2 * NETLINK Kernel-user communication protocol.
4 * Authors: Alan Cox <alan@redhat.com>
5 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
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.
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
24 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/kernel.h>
28 #include <linux/init.h>
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>
36 #include <linux/fcntl.h>
37 #include <linux/termios.h>
38 #include <linux/sockios.h>
39 #include <linux/net.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/notifier.h>
49 #include <linux/security.h>
50 #include <linux/jhash.h>
51 #include <linux/jiffies.h>
52 #include <linux/random.h>
53 #include <linux/bitops.h>
55 #include <linux/types.h>
56 #include <linux/audit.h>
57 #include <linux/selinux.h>
58 #include <linux/mutex.h>
60 #include <net/net_namespace.h>
63 #include <net/netlink.h>
65 #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
66 #define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
69 /* struct sock has to be the first member of netlink_sock */
77 unsigned long *groups
;
79 wait_queue_head_t wait
;
80 struct netlink_callback
*cb
;
81 struct mutex
*cb_mutex
;
82 struct mutex cb_def_mutex
;
83 void (*data_ready
)(struct sock
*sk
, int bytes
);
84 struct module
*module
;
87 #define NETLINK_KERNEL_SOCKET 0x1
88 #define NETLINK_RECV_PKTINFO 0x2
90 static inline struct netlink_sock
*nlk_sk(struct sock
*sk
)
92 return container_of(sk
, struct netlink_sock
, sk
);
96 struct hlist_head
*table
;
97 unsigned long rehash_time
;
102 unsigned int entries
;
103 unsigned int max_shift
;
108 struct netlink_table
{
109 struct nl_pid_hash hash
;
110 struct hlist_head mc_list
;
111 unsigned long *listeners
;
112 unsigned int nl_nonroot
;
114 struct mutex
*cb_mutex
;
115 struct module
*module
;
119 static struct netlink_table
*nl_table
;
121 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait
);
123 static int netlink_dump(struct sock
*sk
);
124 static void netlink_destroy_callback(struct netlink_callback
*cb
);
125 static void netlink_queue_skip(struct nlmsghdr
*nlh
, struct sk_buff
*skb
);
127 static DEFINE_RWLOCK(nl_table_lock
);
128 static atomic_t nl_table_users
= ATOMIC_INIT(0);
130 static ATOMIC_NOTIFIER_HEAD(netlink_chain
);
132 static u32
netlink_group_mask(u32 group
)
134 return group
? 1 << (group
- 1) : 0;
137 static struct hlist_head
*nl_pid_hashfn(struct nl_pid_hash
*hash
, u32 pid
)
139 return &hash
->table
[jhash_1word(pid
, hash
->rnd
) & hash
->mask
];
142 static void netlink_sock_destruct(struct sock
*sk
)
144 struct netlink_sock
*nlk
= nlk_sk(sk
);
148 nlk
->cb
->done(nlk
->cb
);
149 netlink_destroy_callback(nlk
->cb
);
152 skb_queue_purge(&sk
->sk_receive_queue
);
154 if (!sock_flag(sk
, SOCK_DEAD
)) {
155 printk("Freeing alive netlink socket %p\n", sk
);
158 BUG_TRAP(!atomic_read(&sk
->sk_rmem_alloc
));
159 BUG_TRAP(!atomic_read(&sk
->sk_wmem_alloc
));
160 BUG_TRAP(!nlk_sk(sk
)->groups
);
163 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
164 * Look, when several writers sleep and reader wakes them up, all but one
165 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
166 * this, _but_ remember, it adds useless work on UP machines.
169 static void netlink_table_grab(void)
171 write_lock_irq(&nl_table_lock
);
173 if (atomic_read(&nl_table_users
)) {
174 DECLARE_WAITQUEUE(wait
, current
);
176 add_wait_queue_exclusive(&nl_table_wait
, &wait
);
178 set_current_state(TASK_UNINTERRUPTIBLE
);
179 if (atomic_read(&nl_table_users
) == 0)
181 write_unlock_irq(&nl_table_lock
);
183 write_lock_irq(&nl_table_lock
);
186 __set_current_state(TASK_RUNNING
);
187 remove_wait_queue(&nl_table_wait
, &wait
);
191 static __inline__
void netlink_table_ungrab(void)
193 write_unlock_irq(&nl_table_lock
);
194 wake_up(&nl_table_wait
);
197 static __inline__
void
198 netlink_lock_table(void)
200 /* read_lock() synchronizes us to netlink_table_grab */
202 read_lock(&nl_table_lock
);
203 atomic_inc(&nl_table_users
);
204 read_unlock(&nl_table_lock
);
207 static __inline__
void
208 netlink_unlock_table(void)
210 if (atomic_dec_and_test(&nl_table_users
))
211 wake_up(&nl_table_wait
);
214 static __inline__
struct sock
*netlink_lookup(int protocol
, u32 pid
)
216 struct nl_pid_hash
*hash
= &nl_table
[protocol
].hash
;
217 struct hlist_head
*head
;
219 struct hlist_node
*node
;
221 read_lock(&nl_table_lock
);
222 head
= nl_pid_hashfn(hash
, pid
);
223 sk_for_each(sk
, node
, head
) {
224 if (nlk_sk(sk
)->pid
== pid
) {
231 read_unlock(&nl_table_lock
);
235 static inline struct hlist_head
*nl_pid_hash_alloc(size_t size
)
237 if (size
<= PAGE_SIZE
)
238 return kmalloc(size
, GFP_ATOMIC
);
240 return (struct hlist_head
*)
241 __get_free_pages(GFP_ATOMIC
, get_order(size
));
244 static inline void nl_pid_hash_free(struct hlist_head
*table
, size_t size
)
246 if (size
<= PAGE_SIZE
)
249 free_pages((unsigned long)table
, get_order(size
));
252 static int nl_pid_hash_rehash(struct nl_pid_hash
*hash
, int grow
)
254 unsigned int omask
, mask
, shift
;
256 struct hlist_head
*otable
, *table
;
259 omask
= mask
= hash
->mask
;
260 osize
= size
= (mask
+ 1) * sizeof(*table
);
264 if (++shift
> hash
->max_shift
)
270 table
= nl_pid_hash_alloc(size
);
274 memset(table
, 0, size
);
275 otable
= hash
->table
;
279 get_random_bytes(&hash
->rnd
, sizeof(hash
->rnd
));
281 for (i
= 0; i
<= omask
; i
++) {
283 struct hlist_node
*node
, *tmp
;
285 sk_for_each_safe(sk
, node
, tmp
, &otable
[i
])
286 __sk_add_node(sk
, nl_pid_hashfn(hash
, nlk_sk(sk
)->pid
));
289 nl_pid_hash_free(otable
, osize
);
290 hash
->rehash_time
= jiffies
+ 10 * 60 * HZ
;
294 static inline int nl_pid_hash_dilute(struct nl_pid_hash
*hash
, int len
)
296 int avg
= hash
->entries
>> hash
->shift
;
298 if (unlikely(avg
> 1) && nl_pid_hash_rehash(hash
, 1))
301 if (unlikely(len
> avg
) && time_after(jiffies
, hash
->rehash_time
)) {
302 nl_pid_hash_rehash(hash
, 0);
309 static const struct proto_ops netlink_ops
;
312 netlink_update_listeners(struct sock
*sk
)
314 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
315 struct hlist_node
*node
;
319 for (i
= 0; i
< NLGRPLONGS(tbl
->groups
); i
++) {
321 sk_for_each_bound(sk
, node
, &tbl
->mc_list
) {
322 if (i
< NLGRPLONGS(nlk_sk(sk
)->ngroups
))
323 mask
|= nlk_sk(sk
)->groups
[i
];
325 tbl
->listeners
[i
] = mask
;
327 /* this function is only called with the netlink table "grabbed", which
328 * makes sure updates are visible before bind or setsockopt return. */
331 static int netlink_insert(struct sock
*sk
, u32 pid
)
333 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
334 struct hlist_head
*head
;
335 int err
= -EADDRINUSE
;
337 struct hlist_node
*node
;
340 netlink_table_grab();
341 head
= nl_pid_hashfn(hash
, pid
);
343 sk_for_each(osk
, node
, head
) {
344 if (nlk_sk(osk
)->pid
== pid
)
356 if (BITS_PER_LONG
> 32 && unlikely(hash
->entries
>= UINT_MAX
))
359 if (len
&& nl_pid_hash_dilute(hash
, len
))
360 head
= nl_pid_hashfn(hash
, pid
);
362 nlk_sk(sk
)->pid
= pid
;
363 sk_add_node(sk
, head
);
367 netlink_table_ungrab();
371 static void netlink_remove(struct sock
*sk
)
373 netlink_table_grab();
374 if (sk_del_node_init(sk
))
375 nl_table
[sk
->sk_protocol
].hash
.entries
--;
376 if (nlk_sk(sk
)->subscriptions
)
377 __sk_del_bind_node(sk
);
378 netlink_table_ungrab();
381 static struct proto netlink_proto
= {
383 .owner
= THIS_MODULE
,
384 .obj_size
= sizeof(struct netlink_sock
),
387 static int __netlink_create(struct net
*net
, struct socket
*sock
,
388 struct mutex
*cb_mutex
, int protocol
)
391 struct netlink_sock
*nlk
;
393 sock
->ops
= &netlink_ops
;
395 sk
= sk_alloc(net
, PF_NETLINK
, GFP_KERNEL
, &netlink_proto
, 1);
399 sock_init_data(sock
, sk
);
403 nlk
->cb_mutex
= cb_mutex
;
405 nlk
->cb_mutex
= &nlk
->cb_def_mutex
;
406 mutex_init(nlk
->cb_mutex
);
408 init_waitqueue_head(&nlk
->wait
);
410 sk
->sk_destruct
= netlink_sock_destruct
;
411 sk
->sk_protocol
= protocol
;
415 static int netlink_create(struct net
*net
, struct socket
*sock
, int protocol
)
417 struct module
*module
= NULL
;
418 struct mutex
*cb_mutex
;
419 struct netlink_sock
*nlk
;
422 if (net
!= &init_net
)
423 return -EAFNOSUPPORT
;
425 sock
->state
= SS_UNCONNECTED
;
427 if (sock
->type
!= SOCK_RAW
&& sock
->type
!= SOCK_DGRAM
)
428 return -ESOCKTNOSUPPORT
;
430 if (protocol
<0 || protocol
>= MAX_LINKS
)
431 return -EPROTONOSUPPORT
;
433 netlink_lock_table();
435 if (!nl_table
[protocol
].registered
) {
436 netlink_unlock_table();
437 request_module("net-pf-%d-proto-%d", PF_NETLINK
, protocol
);
438 netlink_lock_table();
441 if (nl_table
[protocol
].registered
&&
442 try_module_get(nl_table
[protocol
].module
))
443 module
= nl_table
[protocol
].module
;
444 cb_mutex
= nl_table
[protocol
].cb_mutex
;
445 netlink_unlock_table();
447 if ((err
= __netlink_create(net
, sock
, cb_mutex
, protocol
)) < 0)
450 nlk
= nlk_sk(sock
->sk
);
451 nlk
->module
= module
;
460 static int netlink_release(struct socket
*sock
)
462 struct sock
*sk
= sock
->sk
;
463 struct netlink_sock
*nlk
;
473 * OK. Socket is unlinked, any packets that arrive now
478 wake_up_interruptible_all(&nlk
->wait
);
480 skb_queue_purge(&sk
->sk_write_queue
);
482 if (nlk
->pid
&& !nlk
->subscriptions
) {
483 struct netlink_notify n
= {
484 .protocol
= sk
->sk_protocol
,
487 atomic_notifier_call_chain(&netlink_chain
,
488 NETLINK_URELEASE
, &n
);
491 module_put(nlk
->module
);
493 netlink_table_grab();
494 if (nlk
->flags
& NETLINK_KERNEL_SOCKET
) {
495 kfree(nl_table
[sk
->sk_protocol
].listeners
);
496 nl_table
[sk
->sk_protocol
].module
= NULL
;
497 nl_table
[sk
->sk_protocol
].registered
= 0;
498 } else if (nlk
->subscriptions
)
499 netlink_update_listeners(sk
);
500 netlink_table_ungrab();
509 static int netlink_autobind(struct socket
*sock
)
511 struct sock
*sk
= sock
->sk
;
512 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
513 struct hlist_head
*head
;
515 struct hlist_node
*node
;
516 s32 pid
= current
->tgid
;
518 static s32 rover
= -4097;
522 netlink_table_grab();
523 head
= nl_pid_hashfn(hash
, pid
);
524 sk_for_each(osk
, node
, head
) {
525 if (nlk_sk(osk
)->pid
== pid
) {
526 /* Bind collision, search negative pid values. */
530 netlink_table_ungrab();
534 netlink_table_ungrab();
536 err
= netlink_insert(sk
, pid
);
537 if (err
== -EADDRINUSE
)
540 /* If 2 threads race to autobind, that is fine. */
547 static inline int netlink_capable(struct socket
*sock
, unsigned int flag
)
549 return (nl_table
[sock
->sk
->sk_protocol
].nl_nonroot
& flag
) ||
550 capable(CAP_NET_ADMIN
);
554 netlink_update_subscriptions(struct sock
*sk
, unsigned int subscriptions
)
556 struct netlink_sock
*nlk
= nlk_sk(sk
);
558 if (nlk
->subscriptions
&& !subscriptions
)
559 __sk_del_bind_node(sk
);
560 else if (!nlk
->subscriptions
&& subscriptions
)
561 sk_add_bind_node(sk
, &nl_table
[sk
->sk_protocol
].mc_list
);
562 nlk
->subscriptions
= subscriptions
;
565 static int netlink_realloc_groups(struct sock
*sk
)
567 struct netlink_sock
*nlk
= nlk_sk(sk
);
569 unsigned long *new_groups
;
572 netlink_table_grab();
574 groups
= nl_table
[sk
->sk_protocol
].groups
;
575 if (!nl_table
[sk
->sk_protocol
].registered
) {
580 if (nlk
->ngroups
>= groups
)
583 new_groups
= krealloc(nlk
->groups
, NLGRPSZ(groups
), GFP_ATOMIC
);
584 if (new_groups
== NULL
) {
588 memset((char*)new_groups
+ NLGRPSZ(nlk
->ngroups
), 0,
589 NLGRPSZ(groups
) - NLGRPSZ(nlk
->ngroups
));
591 nlk
->groups
= new_groups
;
592 nlk
->ngroups
= groups
;
594 netlink_table_ungrab();
598 static int netlink_bind(struct socket
*sock
, struct sockaddr
*addr
, int addr_len
)
600 struct sock
*sk
= sock
->sk
;
601 struct netlink_sock
*nlk
= nlk_sk(sk
);
602 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
605 if (nladdr
->nl_family
!= AF_NETLINK
)
608 /* Only superuser is allowed to listen multicasts */
609 if (nladdr
->nl_groups
) {
610 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
612 err
= netlink_realloc_groups(sk
);
618 if (nladdr
->nl_pid
!= nlk
->pid
)
621 err
= nladdr
->nl_pid
?
622 netlink_insert(sk
, nladdr
->nl_pid
) :
623 netlink_autobind(sock
);
628 if (!nladdr
->nl_groups
&& (nlk
->groups
== NULL
|| !(u32
)nlk
->groups
[0]))
631 netlink_table_grab();
632 netlink_update_subscriptions(sk
, nlk
->subscriptions
+
633 hweight32(nladdr
->nl_groups
) -
634 hweight32(nlk
->groups
[0]));
635 nlk
->groups
[0] = (nlk
->groups
[0] & ~0xffffffffUL
) | nladdr
->nl_groups
;
636 netlink_update_listeners(sk
);
637 netlink_table_ungrab();
642 static int netlink_connect(struct socket
*sock
, struct sockaddr
*addr
,
646 struct sock
*sk
= sock
->sk
;
647 struct netlink_sock
*nlk
= nlk_sk(sk
);
648 struct sockaddr_nl
*nladdr
=(struct sockaddr_nl
*)addr
;
650 if (addr
->sa_family
== AF_UNSPEC
) {
651 sk
->sk_state
= NETLINK_UNCONNECTED
;
656 if (addr
->sa_family
!= AF_NETLINK
)
659 /* Only superuser is allowed to send multicasts */
660 if (nladdr
->nl_groups
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
664 err
= netlink_autobind(sock
);
667 sk
->sk_state
= NETLINK_CONNECTED
;
668 nlk
->dst_pid
= nladdr
->nl_pid
;
669 nlk
->dst_group
= ffs(nladdr
->nl_groups
);
675 static int netlink_getname(struct socket
*sock
, struct sockaddr
*addr
, int *addr_len
, int peer
)
677 struct sock
*sk
= sock
->sk
;
678 struct netlink_sock
*nlk
= nlk_sk(sk
);
679 struct sockaddr_nl
*nladdr
=(struct sockaddr_nl
*)addr
;
681 nladdr
->nl_family
= AF_NETLINK
;
683 *addr_len
= sizeof(*nladdr
);
686 nladdr
->nl_pid
= nlk
->dst_pid
;
687 nladdr
->nl_groups
= netlink_group_mask(nlk
->dst_group
);
689 nladdr
->nl_pid
= nlk
->pid
;
690 nladdr
->nl_groups
= nlk
->groups
? nlk
->groups
[0] : 0;
695 static void netlink_overrun(struct sock
*sk
)
697 if (!test_and_set_bit(0, &nlk_sk(sk
)->state
)) {
698 sk
->sk_err
= ENOBUFS
;
699 sk
->sk_error_report(sk
);
703 static struct sock
*netlink_getsockbypid(struct sock
*ssk
, u32 pid
)
705 int protocol
= ssk
->sk_protocol
;
707 struct netlink_sock
*nlk
;
709 sock
= netlink_lookup(protocol
, pid
);
711 return ERR_PTR(-ECONNREFUSED
);
713 /* Don't bother queuing skb if kernel socket has no input function */
715 if ((nlk
->pid
== 0 && !nlk
->data_ready
) ||
716 (sock
->sk_state
== NETLINK_CONNECTED
&&
717 nlk
->dst_pid
!= nlk_sk(ssk
)->pid
)) {
719 return ERR_PTR(-ECONNREFUSED
);
724 struct sock
*netlink_getsockbyfilp(struct file
*filp
)
726 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
729 if (!S_ISSOCK(inode
->i_mode
))
730 return ERR_PTR(-ENOTSOCK
);
732 sock
= SOCKET_I(inode
)->sk
;
733 if (sock
->sk_family
!= AF_NETLINK
)
734 return ERR_PTR(-EINVAL
);
741 * Attach a skb to a netlink socket.
742 * The caller must hold a reference to the destination socket. On error, the
743 * reference is dropped. The skb is not send to the destination, just all
744 * all error checks are performed and memory in the queue is reserved.
746 * < 0: error. skb freed, reference to sock dropped.
748 * 1: repeat lookup - reference dropped while waiting for socket memory.
750 int netlink_attachskb(struct sock
*sk
, struct sk_buff
*skb
, int nonblock
,
751 long timeo
, struct sock
*ssk
)
753 struct netlink_sock
*nlk
;
757 if (atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
758 test_bit(0, &nlk
->state
)) {
759 DECLARE_WAITQUEUE(wait
, current
);
761 if (!ssk
|| nlk_sk(ssk
)->pid
== 0)
768 __set_current_state(TASK_INTERRUPTIBLE
);
769 add_wait_queue(&nlk
->wait
, &wait
);
771 if ((atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
772 test_bit(0, &nlk
->state
)) &&
773 !sock_flag(sk
, SOCK_DEAD
))
774 timeo
= schedule_timeout(timeo
);
776 __set_current_state(TASK_RUNNING
);
777 remove_wait_queue(&nlk
->wait
, &wait
);
780 if (signal_pending(current
)) {
782 return sock_intr_errno(timeo
);
786 skb_set_owner_r(skb
, sk
);
790 int netlink_sendskb(struct sock
*sk
, struct sk_buff
*skb
, int protocol
)
794 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
795 sk
->sk_data_ready(sk
, len
);
800 void netlink_detachskb(struct sock
*sk
, struct sk_buff
*skb
)
806 static inline struct sk_buff
*netlink_trim(struct sk_buff
*skb
,
813 delta
= skb
->end
- skb
->tail
;
814 if (delta
* 2 < skb
->truesize
)
817 if (skb_shared(skb
)) {
818 struct sk_buff
*nskb
= skb_clone(skb
, allocation
);
825 if (!pskb_expand_head(skb
, 0, -delta
, allocation
))
826 skb
->truesize
-= delta
;
831 int netlink_unicast(struct sock
*ssk
, struct sk_buff
*skb
, u32 pid
, int nonblock
)
837 skb
= netlink_trim(skb
, gfp_any());
839 timeo
= sock_sndtimeo(ssk
, nonblock
);
841 sk
= netlink_getsockbypid(ssk
, pid
);
846 err
= netlink_attachskb(sk
, skb
, nonblock
, timeo
, ssk
);
852 return netlink_sendskb(sk
, skb
, ssk
->sk_protocol
);
855 int netlink_has_listeners(struct sock
*sk
, unsigned int group
)
858 unsigned long *listeners
;
860 BUG_ON(!(nlk_sk(sk
)->flags
& NETLINK_KERNEL_SOCKET
));
863 listeners
= rcu_dereference(nl_table
[sk
->sk_protocol
].listeners
);
865 if (group
- 1 < nl_table
[sk
->sk_protocol
].groups
)
866 res
= test_bit(group
- 1, listeners
);
872 EXPORT_SYMBOL_GPL(netlink_has_listeners
);
874 static __inline__
int netlink_broadcast_deliver(struct sock
*sk
, struct sk_buff
*skb
)
876 struct netlink_sock
*nlk
= nlk_sk(sk
);
878 if (atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
&&
879 !test_bit(0, &nlk
->state
)) {
880 skb_set_owner_r(skb
, sk
);
881 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
882 sk
->sk_data_ready(sk
, skb
->len
);
883 return atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
;
888 struct netlink_broadcast_data
{
889 struct sock
*exclude_sk
;
896 struct sk_buff
*skb
, *skb2
;
899 static inline int do_one_broadcast(struct sock
*sk
,
900 struct netlink_broadcast_data
*p
)
902 struct netlink_sock
*nlk
= nlk_sk(sk
);
905 if (p
->exclude_sk
== sk
)
908 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
909 !test_bit(p
->group
- 1, nlk
->groups
))
918 if (p
->skb2
== NULL
) {
919 if (skb_shared(p
->skb
)) {
920 p
->skb2
= skb_clone(p
->skb
, p
->allocation
);
922 p
->skb2
= skb_get(p
->skb
);
924 * skb ownership may have been set when
925 * delivered to a previous socket.
930 if (p
->skb2
== NULL
) {
932 /* Clone failed. Notify ALL listeners. */
934 } else if ((val
= netlink_broadcast_deliver(sk
, p
->skb2
)) < 0) {
947 int netlink_broadcast(struct sock
*ssk
, struct sk_buff
*skb
, u32 pid
,
948 u32 group
, gfp_t allocation
)
950 struct netlink_broadcast_data info
;
951 struct hlist_node
*node
;
954 skb
= netlink_trim(skb
, allocation
);
956 info
.exclude_sk
= ssk
;
962 info
.allocation
= allocation
;
966 /* While we sleep in clone, do not allow to change socket list */
968 netlink_lock_table();
970 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
971 do_one_broadcast(sk
, &info
);
975 netlink_unlock_table();
978 kfree_skb(info
.skb2
);
980 if (info
.delivered
) {
981 if (info
.congested
&& (allocation
& __GFP_WAIT
))
990 struct netlink_set_err_data
{
991 struct sock
*exclude_sk
;
997 static inline int do_one_set_err(struct sock
*sk
,
998 struct netlink_set_err_data
*p
)
1000 struct netlink_sock
*nlk
= nlk_sk(sk
);
1002 if (sk
== p
->exclude_sk
)
1005 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
1006 !test_bit(p
->group
- 1, nlk
->groups
))
1009 sk
->sk_err
= p
->code
;
1010 sk
->sk_error_report(sk
);
1015 void netlink_set_err(struct sock
*ssk
, u32 pid
, u32 group
, int code
)
1017 struct netlink_set_err_data info
;
1018 struct hlist_node
*node
;
1021 info
.exclude_sk
= ssk
;
1026 read_lock(&nl_table_lock
);
1028 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1029 do_one_set_err(sk
, &info
);
1031 read_unlock(&nl_table_lock
);
1034 /* must be called with netlink table grabbed */
1035 static void netlink_update_socket_mc(struct netlink_sock
*nlk
,
1039 int old
, new = !!is_new
, subscriptions
;
1041 old
= test_bit(group
- 1, nlk
->groups
);
1042 subscriptions
= nlk
->subscriptions
- old
+ new;
1044 __set_bit(group
- 1, nlk
->groups
);
1046 __clear_bit(group
- 1, nlk
->groups
);
1047 netlink_update_subscriptions(&nlk
->sk
, subscriptions
);
1048 netlink_update_listeners(&nlk
->sk
);
1051 static int netlink_setsockopt(struct socket
*sock
, int level
, int optname
,
1052 char __user
*optval
, int optlen
)
1054 struct sock
*sk
= sock
->sk
;
1055 struct netlink_sock
*nlk
= nlk_sk(sk
);
1056 unsigned int val
= 0;
1059 if (level
!= SOL_NETLINK
)
1060 return -ENOPROTOOPT
;
1062 if (optlen
>= sizeof(int) &&
1063 get_user(val
, (unsigned int __user
*)optval
))
1067 case NETLINK_PKTINFO
:
1069 nlk
->flags
|= NETLINK_RECV_PKTINFO
;
1071 nlk
->flags
&= ~NETLINK_RECV_PKTINFO
;
1074 case NETLINK_ADD_MEMBERSHIP
:
1075 case NETLINK_DROP_MEMBERSHIP
: {
1076 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
1078 err
= netlink_realloc_groups(sk
);
1081 if (!val
|| val
- 1 >= nlk
->ngroups
)
1083 netlink_table_grab();
1084 netlink_update_socket_mc(nlk
, val
,
1085 optname
== NETLINK_ADD_MEMBERSHIP
);
1086 netlink_table_ungrab();
1096 static int netlink_getsockopt(struct socket
*sock
, int level
, int optname
,
1097 char __user
*optval
, int __user
*optlen
)
1099 struct sock
*sk
= sock
->sk
;
1100 struct netlink_sock
*nlk
= nlk_sk(sk
);
1103 if (level
!= SOL_NETLINK
)
1104 return -ENOPROTOOPT
;
1106 if (get_user(len
, optlen
))
1112 case NETLINK_PKTINFO
:
1113 if (len
< sizeof(int))
1116 val
= nlk
->flags
& NETLINK_RECV_PKTINFO
? 1 : 0;
1117 if (put_user(len
, optlen
) ||
1118 put_user(val
, optval
))
1128 static void netlink_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
1130 struct nl_pktinfo info
;
1132 info
.group
= NETLINK_CB(skb
).dst_group
;
1133 put_cmsg(msg
, SOL_NETLINK
, NETLINK_PKTINFO
, sizeof(info
), &info
);
1136 static inline void netlink_rcv_wake(struct sock
*sk
)
1138 struct netlink_sock
*nlk
= nlk_sk(sk
);
1140 if (skb_queue_empty(&sk
->sk_receive_queue
))
1141 clear_bit(0, &nlk
->state
);
1142 if (!test_bit(0, &nlk
->state
))
1143 wake_up_interruptible(&nlk
->wait
);
1146 static int netlink_sendmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1147 struct msghdr
*msg
, size_t len
)
1149 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1150 struct sock
*sk
= sock
->sk
;
1151 struct netlink_sock
*nlk
= nlk_sk(sk
);
1152 struct sockaddr_nl
*addr
=msg
->msg_name
;
1155 struct sk_buff
*skb
;
1157 struct scm_cookie scm
;
1159 if (msg
->msg_flags
&MSG_OOB
)
1162 if (NULL
== siocb
->scm
)
1164 err
= scm_send(sock
, msg
, siocb
->scm
);
1168 if (msg
->msg_namelen
) {
1169 if (addr
->nl_family
!= AF_NETLINK
)
1171 dst_pid
= addr
->nl_pid
;
1172 dst_group
= ffs(addr
->nl_groups
);
1173 if (dst_group
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
1176 dst_pid
= nlk
->dst_pid
;
1177 dst_group
= nlk
->dst_group
;
1181 err
= netlink_autobind(sock
);
1187 if (len
> sk
->sk_sndbuf
- 32)
1190 skb
= alloc_skb(len
, GFP_KERNEL
);
1194 NETLINK_CB(skb
).pid
= nlk
->pid
;
1195 NETLINK_CB(skb
).dst_group
= dst_group
;
1196 NETLINK_CB(skb
).loginuid
= audit_get_loginuid(current
->audit_context
);
1197 selinux_get_task_sid(current
, &(NETLINK_CB(skb
).sid
));
1198 memcpy(NETLINK_CREDS(skb
), &siocb
->scm
->creds
, sizeof(struct ucred
));
1200 /* What can I do? Netlink is asynchronous, so that
1201 we will have to save current capabilities to
1202 check them, when this message will be delivered
1203 to corresponding kernel module. --ANK (980802)
1207 if (memcpy_fromiovec(skb_put(skb
,len
), msg
->msg_iov
, len
)) {
1212 err
= security_netlink_send(sk
, skb
);
1219 atomic_inc(&skb
->users
);
1220 netlink_broadcast(sk
, skb
, dst_pid
, dst_group
, GFP_KERNEL
);
1222 err
= netlink_unicast(sk
, skb
, dst_pid
, msg
->msg_flags
&MSG_DONTWAIT
);
1228 static int netlink_recvmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1229 struct msghdr
*msg
, size_t len
,
1232 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1233 struct scm_cookie scm
;
1234 struct sock
*sk
= sock
->sk
;
1235 struct netlink_sock
*nlk
= nlk_sk(sk
);
1236 int noblock
= flags
&MSG_DONTWAIT
;
1238 struct sk_buff
*skb
;
1246 skb
= skb_recv_datagram(sk
,flags
,noblock
,&err
);
1250 msg
->msg_namelen
= 0;
1254 msg
->msg_flags
|= MSG_TRUNC
;
1258 skb_reset_transport_header(skb
);
1259 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1261 if (msg
->msg_name
) {
1262 struct sockaddr_nl
*addr
= (struct sockaddr_nl
*)msg
->msg_name
;
1263 addr
->nl_family
= AF_NETLINK
;
1265 addr
->nl_pid
= NETLINK_CB(skb
).pid
;
1266 addr
->nl_groups
= netlink_group_mask(NETLINK_CB(skb
).dst_group
);
1267 msg
->msg_namelen
= sizeof(*addr
);
1270 if (nlk
->flags
& NETLINK_RECV_PKTINFO
)
1271 netlink_cmsg_recv_pktinfo(msg
, skb
);
1273 if (NULL
== siocb
->scm
) {
1274 memset(&scm
, 0, sizeof(scm
));
1277 siocb
->scm
->creds
= *NETLINK_CREDS(skb
);
1278 if (flags
& MSG_TRUNC
)
1280 skb_free_datagram(sk
, skb
);
1282 if (nlk
->cb
&& atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
/ 2)
1285 scm_recv(sock
, msg
, siocb
->scm
, flags
);
1287 netlink_rcv_wake(sk
);
1288 return err
? : copied
;
1291 static void netlink_data_ready(struct sock
*sk
, int len
)
1293 struct netlink_sock
*nlk
= nlk_sk(sk
);
1295 if (nlk
->data_ready
)
1296 nlk
->data_ready(sk
, len
);
1297 netlink_rcv_wake(sk
);
1301 * We export these functions to other modules. They provide a
1302 * complete set of kernel non-blocking support for message
1307 netlink_kernel_create(int unit
, unsigned int groups
,
1308 void (*input
)(struct sock
*sk
, int len
),
1309 struct mutex
*cb_mutex
, struct module
*module
)
1311 struct socket
*sock
;
1313 struct netlink_sock
*nlk
;
1314 unsigned long *listeners
= NULL
;
1318 if (unit
<0 || unit
>=MAX_LINKS
)
1321 if (sock_create_lite(PF_NETLINK
, SOCK_DGRAM
, unit
, &sock
))
1324 if (__netlink_create(&init_net
, sock
, cb_mutex
, unit
) < 0)
1325 goto out_sock_release
;
1330 listeners
= kzalloc(NLGRPSZ(groups
), GFP_KERNEL
);
1332 goto out_sock_release
;
1335 sk
->sk_data_ready
= netlink_data_ready
;
1337 nlk_sk(sk
)->data_ready
= input
;
1339 if (netlink_insert(sk
, 0))
1340 goto out_sock_release
;
1343 nlk
->flags
|= NETLINK_KERNEL_SOCKET
;
1345 netlink_table_grab();
1346 nl_table
[unit
].groups
= groups
;
1347 nl_table
[unit
].listeners
= listeners
;
1348 nl_table
[unit
].cb_mutex
= cb_mutex
;
1349 nl_table
[unit
].module
= module
;
1350 nl_table
[unit
].registered
= 1;
1351 netlink_table_ungrab();
1362 * netlink_change_ngroups - change number of multicast groups
1364 * This changes the number of multicast groups that are available
1365 * on a certain netlink family. Note that it is not possible to
1366 * change the number of groups to below 32. Also note that it does
1367 * not implicitly call netlink_clear_multicast_users() when the
1368 * number of groups is reduced.
1370 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1371 * @groups: The new number of groups.
1373 int netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
1375 unsigned long *listeners
, *old
= NULL
;
1376 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
1382 netlink_table_grab();
1383 if (NLGRPSZ(tbl
->groups
) < NLGRPSZ(groups
)) {
1384 listeners
= kzalloc(NLGRPSZ(groups
), GFP_ATOMIC
);
1389 old
= tbl
->listeners
;
1390 memcpy(listeners
, old
, NLGRPSZ(tbl
->groups
));
1391 rcu_assign_pointer(tbl
->listeners
, listeners
);
1393 tbl
->groups
= groups
;
1396 netlink_table_ungrab();
1401 EXPORT_SYMBOL(netlink_change_ngroups
);
1404 * netlink_clear_multicast_users - kick off multicast listeners
1406 * This function removes all listeners from the given group.
1407 * @ksk: The kernel netlink socket, as returned by
1408 * netlink_kernel_create().
1409 * @group: The multicast group to clear.
1411 void netlink_clear_multicast_users(struct sock
*ksk
, unsigned int group
)
1414 struct hlist_node
*node
;
1415 struct netlink_table
*tbl
= &nl_table
[ksk
->sk_protocol
];
1417 netlink_table_grab();
1419 sk_for_each_bound(sk
, node
, &tbl
->mc_list
)
1420 netlink_update_socket_mc(nlk_sk(sk
), group
, 0);
1422 netlink_table_ungrab();
1424 EXPORT_SYMBOL(netlink_clear_multicast_users
);
1426 void netlink_set_nonroot(int protocol
, unsigned int flags
)
1428 if ((unsigned int)protocol
< MAX_LINKS
)
1429 nl_table
[protocol
].nl_nonroot
= flags
;
1432 static void netlink_destroy_callback(struct netlink_callback
*cb
)
1440 * It looks a bit ugly.
1441 * It would be better to create kernel thread.
1444 static int netlink_dump(struct sock
*sk
)
1446 struct netlink_sock
*nlk
= nlk_sk(sk
);
1447 struct netlink_callback
*cb
;
1448 struct sk_buff
*skb
;
1449 struct nlmsghdr
*nlh
;
1450 int len
, err
= -ENOBUFS
;
1452 skb
= sock_rmalloc(sk
, NLMSG_GOODSIZE
, 0, GFP_KERNEL
);
1456 mutex_lock(nlk
->cb_mutex
);
1464 len
= cb
->dump(skb
, cb
);
1467 mutex_unlock(nlk
->cb_mutex
);
1468 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1469 sk
->sk_data_ready(sk
, len
);
1473 nlh
= nlmsg_put_answer(skb
, cb
, NLMSG_DONE
, sizeof(len
), NLM_F_MULTI
);
1477 memcpy(nlmsg_data(nlh
), &len
, sizeof(len
));
1479 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1480 sk
->sk_data_ready(sk
, skb
->len
);
1485 mutex_unlock(nlk
->cb_mutex
);
1487 netlink_destroy_callback(cb
);
1491 mutex_unlock(nlk
->cb_mutex
);
1497 int netlink_dump_start(struct sock
*ssk
, struct sk_buff
*skb
,
1498 struct nlmsghdr
*nlh
,
1499 int (*dump
)(struct sk_buff
*skb
, struct netlink_callback
*),
1500 int (*done
)(struct netlink_callback
*))
1502 struct netlink_callback
*cb
;
1504 struct netlink_sock
*nlk
;
1506 cb
= kzalloc(sizeof(*cb
), GFP_KERNEL
);
1513 atomic_inc(&skb
->users
);
1516 sk
= netlink_lookup(ssk
->sk_protocol
, NETLINK_CB(skb
).pid
);
1518 netlink_destroy_callback(cb
);
1519 return -ECONNREFUSED
;
1522 /* A dump is in progress... */
1523 mutex_lock(nlk
->cb_mutex
);
1525 mutex_unlock(nlk
->cb_mutex
);
1526 netlink_destroy_callback(cb
);
1531 mutex_unlock(nlk
->cb_mutex
);
1536 /* We successfully started a dump, by returning -EINTR we
1537 * signal the queue mangement to interrupt processing of
1538 * any netlink messages so userspace gets a chance to read
1543 void netlink_ack(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, int err
)
1545 struct sk_buff
*skb
;
1546 struct nlmsghdr
*rep
;
1547 struct nlmsgerr
*errmsg
;
1548 size_t payload
= sizeof(*errmsg
);
1550 /* error messages get the original request appened */
1552 payload
+= nlmsg_len(nlh
);
1554 skb
= nlmsg_new(payload
, GFP_KERNEL
);
1558 sk
= netlink_lookup(in_skb
->sk
->sk_protocol
,
1559 NETLINK_CB(in_skb
).pid
);
1561 sk
->sk_err
= ENOBUFS
;
1562 sk
->sk_error_report(sk
);
1568 rep
= __nlmsg_put(skb
, NETLINK_CB(in_skb
).pid
, nlh
->nlmsg_seq
,
1569 NLMSG_ERROR
, sizeof(struct nlmsgerr
), 0);
1570 errmsg
= nlmsg_data(rep
);
1571 errmsg
->error
= err
;
1572 memcpy(&errmsg
->msg
, nlh
, err
? nlh
->nlmsg_len
: sizeof(*nlh
));
1573 netlink_unicast(in_skb
->sk
, skb
, NETLINK_CB(in_skb
).pid
, MSG_DONTWAIT
);
1576 static int netlink_rcv_skb(struct sk_buff
*skb
, int (*cb
)(struct sk_buff
*,
1579 struct nlmsghdr
*nlh
;
1582 while (skb
->len
>= nlmsg_total_size(0)) {
1583 nlh
= nlmsg_hdr(skb
);
1586 if (nlh
->nlmsg_len
< NLMSG_HDRLEN
|| skb
->len
< nlh
->nlmsg_len
)
1589 /* Only requests are handled by the kernel */
1590 if (!(nlh
->nlmsg_flags
& NLM_F_REQUEST
))
1593 /* Skip control messages */
1594 if (nlh
->nlmsg_type
< NLMSG_MIN_TYPE
)
1598 if (err
== -EINTR
) {
1599 /* Not an error, but we interrupt processing */
1600 netlink_queue_skip(nlh
, skb
);
1604 if (nlh
->nlmsg_flags
& NLM_F_ACK
|| err
)
1605 netlink_ack(skb
, nlh
, err
);
1607 netlink_queue_skip(nlh
, skb
);
1614 * nelink_run_queue - Process netlink receive queue.
1615 * @sk: Netlink socket containing the queue
1616 * @qlen: Place to store queue length upon entry
1617 * @cb: Callback function invoked for each netlink message found
1619 * Processes as much as there was in the queue upon entry and invokes
1620 * a callback function for each netlink message found. The callback
1621 * function may refuse a message by returning a negative error code
1622 * but setting the error pointer to 0 in which case this function
1623 * returns with a qlen != 0.
1625 * qlen must be initialized to 0 before the initial entry, afterwards
1626 * the function may be called repeatedly until qlen reaches 0.
1628 * The callback function may return -EINTR to signal that processing
1629 * of netlink messages shall be interrupted. In this case the message
1630 * currently being processed will NOT be requeued onto the receive
1633 void netlink_run_queue(struct sock
*sk
, unsigned int *qlen
,
1634 int (*cb
)(struct sk_buff
*, struct nlmsghdr
*))
1636 struct sk_buff
*skb
;
1638 if (!*qlen
|| *qlen
> skb_queue_len(&sk
->sk_receive_queue
))
1639 *qlen
= skb_queue_len(&sk
->sk_receive_queue
);
1641 for (; *qlen
; (*qlen
)--) {
1642 skb
= skb_dequeue(&sk
->sk_receive_queue
);
1643 if (netlink_rcv_skb(skb
, cb
)) {
1645 skb_queue_head(&sk
->sk_receive_queue
, skb
);
1658 * netlink_queue_skip - Skip netlink message while processing queue.
1659 * @nlh: Netlink message to be skipped
1660 * @skb: Socket buffer containing the netlink messages.
1662 * Pulls the given netlink message off the socket buffer so the next
1663 * call to netlink_queue_run() will not reconsider the message.
1665 static void netlink_queue_skip(struct nlmsghdr
*nlh
, struct sk_buff
*skb
)
1667 int msglen
= NLMSG_ALIGN(nlh
->nlmsg_len
);
1669 if (msglen
> skb
->len
)
1672 skb_pull(skb
, msglen
);
1676 * nlmsg_notify - send a notification netlink message
1677 * @sk: netlink socket to use
1678 * @skb: notification message
1679 * @pid: destination netlink pid for reports or 0
1680 * @group: destination multicast group or 0
1681 * @report: 1 to report back, 0 to disable
1682 * @flags: allocation flags
1684 int nlmsg_notify(struct sock
*sk
, struct sk_buff
*skb
, u32 pid
,
1685 unsigned int group
, int report
, gfp_t flags
)
1690 int exclude_pid
= 0;
1693 atomic_inc(&skb
->users
);
1697 /* errors reported via destination sk->sk_err */
1698 nlmsg_multicast(sk
, skb
, exclude_pid
, group
, flags
);
1702 err
= nlmsg_unicast(sk
, skb
, pid
);
1707 #ifdef CONFIG_PROC_FS
1708 struct nl_seq_iter
{
1713 static struct sock
*netlink_seq_socket_idx(struct seq_file
*seq
, loff_t pos
)
1715 struct nl_seq_iter
*iter
= seq
->private;
1718 struct hlist_node
*node
;
1721 for (i
=0; i
<MAX_LINKS
; i
++) {
1722 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1724 for (j
= 0; j
<= hash
->mask
; j
++) {
1725 sk_for_each(s
, node
, &hash
->table
[j
]) {
1738 static void *netlink_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1740 read_lock(&nl_table_lock
);
1741 return *pos
? netlink_seq_socket_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1744 static void *netlink_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1747 struct nl_seq_iter
*iter
;
1752 if (v
== SEQ_START_TOKEN
)
1753 return netlink_seq_socket_idx(seq
, 0);
1759 iter
= seq
->private;
1761 j
= iter
->hash_idx
+ 1;
1764 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1766 for (; j
<= hash
->mask
; j
++) {
1767 s
= sk_head(&hash
->table
[j
]);
1776 } while (++i
< MAX_LINKS
);
1781 static void netlink_seq_stop(struct seq_file
*seq
, void *v
)
1783 read_unlock(&nl_table_lock
);
1787 static int netlink_seq_show(struct seq_file
*seq
, void *v
)
1789 if (v
== SEQ_START_TOKEN
)
1791 "sk Eth Pid Groups "
1792 "Rmem Wmem Dump Locks\n");
1795 struct netlink_sock
*nlk
= nlk_sk(s
);
1797 seq_printf(seq
, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1801 nlk
->groups
? (u32
)nlk
->groups
[0] : 0,
1802 atomic_read(&s
->sk_rmem_alloc
),
1803 atomic_read(&s
->sk_wmem_alloc
),
1805 atomic_read(&s
->sk_refcnt
)
1812 static const struct seq_operations netlink_seq_ops
= {
1813 .start
= netlink_seq_start
,
1814 .next
= netlink_seq_next
,
1815 .stop
= netlink_seq_stop
,
1816 .show
= netlink_seq_show
,
1820 static int netlink_seq_open(struct inode
*inode
, struct file
*file
)
1822 struct seq_file
*seq
;
1823 struct nl_seq_iter
*iter
;
1826 iter
= kzalloc(sizeof(*iter
), GFP_KERNEL
);
1830 err
= seq_open(file
, &netlink_seq_ops
);
1836 seq
= file
->private_data
;
1837 seq
->private = iter
;
1841 static const struct file_operations netlink_seq_fops
= {
1842 .owner
= THIS_MODULE
,
1843 .open
= netlink_seq_open
,
1845 .llseek
= seq_lseek
,
1846 .release
= seq_release_private
,
1851 int netlink_register_notifier(struct notifier_block
*nb
)
1853 return atomic_notifier_chain_register(&netlink_chain
, nb
);
1856 int netlink_unregister_notifier(struct notifier_block
*nb
)
1858 return atomic_notifier_chain_unregister(&netlink_chain
, nb
);
1861 static const struct proto_ops netlink_ops
= {
1862 .family
= PF_NETLINK
,
1863 .owner
= THIS_MODULE
,
1864 .release
= netlink_release
,
1865 .bind
= netlink_bind
,
1866 .connect
= netlink_connect
,
1867 .socketpair
= sock_no_socketpair
,
1868 .accept
= sock_no_accept
,
1869 .getname
= netlink_getname
,
1870 .poll
= datagram_poll
,
1871 .ioctl
= sock_no_ioctl
,
1872 .listen
= sock_no_listen
,
1873 .shutdown
= sock_no_shutdown
,
1874 .setsockopt
= netlink_setsockopt
,
1875 .getsockopt
= netlink_getsockopt
,
1876 .sendmsg
= netlink_sendmsg
,
1877 .recvmsg
= netlink_recvmsg
,
1878 .mmap
= sock_no_mmap
,
1879 .sendpage
= sock_no_sendpage
,
1882 static struct net_proto_family netlink_family_ops
= {
1883 .family
= PF_NETLINK
,
1884 .create
= netlink_create
,
1885 .owner
= THIS_MODULE
, /* for consistency 8) */
1888 static int __init
netlink_proto_init(void)
1890 struct sk_buff
*dummy_skb
;
1894 int err
= proto_register(&netlink_proto
, 0);
1899 BUILD_BUG_ON(sizeof(struct netlink_skb_parms
) > sizeof(dummy_skb
->cb
));
1901 nl_table
= kcalloc(MAX_LINKS
, sizeof(*nl_table
), GFP_KERNEL
);
1905 if (num_physpages
>= (128 * 1024))
1906 max
= num_physpages
>> (21 - PAGE_SHIFT
);
1908 max
= num_physpages
>> (23 - PAGE_SHIFT
);
1910 order
= get_bitmask_order(max
) - 1 + PAGE_SHIFT
;
1911 max
= (1UL << order
) / sizeof(struct hlist_head
);
1912 order
= get_bitmask_order(max
> UINT_MAX
? UINT_MAX
: max
) - 1;
1914 for (i
= 0; i
< MAX_LINKS
; i
++) {
1915 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1917 hash
->table
= nl_pid_hash_alloc(1 * sizeof(*hash
->table
));
1920 nl_pid_hash_free(nl_table
[i
].hash
.table
,
1921 1 * sizeof(*hash
->table
));
1925 memset(hash
->table
, 0, 1 * sizeof(*hash
->table
));
1926 hash
->max_shift
= order
;
1929 hash
->rehash_time
= jiffies
;
1932 sock_register(&netlink_family_ops
);
1933 #ifdef CONFIG_PROC_FS
1934 proc_net_fops_create(&init_net
, "netlink", 0, &netlink_seq_fops
);
1936 /* The netlink device handler may be needed early. */
1941 panic("netlink_init: Cannot allocate nl_table\n");
1944 core_initcall(netlink_proto_init
);
1946 EXPORT_SYMBOL(netlink_ack
);
1947 EXPORT_SYMBOL(netlink_run_queue
);
1948 EXPORT_SYMBOL(netlink_broadcast
);
1949 EXPORT_SYMBOL(netlink_dump_start
);
1950 EXPORT_SYMBOL(netlink_kernel_create
);
1951 EXPORT_SYMBOL(netlink_register_notifier
);
1952 EXPORT_SYMBOL(netlink_set_nonroot
);
1953 EXPORT_SYMBOL(netlink_unicast
);
1954 EXPORT_SYMBOL(netlink_unregister_notifier
);
1955 EXPORT_SYMBOL(nlmsg_notify
);