS;
1353
1354 memset(cb, 0, sizeof(*cb));
1355 cb->dump = dump;
1356 cb->done = done;
1357 cb->nlh = nlh;
1358 atomic_inc(&skb->users);
1359 cb->skb = skb;
1360
1361 sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid);
1362 if (sk == NULL) {
1363 netlink_destroy_callback(cb);
1364 return -ECONNREFUSED;
1365 }
1366 nlk = nlk_sk(sk);
1367 /* A dump is in progress... */
1368 spin_lock(&nlk->cb_lock);
1369 if (nlk->cb) {
1370 spin_unlock(&nlk->cb_lock);
1371 netlink_destroy_callback(cb);
1372 sock_put(sk);
1373 return -EBUSY;
1374 }
1375 nlk->cb = cb;
1376 spin_unlock(&nlk->cb_lock);
1377
1378 netlink_dump(sk);
1379 sock_put(sk);
1380 return 0;
1381 }
1382
1383 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1384 {
1385 struct sk_buff *skb;
1386 struct nlmsghdr *rep;
1387 struct nlmsgerr *errmsg;
1388 int size;
1389
1390 if (err == 0)
1391 size = NLMSG_SPACE(sizeof(struct nlmsgerr));
1392 else
1393 size = NLMSG_SPACE(4 + NLMSG_ALIGN(nlh->nlmsg_len));
1394
1395 skb = alloc_skb(size, GFP_KERNEL);
1396 if (!skb) {
1397 struct sock *sk;
1398
1399 sk = netlink_lookup(in_skb->sk->sk_protocol,
1400 NETLINK_CB(in_skb).pid);
1401 if (sk) {
1402 sk->sk_err = ENOBUFS;
1403 sk->sk_error_report(sk);
1404 sock_put(sk);
1405 }
1406 return;
1407 }
1408
1409 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
1410 NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
1411 errmsg = NLMSG_DATA(rep);
1412 errmsg->error = err;
1413 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(struct nlmsghdr));
1414 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1415 }
1416
1417 static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
1418 struct nlmsghdr *, int *))
1419 {
1420 unsigned int total_len;
1421 struct nlmsghdr *nlh;
1422 int err;
1423
1424 while (skb->len >= nlmsg_total_size(0)) {
1425 nlh = (struct nlmsghdr *) skb->data;
1426
1427 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
1428 return 0;
1429
1430 total_len = min(NLMSG_ALIGN(nlh->nlmsg_len), skb->len);
1431
1432 if (cb(skb, nlh, &err) < 0) {
1433 /* Not an error, but we have to interrupt processing
1434 * here. Note: that in this case we do not pull
1435 * message from skb, it will be processed later.
1436 */
1437 if (err == 0)
1438 return -1;
1439 netlink_ack(skb, nlh, err);
1440 } else if (nlh->nlmsg_flags & NLM_F_ACK)
1441 netlink_ack(skb, nlh, 0);
1442
1443 skb_pull(skb, total_len);
1444 }
1445
1446 return 0;
1447 }
1448
1449 /**
1450 * nelink_run_queue - Process netlink receive queue.
1451 * @sk: Netlink socket containing the queue
1452 * @qlen: Place to store queue length upon entry
1453 * @cb: Callback function invoked for each netlink message found
1454 *
1455 * Processes as much as there was in the queue upon entry and invokes
1456 * a callback function for each netlink message found. The callback
1457 * function may refuse a message by returning a negative error code
1458 * but setting the error pointer to 0 in which case this function
1459 * returns with a qlen != 0.
1460 *
1461 * qlen must be initialized to 0 before the initial entry, afterwards
1462 * the function may be called repeatedly until qlen reaches 0.
1463 */
1464 void netlink_run_queue(struct sock *sk, unsigned int *qlen,
1465 int (*cb)(struct sk_buff *, struct nlmsghdr *, int *))
1466 {
1467 struct sk_buff *skb;
1468
1469 if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue))
1470 *qlen = skb_queue_len(&sk->sk_receive_queue);
1471
1472 for (; *qlen; (*qlen)--) {
1473 skb = skb_dequeue(&sk->sk_receive_queue);
1474 if (netlink_rcv_skb(skb, cb)) {
1475 if (skb->len)
1476 skb_queue_head(&sk->sk_receive_queue, skb);
1477 else {
1478 kfree_skb(skb);
1479 (*qlen)--;
1480 }
1481 break;
1482 }
1483
1484 kfree_skb(skb);
1485 }
1486 }
1487
1488 /**
1489 * netlink_queue_skip - Skip netlink message while processing queue.
1490 * @nlh: Netlink message to be skipped
1491 * @skb: Socket buffer containing the netlink messages.
1492 *
1493 * Pulls the given netlink message off the socket buffer so the next
1494 * call to netlink_queue_run() will not reconsider the message.
1495 */
1496 void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb)
1497 {
1498 int msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1499
1500 if (msglen > skb->len)
1501 msglen = skb->len;
1502
1503 skb_pull(skb, msglen);
1504 }
1505
1506 #ifdef CONFIG_PROC_FS
1507 struct nl_seq_iter {
1508 int link;
1509 int hash_idx;
1510 };
1511
1512 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1513 {
1514 struct nl_seq_iter *iter = seq->private;
1515 int i, j;
1516 struct sock *s;
1517 struct hlist_node *node;
1518 loff_t off = 0;
1519
1520 for (i=0; i<MAX_LINKS; i++) {
1521 struct nl_pid_hash *hash = &nl_table[i].hash;
1522
1523 for (j = 0; j <= hash->mask; j++) {
1524 sk_for_each(s, node, &hash->table[j]) {
1525 if (off == pos) {
1526 iter->link = i;
1527 iter->hash_idx = j;
1528 return s;
1529 }
1530 ++off;
1531 }
1532 }
1533 }
1534 return NULL;
1535 }
1536
1537 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
1538 {
1539 read_lock(&nl_table_lock);
1540 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1541 }
1542
1543 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1544 {
1545 struct sock *s;
1546 struct nl_seq_iter *iter;
1547 int i, j;
1548
1549 ++*pos;
1550
1551 if (v == SEQ_START_TOKEN)
1552 return netlink_seq_socket_idx(seq, 0);
1553
1554 s = sk_next(v);
1555 if (s)
1556 return s;
1557
1558 iter = seq->private;
1559 i = iter->link;
1560 j = iter->hash_idx + 1;
1561
1562 do {
1563 struct nl_pid_hash *hash = &nl_table[i].hash;
1564
1565 for (; j <= hash->mask; j++) {
1566 s = sk_head(&hash->table[j]);
1567 if (s) {
1568 iter->link = i;
1569 iter->hash_idx = j;
1570 return s;
1571 }
1572 }
1573
1574 j = 0;
1575 } while (++i < MAX_LINKS);
1576
1577 return NULL;
1578 }
1579
1580 static void netlink_seq_stop(struct seq_file *seq, void *v)
1581 {
1582 read_unlock(&nl_table_lock);
1583 }
1584
1585
1586 static int netlink_seq_show(struct seq_file *seq, void *v)
1587 {
1588 if (v == SEQ_START_TOKEN)
1589 seq_puts(seq,
1590 "sk Eth Pid Groups "
1591 "Rmem Wmem Dump Locks\n");
1592 else {
1593 struct sock *s = v;
1594 struct netlink_sock *nlk = nlk_sk(s);
1595
1596 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1597 s,
1598 s->sk_protocol,
1599 nlk->pid,
1600 nlk->groups ? (u32)nlk->groups[0] : 0,
1601 atomic_read(&s->sk_rmem_alloc),
1602 atomic_read(&s->sk_wmem_alloc),
1603 nlk->cb,
1604 atomic_read(&s->sk_refcnt)
1605 );
1606
1607 }
1608 return 0;
1609 }
1610
1611 static struct seq_operations netlink_seq_ops = {
1612 .start = netlink_seq_start,
1613 .next = netlink_seq_next,
1614 .stop = netlink_seq_stop,
1615 .show = netlink_seq_show,
1616 };
1617
1618
1619 static int netlink_seq_open(struct inode *inode, struct file *file)
1620 {
1621 struct seq_file *seq;
1622 struct nl_seq_iter *iter;
1623 int err;
1624
1625 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
1626 if (!iter)
1627 return -ENOMEM;
1628
1629 err = seq_open(file, &netlink_seq_ops);
1630 if (err) {
1631 kfree(iter);
1632 return err;
1633 }
1634
1635 memset(iter, 0, sizeof(*iter));
1636 seq = file->private_data;
1637 seq->private = iter;
1638 return 0;
1639 }
1640
1641 static struct file_operations netlink_seq_fops = {
1642 .owner = THIS_MODULE,
1643 .open = netlink_seq_open,
1644 .read = seq_read,
1645 .llseek = seq_lseek,
1646 .release = seq_release_private,
1647 };
1648
1649 #endif
1650
1651 int netlink_register_notifier(struct notifier_block *nb)
1652 {
1653 return notifier_chain_register(&netlink_chain, nb);
1654 }
1655
1656 int netlink_unregister_notifier(struct notifier_block *nb)
1657 {
1658 return notifier_chain_unregister(&netlink_chain, nb);
1659 }
1660
1661 static const struct proto_ops netlink_ops = {
1662 .family = PF_NETLINK,
1663 .owner = THIS_MODULE,
1664 .release = netlink_release,
1665 .bind = netlink_bind,
1666 .connect = netlink_connect,
1667 .socketpair = sock_no_socketpair,
1668 .accept = sock_no_accept,
1669 .getname = netlink_getname,
1670 .poll = datagram_poll,
1671 .ioctl = sock_no_ioctl,
1672 .listen = sock_no_listen,
1673 .shutdown = sock_no_shutdown,
1674 .setsockopt = netlink_setsockopt,
1675 .getsockopt = netlink_getsockopt,
1676 .sendmsg = netlink_sendmsg,
1677 .recvmsg = netlink_recvmsg,
1678 .mmap = sock_no_mmap,
1679 .sendpage = sock_no_sendpage,
1680 };
1681
1682 static struct net_proto_family netlink_family_ops = {
1683 .family = PF_NETLINK,
1684 .create = netlink_create,
1685 .owner = THIS_MODULE, /* for consistency 8) */
1686 };
1687
1688 extern void netlink_skb_parms_too_large(void);
1689
1690 static int __init netlink_proto_init(void)
1691 {
1692 struct sk_buff *dummy_skb;
1693 int i;
1694 unsigned long max;
1695 unsigned int order;
1696 int err = proto_register(&netlink_proto, 0);
1697
1698 if (err != 0)
1699 goto out;
1700
1701 if (sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb))
1702 netlink_skb_parms_too_large();
1703
1704 nl_table = kmalloc(sizeof(*nl_table) * MAX_LINKS, GFP_KERNEL);
1705 if (!nl_table) {
1706 enomem:
1707 printk(KERN_CRIT "netlink_init: Cannot allocate nl_table\n");
1708 return -ENOMEM;
1709 }
1710
1711 memset(nl_table, 0, sizeof(*nl_table) * MAX_LINKS);
1712
1713 if (num_physpages >= (128 * 1024))
1714 max = num_physpages >> (21 - PAGE_SHIFT);
1715 else
1716 max = num_physpages >> (23 - PAGE_SHIFT);
1717
1718 order = get_bitmask_order(max) - 1 + PAGE_SHIFT;
1719 max = (1UL << order) / sizeof(struct hlist_head);
1720 order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1;
1721
1722 for (i = 0; i < MAX_LINKS; i++) {
1723 struct nl_pid_hash *hash = &nl_table[i].hash;
1724
1725 hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table));
1726 if (!hash->table) {
1727 while (i-- > 0)
1728 nl_pid_hash_free(nl_table[i].hash.table,
1729 1 * sizeof(*hash->table));
1730 kfree(nl_table);
1731 goto enomem;
1732 }
1733 memset(hash->table, 0, 1 * sizeof(*hash->table));
1734 hash->max_shift = order;
1735 hash->shift = 0;
1736 hash->mask = 0;
1737 hash->rehash_time = jiffies;
1738 }
1739
1740 sock_register(&netlink_family_ops);
1741 #ifdef CONFIG_PROC_FS
1742 proc_net_fops_create("netlink", 0, &netlink_seq_fops);
1743 #endif
1744 /* The netlink device handler may be needed early. */
1745 rtnetlink_init();
1746 out:
1747 return err;
1748 }
1749
1750 core_initcall(netlink_proto_init);
1751
1752 EXPORT_SYMBOL(netlink_ack);
1753 EXPORT_SYMBOL(netlink_run_queue);
1754 EXPORT_SYMBOL(netlink_queue_skip);
1755 EXPORT_SYMBOL(netlink_broadcast);
1756 EXPORT_SYMBOL(netlink_dump_start);
1757 EXPORT_SYMBOL(netlink_kernel_create);
1758 EXPORT_SYMBOL(netlink_register_notifier);
1759 EXPORT_SYMBOL(netlink_set_err);
1760 EXPORT_SYMBOL(netlink_set_nonroot);
1761 EXPORT_SYMBOL(netlink_unicast);
1762 EXPORT_SYMBOL(netlink_unregister_notifier);
1763
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree