?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[PATCH] Ext2 flags shouldn't report "nogrpid"
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / netlink / af_netlink.c
blobd00a9034cb5f9177aa31ad4d533ce9490cfa77c6
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 long *listeners;
110 unsigned int nl_nonroot;
111 unsigned int groups;
112 struct module *module;
113 int registered;
114 };
115
116 static struct netlink_table *nl_table;
117
118 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
119
120 static int netlink_dump(struct sock *sk);
121 static void netlink_destroy_callback(struct netlink_callback *cb);
122
123 static DEFINE_RWLOCK(nl_table_lock);
124 static atomic_t nl_table_users = ATOMIC_INIT(0);
125
126 static struct notifier_block *netlink_chain;
127
128 static u32 netlink_group_mask(u32 group)
129 {
130 return group ? 1 << (group - 1) : 0;
131 }
132
133 static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
134 {
135 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
136 }
137
138 static void netlink_sock_destruct(struct sock *sk)
139 {
140 skb_queue_purge(&sk->sk_receive_queue);
141
142 if (!sock_flag(sk, SOCK_DEAD)) {
143 printk("Freeing alive netlink socket %p\n", sk);
144 return;
145 }
146 BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
147 BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
148 BUG_TRAP(!nlk_sk(sk)->cb);
149 BUG_TRAP(!nlk_sk(sk)->groups);
150 }
151
152 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
153 * Look, when several writers sleep and reader wakes them up, all but one
154 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
155 * this, _but_ remember, it adds useless work on UP machines.
156 */
157
158 static void netlink_table_grab(void)
159 {
160 write_lock_bh(&nl_table_lock);
161
162 if (atomic_read(&nl_table_users)) {
163 DECLARE_WAITQUEUE(wait, current);
164
165 add_wait_queue_exclusive(&nl_table_wait, &wait);
166 for(;;) {
167 set_current_state(TASK_UNINTERRUPTIBLE);
168 if (atomic_read(&nl_table_users) == 0)
169 break;
170 write_unlock_bh(&nl_table_lock);
171 schedule();
172 write_lock_bh(&nl_table_lock);
173 }
174
175 __set_current_state(TASK_RUNNING);
176 remove_wait_queue(&nl_table_wait, &wait);
177 }
178 }
179
180 static __inline__ void netlink_table_ungrab(void)
181 {
182 write_unlock_bh(&nl_table_lock);
183 wake_up(&nl_table_wait);
184 }
185
186 static __inline__ void
187 netlink_lock_table(void)
188 {
189 /* read_lock() synchronizes us to netlink_table_grab */
190
191 read_lock(&nl_table_lock);
192 atomic_inc(&nl_table_users);
193 read_unlock(&nl_table_lock);
194 }
195
196 static __inline__ void
197 netlink_unlock_table(void)
198 {
199 if (atomic_dec_and_test(&nl_table_users))
200 wake_up(&nl_table_wait);
201 }
202
203 static __inline__ struct sock *netlink_lookup(int protocol, u32 pid)
204 {
205 struct nl_pid_hash *hash = &nl_table[protocol].hash;
206 struct hlist_head *head;
207 struct sock *sk;
208 struct hlist_node *node;
209
210 read_lock(&nl_table_lock);
211 head = nl_pid_hashfn(hash, pid);
212 sk_for_each(sk, node, head) {
213 if (nlk_sk(sk)->pid == pid) {
214 sock_hold(sk);
215 goto found;
216 }
217 }
218 sk = NULL;
219 found:
220 read_unlock(&nl_table_lock);
221 return sk;
222 }
223
224 static inline struct hlist_head *nl_pid_hash_alloc(size_t size)
225 {
226 if (size <= PAGE_SIZE)
227 return kmalloc(size, GFP_ATOMIC);
228 else
229 return (struct hlist_head *)
230 __get_free_pages(GFP_ATOMIC, get_order(size));
231 }
232
233 static inline void nl_pid_hash_free(struct hlist_head *table, size_t size)
234 {
235 if (size <= PAGE_SIZE)
236 kfree(table);
237 else
238 free_pages((unsigned long)table, get_order(size));
239 }
240
241 static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
242 {
243 unsigned int omask, mask, shift;
244 size_t osize, size;
245 struct hlist_head *otable, *table;
246 int i;
247
248 omask = mask = hash->mask;
249 osize = size = (mask + 1) * sizeof(*table);
250 shift = hash->shift;
251
252 if (grow) {
253 if (++shift > hash->max_shift)
254 return 0;
255 mask = mask * 2 + 1;
256 size *= 2;
257 }
258
259 table = nl_pid_hash_alloc(size);
260 if (!table)
261 return 0;
262
263 memset(table, 0, size);
264 otable = hash->table;
265 hash->table = table;
266 hash->mask = mask;
267 hash->shift = shift;
268 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
269
270 for (i = 0; i <= omask; i++) {
271 struct sock *sk;
272 struct hlist_node *node, *tmp;
273
274 sk_for_each_safe(sk, node, tmp, &otable[i])
275 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
276 }
277
278 nl_pid_hash_free(otable, osize);
279 hash->rehash_time = jiffies + 10 * 60 * HZ;
280 return 1;
281 }
282
283 static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
284 {
285 int avg = hash->entries >> hash->shift;
286
287 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
288 return 1;
289
290 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
291 nl_pid_hash_rehash(hash, 0);
292 return 1;
293 }
294
295 return 0;
296 }
297
298 static const struct proto_ops netlink_ops;
299
300 static void
301 netlink_update_listeners(struct sock *sk)
302 {
303 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
304 struct hlist_node *node;
305 unsigned long mask;
306 unsigned int i;
307
308 for (i = 0; i < NLGRPSZ(tbl->groups)/sizeof(unsigned long); i++) {
309 mask = 0;
310 sk_for_each_bound(sk, node, &tbl->mc_list)
311 mask |= nlk_sk(sk)->groups[i];
312 tbl->listeners[i] = mask;
313 }
314 /* this function is only called with the netlink table "grabbed", which
315 * makes sure updates are visible before bind or setsockopt return. */
316 }
317
318 static int netlink_insert(struct sock *sk, u32 pid)
319 {
320 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
321 struct hlist_head *head;
322 int err = -EADDRINUSE;
323 struct sock *osk;
324 struct hlist_node *node;
325 int len;
326
327 netlink_table_grab();
328 head = nl_pid_hashfn(hash, pid);
329 len = 0;
330 sk_for_each(osk, node, head) {
331 if (nlk_sk(osk)->pid == pid)
332 break;
333 len++;
334 }
335 if (node)
336 goto err;
337
338 err = -EBUSY;
339 if (nlk_sk(sk)->pid)
340 goto err;
341
342 err = -ENOMEM;
343 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
344 goto err;
345
346 if (len && nl_pid_hash_dilute(hash, len))
347 head = nl_pid_hashfn(hash, pid);
348 hash->entries++;
349 nlk_sk(sk)->pid = pid;
350 sk_add_node(sk, head);
351 err = 0;
352
353 err:
354 netlink_table_ungrab();
355 return err;
356 }
357
358 static void netlink_remove(struct sock *sk)
359 {
360 netlink_table_grab();
361 if (sk_del_node_init(sk))
362 nl_table[sk->sk_protocol].hash.entries--;
363 if (nlk_sk(sk)->subscriptions)
364 __sk_del_bind_node(sk);
365 netlink_table_ungrab();
366 }
367
368 static struct proto netlink_proto = {
369 .name = "NETLINK",
370 .owner = THIS_MODULE,
371 .obj_size = sizeof(struct netlink_sock),
372 };
373
374 static int __netlink_create(struct socket *sock, int protocol)
375 {
376 struct sock *sk;
377 struct netlink_sock *nlk;
378
379 sock->ops = &netlink_ops;
380
381 sk = sk_alloc(PF_NETLINK, GFP_KERNEL, &netlink_proto, 1);
382 if (!sk)
383 return -ENOMEM;
384
385 sock_init_data(sock, sk);
386
387 nlk = nlk_sk(sk);
388 spin_lock_init(&nlk->cb_lock);
389 init_waitqueue_head(&nlk->wait);
390
391 sk->sk_destruct = netlink_sock_destruct;
392 sk->sk_protocol = protocol;
393 return 0;
394 }
395
396 static int netlink_create(struct socket *sock, int protocol)
397 {
398 struct module *module = NULL;
399 struct netlink_sock *nlk;
400 unsigned int groups;
401 int err = 0;
402
403 sock->state = SS_UNCONNECTED;
404
405 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
406 return -ESOCKTNOSUPPORT;
407
408 if (protocol<0 || protocol >= MAX_LINKS)
409 return -EPROTONOSUPPORT;
410
411 netlink_lock_table();
412 #ifdef CONFIG_KMOD
413 if (!nl_table[protocol].registered) {
414 netlink_unlock_table();
415 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
416 netlink_lock_table();
417 }
418 #endif
419 if (nl_table[protocol].registered &&
420 try_module_get(nl_table[protocol].module))
421 module = nl_table[protocol].module;
422 groups = nl_table[protocol].groups;
423 netlink_unlock_table();
424
425 if ((err = __netlink_create(sock, protocol)) < 0)
426 goto out_module;
427
428 nlk = nlk_sk(sock->sk);
429 nlk->module = module;
430 out:
431 return err;
432
433 out_module:
434 module_put(module);
435 goto out;
436 }
437
438 static int netlink_release(struct socket *sock)
439 {
440 struct sock *sk = sock->sk;
441 struct netlink_sock *nlk;
442
443 if (!sk)
444 return 0;
445
446 netlink_remove(sk);
447 nlk = nlk_sk(sk);
448
449 spin_lock(&nlk->cb_lock);
450 if (nlk->cb) {
451 if (nlk->cb->done)
452 nlk->cb->done(nlk->cb);
453 netlink_destroy_callback(nlk->cb);
454 nlk->cb = NULL;
455 }
456 spin_unlock(&nlk->cb_lock);
457
458 /* OK. Socket is unlinked, and, therefore,
459 no new packets will arrive */
460
461 sock_orphan(sk);
462 sock->sk = NULL;
463 wake_up_interruptible_all(&nlk->wait);
464
465 skb_queue_purge(&sk->sk_write_queue);
466
467 if (nlk->pid && !nlk->subscriptions) {
468 struct netlink_notify n = {
469 .protocol = sk->sk_protocol,
470 .pid = nlk->pid,
471 };
472 notifier_call_chain(&netlink_chain, NETLINK_URELEASE, &n);
473 }
474
475 if (nlk->module)
476 module_put(nlk->module);
477
478 netlink_table_grab();
479 if (nlk->flags & NETLINK_KERNEL_SOCKET) {
480 kfree(nl_table[sk->sk_protocol].listeners);
481 nl_table[sk->sk_protocol].module = NULL;
482 nl_table[sk->sk_protocol].registered = 0;
483 } else if (nlk->subscriptions)
484 netlink_update_listeners(sk);
485 netlink_table_ungrab();
486
487 kfree(nlk->groups);
488 nlk->groups = NULL;
489
490 sock_put(sk);
491 return 0;
492 }
493
494 static int netlink_autobind(struct socket *sock)
495 {
496 struct sock *sk = sock->sk;
497 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
498 struct hlist_head *head;
499 struct sock *osk;
500 struct hlist_node *node;
501 s32 pid = current->tgid;
502 int err;
503 static s32 rover = -4097;
504
505 retry:
506 cond_resched();
507 netlink_table_grab();
508 head = nl_pid_hashfn(hash, pid);
509 sk_for_each(osk, node, head) {
510 if (nlk_sk(osk)->pid == pid) {
511 /* Bind collision, search negative pid values. */
512 pid = rover--;
513 if (rover > -4097)
514 rover = -4097;
515 netlink_table_ungrab();
516 goto retry;
517 }
518 }
519 netlink_table_ungrab();
520
521 err = netlink_insert(sk, pid);
522 if (err == -EADDRINUSE)
523 goto retry;
524
525 /* If 2 threads race to autobind, that is fine. */
526 if (err == -EBUSY)
527 err = 0;
528
529 return err;
530 }
531
532 static inline int netlink_capable(struct socket *sock, unsigned int flag)
533 {
534 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
535 capable(CAP_NET_ADMIN);
536 }
537
538 static void
539 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
540 {
541 struct netlink_sock *nlk = nlk_sk(sk);
542
543 if (nlk->subscriptions && !subscriptions)
544 __sk_del_bind_node(sk);
545 else if (!nlk->subscriptions && subscriptions)
546 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
547 nlk->subscriptions = subscriptions;
548 }
549
550 static int netlink_alloc_groups(struct sock *sk)
551 {
552 struct netlink_sock *nlk = nlk_sk(sk);
553 unsigned int groups;
554 int err = 0;
555
556 netlink_lock_table();
557 groups = nl_table[sk->sk_protocol].groups;
558 if (!nl_table[sk->sk_protocol].registered)
559 err = -ENOENT;
560 netlink_unlock_table();
561
562 if (err)
563 return err;
564
565 nlk->groups = kmalloc(NLGRPSZ(groups), GFP_KERNEL);
566 if (nlk->groups == NULL)
567 return -ENOMEM;
568 memset(nlk->groups, 0, NLGRPSZ(groups));
569 nlk->ngroups = groups;
570 return 0;
571 }
572
573 static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
574 {
575 struct sock *sk = sock->sk;
576 struct netlink_sock *nlk = nlk_sk(sk);
577 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
578 int err;
579
580 if (nladdr->nl_family != AF_NETLINK)
581 return -EINVAL;
582
583 /* Only superuser is allowed to listen multicasts */
584 if (nladdr->nl_groups) {
585 if (!netlink_capable(sock, NL_NONROOT_RECV))
586 return -EPERM;
587 if (nlk->groups == NULL) {
588 err = netlink_alloc_groups(sk);
589 if (err)
590 return err;
591 }
592 }
593
594 if (nlk->pid) {
595 if (nladdr->nl_pid != nlk->pid)
596 return -EINVAL;
597 } else {
598 err = nladdr->nl_pid ?
599 netlink_insert(sk, nladdr->nl_pid) :
600 netlink_autobind(sock);
601 if (err)
602 return err;
603 }
604
605 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
606 return 0;
607
608 netlink_table_grab();
609 netlink_update_subscriptions(sk, nlk->subscriptions +
610 hweight32(nladdr->nl_groups) -
611 hweight32(nlk->groups[0]));
612 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
613 netlink_update_listeners(sk);
614 netlink_table_ungrab();
615
616 return 0;
617 }
618
619 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
620 int alen, int flags)
621 {
622 int err = 0;
623 struct sock *sk = sock->sk;
624 struct netlink_sock *nlk = nlk_sk(sk);
625 struct sockaddr_nl *nladdr=(struct sockaddr_nl*)addr;
626
627 if (addr->sa_family == AF_UNSPEC) {
628 sk->sk_state = NETLINK_UNCONNECTED;
629 nlk->dst_pid = 0;
630 nlk->dst_group = 0;
631 return 0;
632 }
633 if (addr->sa_family != AF_NETLINK)
634 return -EINVAL;
635
636 /* Only superuser is allowed to send multicasts */
637 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
638 return -EPERM;
639
640 if (!nlk->pid)
641 err = netlink_autobind(sock);
642
643 if (err == 0) {
644 sk->sk_state = NETLINK_CONNECTED;
645 nlk->dst_pid = nladdr->nl_pid;
646 nlk->dst_group = ffs(nladdr->nl_groups);
647 }
648
649 return err;
650 }
651
652 static int netlink_getname(struct socket *sock, struct sockaddr *addr, int *addr_len, int peer)
653 {
654 struct sock *sk = sock->sk;
655 struct netlink_sock *nlk = nlk_sk(sk);
656 struct sockaddr_nl *nladdr=(struct sockaddr_nl *)addr;
657
658 nladdr->nl_family = AF_NETLINK;
659 nladdr->nl_pad = 0;
660 *addr_len = sizeof(*nladdr);
661
662 if (peer) {
663 nladdr->nl_pid = nlk->dst_pid;
664 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
665 } else {
666 nladdr->nl_pid = nlk->pid;
667 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
668 }
669 return 0;
670 }
671
672 static void netlink_overrun(struct sock *sk)
673 {
674 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
675 sk->sk_err = ENOBUFS;
676 sk->sk_error_report(sk);
677 }
678 }
679
680 static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
681 {
682 int protocol = ssk->sk_protocol;
683 struct sock *sock;
684 struct netlink_sock *nlk;
685
686 sock = netlink_lookup(protocol, pid);
687 if (!sock)
688 return ERR_PTR(-ECONNREFUSED);
689
690 /* Don't bother queuing skb if kernel socket has no input function */
691 nlk = nlk_sk(sock);
692 if ((nlk->pid == 0 && !nlk->data_ready) ||
693 (sock->sk_state == NETLINK_CONNECTED &&
694 nlk->dst_pid != nlk_sk(ssk)->pid)) {
695 sock_put(sock);
696 return ERR_PTR(-ECONNREFUSED);
697 }
698 return sock;
699 }
700
701 struct sock *netlink_getsockbyfilp(struct file *filp)
702 {
703 struct inode *inode = filp->f_dentry->d_inode;
704 struct sock *sock;
705
706 if (!S_ISSOCK(inode->i_mode))
707 return ERR_PTR(-ENOTSOCK);
708
709 sock = SOCKET_I(inode)->sk;
710 if (sock->sk_family != AF_NETLINK)
711 return ERR_PTR(-EINVAL);
712
713 sock_hold(sock);
714 return sock;
715 }
716
717 /*
718 * Attach a skb to a netlink socket.
719 * The caller must hold a reference to the destination socket. On error, the
720 * reference is dropped. The skb is not send to the destination, just all
721 * all error checks are performed and memory in the queue is reserved.
722 * Return values:
723 * < 0: error. skb freed, reference to sock dropped.
724 * 0: continue
725 * 1: repeat lookup - reference dropped while waiting for socket memory.
726 */
727 int netlink_attachskb(struct sock *sk, struct sk_buff *skb, int nonblock,
728 long timeo, struct sock *ssk)
729 {
730 struct netlink_sock *nlk;
731
732 nlk = nlk_sk(sk);
733
734 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
735 test_bit(0, &nlk->state)) {
736 DECLARE_WAITQUEUE(wait, current);
737 if (!timeo) {
738 if (!ssk || nlk_sk(ssk)->pid == 0)
739 netlink_overrun(sk);
740 sock_put(sk);
741 kfree_skb(skb);
742 return -EAGAIN;
743 }
744
745 __set_current_state(TASK_INTERRUPTIBLE);
746 add_wait_queue(&nlk->wait, &wait);
747
748 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
749 test_bit(0, &nlk->state)) &&
750 !sock_flag(sk, SOCK_DEAD))
751 timeo = schedule_timeout(timeo);
752
753 __set_current_state(TASK_RUNNING);
754 remove_wait_queue(&nlk->wait, &wait);
755 sock_put(sk);
756
757 if (signal_pending(current)) {
758 kfree_skb(skb);
759 return sock_intr_errno(timeo);
760 }
761 return 1;
762 }
763 skb_set_owner_r(skb, sk);
764 return 0;
765 }
766
767 int netlink_sendskb(struct sock *sk, struct sk_buff *skb, int protocol)
768 {
769 int len = skb->len;
770
771 skb_queue_tail(&sk->sk_receive_queue, skb);
772 sk->sk_data_ready(sk, len);
773 sock_put(sk);
774 return len;
775 }
776
777 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
778 {
779 kfree_skb(skb);
780 sock_put(sk);
781 }
782
783 static inline struct sk_buff *netlink_trim(struct sk_buff *skb,
784 gfp_t allocation)
785 {
786 int delta;
787
788 skb_orphan(skb);
789
790 delta = skb->end - skb->tail;
791 if (delta * 2 < skb->truesize)
792 return skb;
793
794 if (skb_shared(skb)) {
795 struct sk_buff *nskb = skb_clone(skb, allocation);
796 if (!nskb)
797 return skb;
798 kfree_skb(skb);
799 skb = nskb;
800 }
801
802 if (!pskb_expand_head(skb, 0, -delta, allocation))
803 skb->truesize -= delta;
804
805 return skb;
806 }
807
808 int netlink_unicast(struct sock *ssk, struct sk_buff *skb, u32 pid, int nonblock)
809 {
810 struct sock *sk;
811 int err;
812 long timeo;
813
814 skb = netlink_trim(skb, gfp_any());
815
816 timeo = sock_sndtimeo(ssk, nonblock);
817 retry:
818 sk = netlink_getsockbypid(ssk, pid);
819 if (IS_ERR(sk)) {
820 kfree_skb(skb);
821 return PTR_ERR(sk);
822 }
823 err = netlink_attachskb(sk, skb, nonblock, timeo, ssk);
824 if (err == 1)
825 goto retry;
826 if (err)
827 return err;
828
829 return netlink_sendskb(sk, skb, ssk->sk_protocol);
830 }
831
832 int netlink_has_listeners(struct sock *sk, unsigned int group)
833 {
834 int res = 0;
835
836 BUG_ON(!(nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET));
837 if (group - 1 < nl_table[sk->sk_protocol].groups)
838 res = test_bit(group - 1, nl_table[sk->sk_protocol].listeners);
839 return res;
840 }
841 EXPORT_SYMBOL_GPL(netlink_has_listeners);
842
843 static __inline__ int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
844 {
845 struct netlink_sock *nlk = nlk_sk(sk);
846
847 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
848 !test_bit(0, &nlk->state)) {
849 skb_set_owner_r(skb, sk);
850 skb_queue_tail(&sk->sk_receive_queue, skb);
851 sk->sk_data_ready(sk, skb->len);
852 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf;
853 }
854 return -1;
855 }
856
857 struct netlink_broadcast_data {
858 struct sock *exclude_sk;
859 u32 pid;
860 u32 group;
861 int failure;
862 int congested;
863 int delivered;
864 gfp_t allocation;
865 struct sk_buff *skb, *skb2;
866 };
867
868 static inline int do_one_broadcast(struct sock *sk,
869 struct netlink_broadcast_data *p)
870 {
871 struct netlink_sock *nlk = nlk_sk(sk);
872 int val;
873
874 if (p->exclude_sk == sk)
875 goto out;
876
877 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
878 !test_bit(p->group - 1, nlk->groups))
879 goto out;
880
881 if (p->failure) {
882 netlink_overrun(sk);
883 goto out;
884 }
885
886 sock_hold(sk);
887 if (p->skb2 == NULL) {
888 if (skb_shared(p->skb)) {
889 p->skb2 = skb_clone(p->skb, p->allocation);
890 } else {
891 p->skb2 = skb_get(p->skb);
892 /*
893 * skb ownership may have been set when
894 * delivered to a previous socket.
895 */
896 skb_orphan(p->skb2);
897 }
898 }
899 if (p->skb2 == NULL) {
900 netlink_overrun(sk);
901 /* Clone failed. Notify ALL listeners. */
902 p->failure = 1;
903 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
904 netlink_overrun(sk);
905 } else {
906 p->congested |= val;
907 p->delivered = 1;
908 p->skb2 = NULL;
909 }
910 sock_put(sk);
911
912 out:
913 return 0;
914 }
915
916 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
917 u32 group, gfp_t allocation)
918 {
919 struct netlink_broadcast_data info;
920 struct hlist_node *node;
921 struct sock *sk;
922
923 skb = netlink_trim(skb, allocation);
924
925 info.exclude_sk = ssk;
926 info.pid = pid;
927 info.group = group;
928 info.failure = 0;
929 info.congested = 0;
930 info.delivered = 0;
931 info.allocation = allocation;
932 info.skb = skb;
933 info.skb2 = NULL;
934
935 /* While we sleep in clone, do not allow to change socket list */
936
937 netlink_lock_table();
938
939 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
940 do_one_broadcast(sk, &info);
941
942 kfree_skb(skb);
943
944 netlink_unlock_table();
945
946 if (info.skb2)
947 kfree_skb(info.skb2);
948
949 if (info.delivered) {
950 if (info.congested && (allocation & __GFP_WAIT))
951 yield();
952 return 0;
953 }
954 if (info.failure)
955 return -ENOBUFS;
956 return -ESRCH;
957 }
958
959 struct netlink_set_err_data {
960 struct sock *exclude_sk;
961 u32 pid;
962 u32 group;
963 int code;
964 };
965
966 static inline int do_one_set_err(struct sock *sk,
967 struct netlink_set_err_data *p)
968 {
969 struct netlink_sock *nlk = nlk_sk(sk);
970
971 if (sk == p->exclude_sk)
972 goto out;
973
974 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
975 !test_bit(p->group - 1, nlk->groups))
976 goto out;
977
978 sk->sk_err = p->code;
979 sk->sk_error_report(sk);
980 out:
981 return 0;
982 }
983
984 void netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
985 {
986 struct netlink_set_err_data info;
987 struct hlist_node *node;
988 struct sock *sk;
989
990 info.exclude_sk = ssk;
991 info.pid = pid;
992 info.group = group;
993 info.code = code;
994
995 read_lock(&nl_table_lock);
996
997 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
998 do_one_set_err(sk, &info);
999
1000 read_unlock(&nl_table_lock);
1001 }
1002
1003 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1004 char __user *optval, int optlen)
1005 {
1006 struct sock *sk = sock->sk;
1007 struct netlink_sock *nlk = nlk_sk(sk);
1008 int val = 0, err;
1009
1010 if (level != SOL_NETLINK)
1011 return -ENOPROTOOPT;
1012
1013 if (optlen >= sizeof(int) &&
1014 get_user(val, (int __user *)optval))
1015 return -EFAULT;
1016
1017 switch (optname) {
1018 case NETLINK_PKTINFO:
1019 if (val)
1020 nlk->flags |= NETLINK_RECV_PKTINFO;
1021 else
1022 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1023 err = 0;
1024 break;
1025 case NETLINK_ADD_MEMBERSHIP:
1026 case NETLINK_DROP_MEMBERSHIP: {
1027 unsigned int subscriptions;
1028 int old, new = optname == NETLINK_ADD_MEMBERSHIP ? 1 : 0;
1029
1030 if (!netlink_capable(sock, NL_NONROOT_RECV))
1031 return -EPERM;
1032 if (nlk->groups == NULL) {
1033 err = netlink_alloc_groups(sk);
1034 if (err)
1035 return err;
1036 }
1037 if (!val || val - 1 >= nlk->ngroups)
1038 return -EINVAL;
1039 netlink_table_grab();
1040 old = test_bit(val - 1, nlk->groups);
1041 subscriptions = nlk->subscriptions - old + new;
1042 if (new)
1043 __set_bit(val - 1, nlk->groups);
1044 else
1045 __clear_bit(val - 1, nlk->groups);
1046 netlink_update_subscriptions(sk, subscriptions);
1047 netlink_update_listeners(sk);
1048 netlink_table_ungrab();
1049 err = 0;
1050 break;
1051 }
1052 default:
1053 err = -ENOPROTOOPT;
1054 }
1055 return err;
1056 }
1057
1058 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1059 char __user *optval, int __user *optlen)
1060 {
1061 struct sock *sk = sock->sk;
1062 struct netlink_sock *nlk = nlk_sk(sk);
1063 int len, val, err;
1064
1065 if (level != SOL_NETLINK)
1066 return -ENOPROTOOPT;
1067
1068 if (get_user(len, optlen))
1069 return -EFAULT;
1070 if (len < 0)
1071 return -EINVAL;
1072
1073 switch (optname) {
1074 case NETLINK_PKTINFO:
1075 if (len < sizeof(int))
1076 return -EINVAL;
1077 len = sizeof(int);
1078 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
1079 put_user(len, optlen);
1080 put_user(val, optval);
1081 err = 0;
1082 break;
1083 default:
1084 err = -ENOPROTOOPT;
1085 }
1086 return err;
1087 }
1088
1089 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1090 {
1091 struct nl_pktinfo info;
1092
1093 info.group = NETLINK_CB(skb).dst_group;
1094 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1095 }
1096
1097 static inline void netlink_rcv_wake(struct sock *sk)
1098 {
1099 struct netlink_sock *nlk = nlk_sk(sk);
1100
1101 if (skb_queue_empty(&sk->sk_receive_queue))
1102 clear_bit(0, &nlk->state);
1103 if (!test_bit(0, &nlk->state))
1104 wake_up_interruptible(&nlk->wait);
1105 }
1106
1107 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1108 struct msghdr *msg, size_t len)
1109 {
1110 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1111 struct sock *sk = sock->sk;
1112 struct netlink_sock *nlk = nlk_sk(sk);
1113 struct sockaddr_nl *addr=msg->msg_name;
1114 u32 dst_pid;
1115 u32 dst_group;
1116 struct sk_buff *skb;
1117 int err;
1118 struct scm_cookie scm;
1119
1120 if (msg->msg_flags&MSG_OOB)
1121 return -EOPNOTSUPP;
1122
1123 if (NULL == siocb->scm)
1124 siocb->scm = &scm;
1125 err = scm_send(sock, msg, siocb->scm);
1126 if (err < 0)
1127 return err;
1128
1129 if (msg->msg_namelen) {
1130 if (addr->nl_family != AF_NETLINK)
1131 return -EINVAL;
1132 dst_pid = addr->nl_pid;
1133 dst_group = ffs(addr->nl_groups);
1134 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
1135 return -EPERM;
1136 } else {
1137 dst_pid = nlk->dst_pid;
1138 dst_group = nlk->dst_group;
1139 }
1140
1141 if (!nlk->pid) {
1142 err = netlink_autobind(sock);
1143 if (err)
1144 goto out;
1145 }
1146
1147 err = -EMSGSIZE;
1148 if (len > sk->sk_sndbuf - 32)
1149 goto out;
1150 err = -ENOBUFS;
1151 skb = alloc_skb(len, GFP_KERNEL);
1152 if (skb==NULL)
1153 goto out;
1154
1155 NETLINK_CB(skb).pid = nlk->pid;
1156 NETLINK_CB(skb).dst_pid = dst_pid;
1157 NETLINK_CB(skb).dst_group = dst_group;
1158 NETLINK_CB(skb).loginuid = audit_get_loginuid(current->audit_context);
1159 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1160
1161 /* What can I do? Netlink is asynchronous, so that
1162 we will have to save current capabilities to
1163 check them, when this message will be delivered
1164 to corresponding kernel module. --ANK (980802)
1165 */
1166
1167 err = -EFAULT;
1168 if (memcpy_fromiovec(skb_put(skb,len), msg->msg_iov, len)) {
1169 kfree_skb(skb);
1170 goto out;
1171 }
1172
1173 err = security_netlink_send(sk, skb);
1174 if (err) {
1175 kfree_skb(skb);
1176 goto out;
1177 }
1178
1179 if (dst_group) {
1180 atomic_inc(&skb->users);
1181 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
1182 }
1183 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1184
1185 out:
1186 return err;
1187 }
1188
1189 static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1190 struct msghdr *msg, size_t len,
1191 int flags)
1192 {
1193 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1194 struct scm_cookie scm;
1195 struct sock *sk = sock->sk;
1196 struct netlink_sock *nlk = nlk_sk(sk);
1197 int noblock = flags&MSG_DONTWAIT;
1198 size_t copied;
1199 struct sk_buff *skb;
1200 int err;
1201
1202 if (flags&MSG_OOB)
1203 return -EOPNOTSUPP;
1204
1205 copied = 0;
1206
1207 skb = skb_recv_datagram(sk,flags,noblock,&err);
1208 if (skb==NULL)
1209 goto out;
1210
1211 msg->msg_namelen = 0;
1212
1213 copied = skb->len;
1214 if (len < copied) {
1215 msg->msg_flags |= MSG_TRUNC;
1216 copied = len;
1217 }
1218
1219 skb->h.raw = skb->data;
1220 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1221
1222 if (msg->msg_name) {
1223 struct sockaddr_nl *addr = (struct sockaddr_nl*)msg->msg_name;
1224 addr->nl_family = AF_NETLINK;
1225 addr->nl_pad = 0;
1226 addr->nl_pid = NETLINK_CB(skb).pid;
1227 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1228 msg->msg_namelen = sizeof(*addr);
1229 }
1230
1231 if (nlk->flags & NETLINK_RECV_PKTINFO)
1232 netlink_cmsg_recv_pktinfo(msg, skb);
1233
1234 if (NULL == siocb->scm) {
1235 memset(&scm, 0, sizeof(scm));
1236 siocb->scm = &scm;
1237 }
1238 siocb->scm->creds = *NETLINK_CREDS(skb);
1239 skb_free_datagram(sk, skb);
1240
1241 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2)
1242 netlink_dump(sk);
1243
1244 scm_recv(sock, msg, siocb->scm, flags);
1245
1246 out:
1247 netlink_rcv_wake(sk);
1248 return err ? : copied;
1249 }
1250
1251 static void netlink_data_ready(struct sock *sk, int len)
1252 {
1253 struct netlink_sock *nlk = nlk_sk(sk);
1254
1255 if (nlk->data_ready)
1256 nlk->data_ready(sk, len);
1257 netlink_rcv_wake(sk);
1258 }
1259
1260 /*
1261 * We export these functions to other modules. They provide a
1262 * complete set of kernel non-blocking support for message
1263 * queueing.
1264 */
1265
1266 struct sock *
1267 netlink_kernel_create(int unit, unsigned int groups,
1268 void (*input)(struct sock *sk, int len),
1269 struct module *module)
1270 {
1271 struct socket *sock;
1272 struct sock *sk;
1273 struct netlink_sock *nlk;
1274 unsigned long *listeners = NULL;
1275
1276 if (!nl_table)
1277 return NULL;
1278
1279 if (unit<0 || unit>=MAX_LINKS)
1280 return NULL;
1281
1282 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1283 return NULL;
1284
1285 if (__netlink_create(sock, unit) < 0)
1286 goto out_sock_release;
1287
1288 if (groups < 32)
1289 groups = 32;
1290
1291 listeners = kzalloc(NLGRPSZ(groups), GFP_KERNEL);
1292 if (!listeners)
1293 goto out_sock_release;
1294
1295 sk = sock->sk;
1296 sk->sk_data_ready = netlink_data_ready;
1297 if (input)
1298 nlk_sk(sk)->data_ready = input;
1299
1300 if (netlink_insert(sk, 0))
1301 goto out_sock_release;
1302
1303 nlk = nlk_sk(sk);
1304 nlk->flags |= NETLINK_KERNEL_SOCKET;
1305
1306 netlink_table_grab();
1307 nl_table[unit].groups = groups;
1308 nl_table[unit].listeners = listeners;
1309 nl_table[unit].module = module;
1310 nl_table[unit].registered = 1;
1311 netlink_table_ungrab();
1312
1313 return sk;
1314
1315 out_sock_release:
1316 kfree(listeners);
1317 sock_release(sock);
1318 return NULL;
1319 }
1320
1321 void netlink_set_nonroot(int protocol, unsigned int flags)
1322 {
1323 if ((unsigned int)protocol < MAX_LINKS)
1324 nl_table[protocol].nl_nonroot = flags;
1325 }
1326
1327 static void netlink_destroy_callback(struct netlink_callback *cb)
1328 {
1329 if (cb->skb)
1330 kfree_skb(cb->skb);
1331 kfree(cb);
1332 }
1333
1334 /*
1335 * It looks a bit ugly.
1336 * It would be better to create kernel thread.
1337 */
1338
1339 static int netlink_dump(struct sock *sk)
1340 {
1341 struct netlink_sock *nlk = nlk_sk(sk);
1342 struct netlink_callback *cb;
1343 struct sk_buff *skb;
1344 struct nlmsghdr *nlh;
1345 int len;
1346
1347 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL);
1348 if (!skb)
1349 return -ENOBUFS;
1350
1351 spin_lock(&nlk->cb_lock);
1352
1353 cb = nlk->cb;
1354 if (cb == NULL) {
1355 spin_unlock(&nlk->cb_lock);
1356 kfree_skb(skb);
1357 return -EINVAL;
1358 }
1359
1360 len = cb->dump(skb, cb);
1361
1362 if (len > 0) {
1363 spin_unlock(&nlk->cb_lock);
1364 skb_queue_tail(&sk->sk_receive_queue, skb);
1365 sk->sk_data_ready(sk, len);
1366 return 0;
1367 }
1368
1369 nlh = NLMSG_NEW_ANSWER(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1370 memcpy(NLMSG_DATA(nlh), &len, sizeof(len));
1371 skb_queue_tail(&sk->sk_receive_queue, skb);
1372 sk->sk_data_ready(sk, skb->len);
1373
1374 if (cb->done)
1375 cb->done(cb);
1376 nlk->cb = NULL;
1377 spin_unlock(&nlk->cb_lock);
1378
1379 netlink_destroy_callback(cb);
1380 return 0;
1381
1382 nlmsg_failure:
1383 return -ENOBUFS;
1384 }
1385
1386 int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1387 struct nlmsghdr *nlh,
1388 int (*dump)(struct sk_buff *skb, struct netlink_callback*),
1389 int (*done)(struct netlink_callback*))
1390 {
1391 struct netlink_callback *cb;
1392 struct sock *sk;
1393 struct netlink_sock *nlk;
1394
1395 cb = kmalloc(sizeof(*cb), GFP_KERNEL);
1396 if (cb == NULL)
1397 return -ENOBUFS;
1398
1399 memset(cb, 0, sizeof(*cb));
1400 cb->dump = dump;
1401 cb->done = done;
1402 cb->nlh = nlh;
1403 atomic_inc(&skb->users);
1404 cb->skb = skb;
1405
1406 sk = netlink_lookup(ssk->sk_protocol, NETLINK_CB(skb).pid);
1407 if (sk == NULL) {
1408 netlink_destroy_callback(cb);
1409 return -ECONNREFUSED;
1410 }
1411 nlk = nlk_sk(sk);
1412 /* A dump is in progress... */
1413 spin_lock(&nlk->cb_lock);
1414 if (nlk->cb) {
1415 spin_unlock(&nlk->cb_lock);
1416 netlink_destroy_callback(cb);
1417 sock_put(sk);
1418 return -EBUSY;
1419 }
1420 nlk->cb = cb;
1421 spin_unlock(&nlk->cb_lock);
1422
1423 netlink_dump(sk);
1424 sock_put(sk);
1425 return 0;
1426 }
1427
1428 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1429 {
1430 struct sk_buff *skb;
1431 struct nlmsghdr *rep;
1432 struct nlmsgerr *errmsg;
1433 int size;
1434
1435 if (err == 0)
1436 size = NLMSG_SPACE(sizeof(struct nlmsgerr));
1437 else
1438 size = NLMSG_SPACE(4 + NLMSG_ALIGN(nlh->nlmsg_len));
1439
1440 skb = alloc_skb(size, GFP_KERNEL);
1441 if (!skb) {
1442 struct sock *sk;
1443
1444 sk = netlink_lookup(in_skb->sk->sk_protocol,
1445 NETLINK_CB(in_skb).pid);
1446 if (sk) {
1447 sk->sk_err = ENOBUFS;
1448 sk->sk_error_report(sk);
1449 sock_put(sk);
1450 }
1451 return;
1452 }
1453
1454 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
1455 NLMSG_ERROR, sizeof(struct nlmsgerr), 0);
1456 errmsg = NLMSG_DATA(rep);
1457 errmsg->error = err;
1458 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(struct nlmsghdr));
1459 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1460 }
1461
1462 static int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
1463 struct nlmsghdr *, int *))
1464 {
1465 unsigned int total_len;
1466 struct nlmsghdr *nlh;
1467 int err;
1468
1469 while (skb->len >= nlmsg_total_size(0)) {
1470 nlh = (struct nlmsghdr *) skb->data;
1471
1472 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
1473 return 0;
1474
1475 total_len = min(NLMSG_ALIGN(nlh->nlmsg_len), skb->len);
1476
1477 if (cb(skb, nlh, &err) < 0) {
1478 /* Not an error, but we have to interrupt processing
1479 * here. Note: that in this case we do not pull
1480 * message from skb, it will be processed later.
1481 */
1482 if (err == 0)
1483 return -1;
1484 netlink_ack(skb, nlh, err);
1485 } else if (nlh->nlmsg_flags & NLM_F_ACK)
1486 netlink_ack(skb, nlh, 0);
1487
1488 skb_pull(skb, total_len);
1489 }
1490
1491 return 0;
1492 }
1493
1494 /**
1495 * nelink_run_queue - Process netlink receive queue.
1496 * @sk: Netlink socket containing the queue
1497 * @qlen: Place to store queue length upon entry
1498 * @cb: Callback function invoked for each netlink message found
1499 *
1500 * Processes as much as there was in the queue upon entry and invokes
1501 * a callback function for each netlink message found. The callback
1502 * function may refuse a message by returning a negative error code
1503 * but setting the error pointer to 0 in which case this function
1504 * returns with a qlen != 0.
1505 *
1506 * qlen must be initialized to 0 before the initial entry, afterwards
1507 * the function may be called repeatedly until qlen reaches 0.
1508 */
1509 void netlink_run_queue(struct sock *sk, unsigned int *qlen,
1510 int (*cb)(struct sk_buff *, struct nlmsghdr *, int *))
1511 {
1512 struct sk_buff *skb;
1513
1514 if (!*qlen || *qlen > skb_queue_len(&sk->sk_receive_queue))
1515 *qlen = skb_queue_len(&sk->sk_receive_queue);
1516
1517 for (; *qlen; (*qlen)--) {
1518 skb = skb_dequeue(&sk->sk_receive_queue);
1519 if (netlink_rcv_skb(skb, cb)) {
1520 if (skb->len)
1521 skb_queue_head(&sk->sk_receive_queue, skb);
1522 else {
1523 kfree_skb(skb);
1524 (*qlen)--;
1525 }
1526 break;
1527 }
1528
1529 kfree_skb(skb);
1530 }
1531 }
1532
1533 /**
1534 * netlink_queue_skip - Skip netlink message while processing queue.
1535 * @nlh: Netlink message to be skipped
1536 * @skb: Socket buffer containing the netlink messages.
1537 *
1538 * Pulls the given netlink message off the socket buffer so the next
1539 * call to netlink_queue_run() will not reconsider the message.
1540 */
1541 void netlink_queue_skip(struct nlmsghdr *nlh, struct sk_buff *skb)
1542 {
1543 int msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1544
1545 if (msglen > skb->len)
1546 msglen = skb->len;
1547
1548 skb_pull(skb, msglen);
1549 }
1550
1551 #ifdef CONFIG_PROC_FS
1552 struct nl_seq_iter {
1553 int link;
1554 int hash_idx;
1555 };
1556
1557 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1558 {
1559 struct nl_seq_iter *iter = seq->private;
1560 int i, j;
1561 struct sock *s;
1562 struct hlist_node *node;
1563 loff_t off = 0;
1564
1565 for (i=0; i<MAX_LINKS; i++) {
1566 struct nl_pid_hash *hash = &nl_table[i].hash;
1567
1568 for (j = 0; j <= hash->mask; j++) {
1569 sk_for_each(s, node, &hash->table[j]) {
1570 if (off == pos) {
1571 iter->link = i;
1572 iter->hash_idx = j;
1573 return s;
1574 }
1575 ++off;
1576 }
1577 }
1578 }
1579 return NULL;
1580 }
1581
1582 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
1583 {
1584 read_lock(&nl_table_lock);
1585 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1586 }
1587
1588 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1589 {
1590 struct sock *s;
1591 struct nl_seq_iter *iter;
1592 int i, j;
1593
1594 ++*pos;
1595
1596 if (v == SEQ_START_TOKEN)
1597 return netlink_seq_socket_idx(seq, 0);
1598
1599 s = sk_next(v);
1600 if (s)
1601 return s;
1602
1603 iter = seq->private;
1604 i = iter->link;
1605 j = iter->hash_idx + 1;
1606
1607 do {
1608 struct nl_pid_hash *hash = &nl_table[i].hash;
1609
1610 for (; j <= hash->mask; j++) {
1611 s = sk_head(&hash->table[j]);
1612 if (s) {
1613 iter->link = i;
1614 iter->hash_idx = j;
1615 return s;
1616 }
1617 }
1618
1619 j = 0;
1620 } while (++i < MAX_LINKS);
1621
1622 return NULL;
1623 }
1624
1625 static void netlink_seq_stop(struct seq_file *seq, void *v)
1626 {
1627 read_unlock(&nl_table_lock);
1628 }
1629
1630
1631 static int netlink_seq_show(struct seq_file *seq, void *v)
1632 {
1633 if (v == SEQ_START_TOKEN)
1634 seq_puts(seq,
1635 "sk Eth Pid Groups "
1636 "Rmem Wmem Dump Locks\n");
1637 else {
1638 struct sock *s = v;
1639 struct netlink_sock *nlk = nlk_sk(s);
1640
1641 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1642 s,
1643 s->sk_protocol,
1644 nlk->pid,
1645 nlk->groups ? (u32)nlk->groups[0] : 0,
1646 atomic_read(&s->sk_rmem_alloc),
1647 atomic_read(&s->sk_wmem_alloc),
1648 nlk->cb,
1649 atomic_read(&s->sk_refcnt)
1650 );
1651
1652 }
1653 return 0;
1654 }
1655
1656 static struct seq_operations netlink_seq_ops = {
1657 .start = netlink_seq_start,
1658 .next = netlink_seq_next,
1659 .stop = netlink_seq_stop,
1660 .show = netlink_seq_show,
1661 };
1662
1663
1664 static int netlink_seq_open(struct inode *inode, struct file *file)
1665 {
1666 struct seq_file *seq;
1667 struct nl_seq_iter *iter;
1668 int err;
1669
1670 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
1671 if (!iter)
1672 return -ENOMEM;
1673
1674 err = seq_open(file, &netlink_seq_ops);
1675 if (err) {
1676 kfree(iter);
1677 return err;
1678 }
1679
1680 memset(iter, 0, sizeof(*iter));
1681 seq = file->private_data;
1682 seq->private = iter;
1683 return 0;
1684 }
1685
1686 static struct file_operations netlink_seq_fops = {
1687 .owner = THIS_MODULE,
1688 .open = netlink_seq_open,
1689 .read = seq_read,
1690 .llseek = seq_lseek,
1691 .release = seq_release_private,
1692 };
1693
1694 #endif
1695
1696 int netlink_register_notifier(struct notifier_block *nb)
1697 {
1698 return notifier_chain_register(&netlink_chain, nb);
1699 }
1700
1701 int netlink_unregister_notifier(struct notifier_block *nb)
1702 {
1703 return notifier_chain_unregister(&netlink_chain, nb);
1704 }
1705
1706 static const struct proto_ops netlink_ops = {
1707 .family = PF_NETLINK,
1708 .owner = THIS_MODULE,
1709 .release = netlink_release,
1710 .bind = netlink_bind,
1711 .connect = netlink_connect,
1712 .socketpair = sock_no_socketpair,
1713 .accept = sock_no_accept,
1714 .getname = netlink_getname,
1715 .poll = datagram_poll,
1716 .ioctl = sock_no_ioctl,
1717 .listen = sock_no_listen,
1718 .shutdown = sock_no_shutdown,
1719 .setsockopt = netlink_setsockopt,
1720 .getsockopt = netlink_getsockopt,
1721 .sendmsg = netlink_sendmsg,
1722 .recvmsg = netlink_recvmsg,
1723 .mmap = sock_no_mmap,
1724 .sendpage = sock_no_sendpage,
1725 };
1726
1727 static struct net_proto_family netlink_family_ops = {
1728 .family = PF_NETLINK,
1729 .create = netlink_create,
1730 .owner = THIS_MODULE, /* for consistency 8) */
1731 };
1732
1733 extern void netlink_skb_parms_too_large(void);
1734
1735 static int __init netlink_proto_init(void)
1736 {
1737 struct sk_buff *dummy_skb;
1738 int i;
1739 unsigned long max;
1740 unsigned int order;
1741 int err = proto_register(&netlink_proto, 0);
1742
1743 if (err != 0)
1744 goto out;
1745
1746 if (sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb))
1747 netlink_skb_parms_too_large();
1748
1749 nl_table = kmalloc(sizeof(*nl_table) * MAX_LINKS, GFP_KERNEL);
1750 if (!nl_table) {
1751 enomem:
1752 printk(KERN_CRIT "netlink_init: Cannot allocate nl_table\n");
1753 return -ENOMEM;
1754 }
1755
1756 memset(nl_table, 0, sizeof(*nl_table) * MAX_LINKS);
1757
1758 if (num_physpages >= (128 * 1024))
1759 max = num_physpages >> (21 - PAGE_SHIFT);
1760 else
1761 max = num_physpages >> (23 - PAGE_SHIFT);
1762
1763 order = get_bitmask_order(max) - 1 + PAGE_SHIFT;
1764 max = (1UL << order) / sizeof(struct hlist_head);
1765 order = get_bitmask_order(max > UINT_MAX ? UINT_MAX : max) - 1;
1766
1767 for (i = 0; i < MAX_LINKS; i++) {
1768 struct nl_pid_hash *hash = &nl_table[i].hash;
1769
1770 hash->table = nl_pid_hash_alloc(1 * sizeof(*hash->table));
1771 if (!hash->table) {
1772 while (i-- > 0)
1773 nl_pid_hash_free(nl_table[i].hash.table,
1774 1 * sizeof(*hash->table));
1775 kfree(nl_table);
1776 goto enomem;
1777 }
1778 memset(hash->table, 0, 1 * sizeof(*hash->table));
1779 hash->max_shift = order;
1780 hash->shift = 0;
1781 hash->mask = 0;
1782 hash->rehash_time = jiffies;
1783 }
1784
1785 sock_register(&netlink_family_ops);
1786 #ifdef CONFIG_PROC_FS
1787 proc_net_fops_create("netlink", 0, &netlink_seq_fops);
1788 #endif
1789 /* The netlink device handler may be needed early. */
1790 rtnetlink_init();
1791 out:
1792 return err;
1793 }
1794
1795 core_initcall(netlink_proto_init);
1796
1797 EXPORT_SYMBOL(netlink_ack);
1798 EXPORT_SYMBOL(netlink_run_queue);
1799 EXPORT_SYMBOL(netlink_queue_skip);
1800 EXPORT_SYMBOL(netlink_broadcast);
1801 EXPORT_SYMBOL(netlink_dump_start);
1802 EXPORT_SYMBOL(netlink_kernel_create);
1803 EXPORT_SYMBOL(netlink_register_notifier);
1804 EXPORT_SYMBOL(netlink_set_err);
1805 EXPORT_SYMBOL(netlink_set_nonroot);
1806 EXPORT_SYMBOL(netlink_unicast);
1807 EXPORT_SYMBOL(netlink_unregister_notifier);
1808
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree