2 * NETLINK Kernel-user communication protocol.
4 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
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/mutex.h>
59 #include <net/net_namespace.h>
62 #include <net/netlink.h>
64 #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
65 #define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
68 /* struct sock has to be the first member of netlink_sock */
76 unsigned long *groups
;
78 wait_queue_head_t wait
;
79 struct netlink_callback
*cb
;
80 struct mutex
*cb_mutex
;
81 struct mutex cb_def_mutex
;
82 void (*netlink_rcv
)(struct sk_buff
*skb
);
83 struct module
*module
;
86 #define NETLINK_KERNEL_SOCKET 0x1
87 #define NETLINK_RECV_PKTINFO 0x2
89 static inline struct netlink_sock
*nlk_sk(struct sock
*sk
)
91 return container_of(sk
, struct netlink_sock
, sk
);
94 static inline int netlink_is_kernel(struct sock
*sk
)
96 return nlk_sk(sk
)->flags
& NETLINK_KERNEL_SOCKET
;
100 struct hlist_head
*table
;
101 unsigned long rehash_time
;
106 unsigned int entries
;
107 unsigned int max_shift
;
112 struct netlink_table
{
113 struct nl_pid_hash hash
;
114 struct hlist_head mc_list
;
115 unsigned long *listeners
;
116 unsigned int nl_nonroot
;
118 struct mutex
*cb_mutex
;
119 struct module
*module
;
123 static struct netlink_table
*nl_table
;
125 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait
);
127 static int netlink_dump(struct sock
*sk
);
128 static void netlink_destroy_callback(struct netlink_callback
*cb
);
130 static DEFINE_RWLOCK(nl_table_lock
);
131 static atomic_t nl_table_users
= ATOMIC_INIT(0);
133 static ATOMIC_NOTIFIER_HEAD(netlink_chain
);
135 static u32
netlink_group_mask(u32 group
)
137 return group
? 1 << (group
- 1) : 0;
140 static struct hlist_head
*nl_pid_hashfn(struct nl_pid_hash
*hash
, u32 pid
)
142 return &hash
->table
[jhash_1word(pid
, hash
->rnd
) & hash
->mask
];
145 static void netlink_sock_destruct(struct sock
*sk
)
147 struct netlink_sock
*nlk
= nlk_sk(sk
);
151 nlk
->cb
->done(nlk
->cb
);
152 netlink_destroy_callback(nlk
->cb
);
155 skb_queue_purge(&sk
->sk_receive_queue
);
157 if (!sock_flag(sk
, SOCK_DEAD
)) {
158 printk(KERN_ERR
"Freeing alive netlink socket %p\n", sk
);
162 WARN_ON(atomic_read(&sk
->sk_rmem_alloc
));
163 WARN_ON(atomic_read(&sk
->sk_wmem_alloc
));
164 WARN_ON(nlk_sk(sk
)->groups
);
167 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
168 * SMP. Look, when several writers sleep and reader wakes them up, all but one
169 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
170 * this, _but_ remember, it adds useless work on UP machines.
173 static void netlink_table_grab(void)
174 __acquires(nl_table_lock
)
176 write_lock_irq(&nl_table_lock
);
178 if (atomic_read(&nl_table_users
)) {
179 DECLARE_WAITQUEUE(wait
, current
);
181 add_wait_queue_exclusive(&nl_table_wait
, &wait
);
183 set_current_state(TASK_UNINTERRUPTIBLE
);
184 if (atomic_read(&nl_table_users
) == 0)
186 write_unlock_irq(&nl_table_lock
);
188 write_lock_irq(&nl_table_lock
);
191 __set_current_state(TASK_RUNNING
);
192 remove_wait_queue(&nl_table_wait
, &wait
);
196 static void netlink_table_ungrab(void)
197 __releases(nl_table_lock
)
199 write_unlock_irq(&nl_table_lock
);
200 wake_up(&nl_table_wait
);
204 netlink_lock_table(void)
206 /* read_lock() synchronizes us to netlink_table_grab */
208 read_lock(&nl_table_lock
);
209 atomic_inc(&nl_table_users
);
210 read_unlock(&nl_table_lock
);
214 netlink_unlock_table(void)
216 if (atomic_dec_and_test(&nl_table_users
))
217 wake_up(&nl_table_wait
);
220 static inline struct sock
*netlink_lookup(struct net
*net
, int protocol
,
223 struct nl_pid_hash
*hash
= &nl_table
[protocol
].hash
;
224 struct hlist_head
*head
;
226 struct hlist_node
*node
;
228 read_lock(&nl_table_lock
);
229 head
= nl_pid_hashfn(hash
, pid
);
230 sk_for_each(sk
, node
, head
) {
231 if (net_eq(sock_net(sk
), net
) && (nlk_sk(sk
)->pid
== pid
)) {
238 read_unlock(&nl_table_lock
);
242 static inline struct hlist_head
*nl_pid_hash_zalloc(size_t size
)
244 if (size
<= PAGE_SIZE
)
245 return kzalloc(size
, GFP_ATOMIC
);
247 return (struct hlist_head
*)
248 __get_free_pages(GFP_ATOMIC
| __GFP_ZERO
,
252 static inline void nl_pid_hash_free(struct hlist_head
*table
, size_t size
)
254 if (size
<= PAGE_SIZE
)
257 free_pages((unsigned long)table
, get_order(size
));
260 static int nl_pid_hash_rehash(struct nl_pid_hash
*hash
, int grow
)
262 unsigned int omask
, mask
, shift
;
264 struct hlist_head
*otable
, *table
;
267 omask
= mask
= hash
->mask
;
268 osize
= size
= (mask
+ 1) * sizeof(*table
);
272 if (++shift
> hash
->max_shift
)
278 table
= nl_pid_hash_zalloc(size
);
282 otable
= hash
->table
;
286 get_random_bytes(&hash
->rnd
, sizeof(hash
->rnd
));
288 for (i
= 0; i
<= omask
; i
++) {
290 struct hlist_node
*node
, *tmp
;
292 sk_for_each_safe(sk
, node
, tmp
, &otable
[i
])
293 __sk_add_node(sk
, nl_pid_hashfn(hash
, nlk_sk(sk
)->pid
));
296 nl_pid_hash_free(otable
, osize
);
297 hash
->rehash_time
= jiffies
+ 10 * 60 * HZ
;
301 static inline int nl_pid_hash_dilute(struct nl_pid_hash
*hash
, int len
)
303 int avg
= hash
->entries
>> hash
->shift
;
305 if (unlikely(avg
> 1) && nl_pid_hash_rehash(hash
, 1))
308 if (unlikely(len
> avg
) && time_after(jiffies
, hash
->rehash_time
)) {
309 nl_pid_hash_rehash(hash
, 0);
316 static const struct proto_ops netlink_ops
;
319 netlink_update_listeners(struct sock
*sk
)
321 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
322 struct hlist_node
*node
;
326 for (i
= 0; i
< NLGRPLONGS(tbl
->groups
); i
++) {
328 sk_for_each_bound(sk
, node
, &tbl
->mc_list
) {
329 if (i
< NLGRPLONGS(nlk_sk(sk
)->ngroups
))
330 mask
|= nlk_sk(sk
)->groups
[i
];
332 tbl
->listeners
[i
] = mask
;
334 /* this function is only called with the netlink table "grabbed", which
335 * makes sure updates are visible before bind or setsockopt return. */
338 static int netlink_insert(struct sock
*sk
, struct net
*net
, u32 pid
)
340 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
341 struct hlist_head
*head
;
342 int err
= -EADDRINUSE
;
344 struct hlist_node
*node
;
347 netlink_table_grab();
348 head
= nl_pid_hashfn(hash
, pid
);
350 sk_for_each(osk
, node
, head
) {
351 if (net_eq(sock_net(osk
), net
) && (nlk_sk(osk
)->pid
== pid
))
363 if (BITS_PER_LONG
> 32 && unlikely(hash
->entries
>= UINT_MAX
))
366 if (len
&& nl_pid_hash_dilute(hash
, len
))
367 head
= nl_pid_hashfn(hash
, pid
);
369 nlk_sk(sk
)->pid
= pid
;
370 sk_add_node(sk
, head
);
374 netlink_table_ungrab();
378 static void netlink_remove(struct sock
*sk
)
380 netlink_table_grab();
381 if (sk_del_node_init(sk
))
382 nl_table
[sk
->sk_protocol
].hash
.entries
--;
383 if (nlk_sk(sk
)->subscriptions
)
384 __sk_del_bind_node(sk
);
385 netlink_table_ungrab();
388 static struct proto netlink_proto
= {
390 .owner
= THIS_MODULE
,
391 .obj_size
= sizeof(struct netlink_sock
),
394 static int __netlink_create(struct net
*net
, struct socket
*sock
,
395 struct mutex
*cb_mutex
, int protocol
)
398 struct netlink_sock
*nlk
;
400 sock
->ops
= &netlink_ops
;
402 sk
= sk_alloc(net
, PF_NETLINK
, GFP_KERNEL
, &netlink_proto
);
406 sock_init_data(sock
, sk
);
410 nlk
->cb_mutex
= cb_mutex
;
412 nlk
->cb_mutex
= &nlk
->cb_def_mutex
;
413 mutex_init(nlk
->cb_mutex
);
415 init_waitqueue_head(&nlk
->wait
);
417 sk
->sk_destruct
= netlink_sock_destruct
;
418 sk
->sk_protocol
= protocol
;
422 static int netlink_create(struct net
*net
, struct socket
*sock
, int protocol
)
424 struct module
*module
= NULL
;
425 struct mutex
*cb_mutex
;
426 struct netlink_sock
*nlk
;
429 sock
->state
= SS_UNCONNECTED
;
431 if (sock
->type
!= SOCK_RAW
&& sock
->type
!= SOCK_DGRAM
)
432 return -ESOCKTNOSUPPORT
;
434 if (protocol
< 0 || protocol
>= MAX_LINKS
)
435 return -EPROTONOSUPPORT
;
437 netlink_lock_table();
438 #ifdef CONFIG_MODULES
439 if (!nl_table
[protocol
].registered
) {
440 netlink_unlock_table();
441 request_module("net-pf-%d-proto-%d", PF_NETLINK
, protocol
);
442 netlink_lock_table();
445 if (nl_table
[protocol
].registered
&&
446 try_module_get(nl_table
[protocol
].module
))
447 module
= nl_table
[protocol
].module
;
448 cb_mutex
= nl_table
[protocol
].cb_mutex
;
449 netlink_unlock_table();
451 err
= __netlink_create(net
, sock
, cb_mutex
, protocol
);
455 sock_prot_inuse_add(net
, &netlink_proto
, 1);
456 nlk
= nlk_sk(sock
->sk
);
457 nlk
->module
= module
;
466 static int netlink_release(struct socket
*sock
)
468 struct sock
*sk
= sock
->sk
;
469 struct netlink_sock
*nlk
;
479 * OK. Socket is unlinked, any packets that arrive now
484 wake_up_interruptible_all(&nlk
->wait
);
486 skb_queue_purge(&sk
->sk_write_queue
);
488 if (nlk
->pid
&& !nlk
->subscriptions
) {
489 struct netlink_notify n
= {
491 .protocol
= sk
->sk_protocol
,
494 atomic_notifier_call_chain(&netlink_chain
,
495 NETLINK_URELEASE
, &n
);
498 module_put(nlk
->module
);
500 netlink_table_grab();
501 if (netlink_is_kernel(sk
)) {
502 BUG_ON(nl_table
[sk
->sk_protocol
].registered
== 0);
503 if (--nl_table
[sk
->sk_protocol
].registered
== 0) {
504 kfree(nl_table
[sk
->sk_protocol
].listeners
);
505 nl_table
[sk
->sk_protocol
].module
= NULL
;
506 nl_table
[sk
->sk_protocol
].registered
= 0;
508 } else if (nlk
->subscriptions
)
509 netlink_update_listeners(sk
);
510 netlink_table_ungrab();
515 sock_prot_inuse_add(sock_net(sk
), &netlink_proto
, -1);
520 static int netlink_autobind(struct socket
*sock
)
522 struct sock
*sk
= sock
->sk
;
523 struct net
*net
= sock_net(sk
);
524 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
525 struct hlist_head
*head
;
527 struct hlist_node
*node
;
528 s32 pid
= current
->tgid
;
530 static s32 rover
= -4097;
534 netlink_table_grab();
535 head
= nl_pid_hashfn(hash
, pid
);
536 sk_for_each(osk
, node
, head
) {
537 if (!net_eq(sock_net(osk
), net
))
539 if (nlk_sk(osk
)->pid
== pid
) {
540 /* Bind collision, search negative pid values. */
544 netlink_table_ungrab();
548 netlink_table_ungrab();
550 err
= netlink_insert(sk
, net
, pid
);
551 if (err
== -EADDRINUSE
)
554 /* If 2 threads race to autobind, that is fine. */
561 static inline int netlink_capable(struct socket
*sock
, unsigned int flag
)
563 return (nl_table
[sock
->sk
->sk_protocol
].nl_nonroot
& flag
) ||
564 capable(CAP_NET_ADMIN
);
568 netlink_update_subscriptions(struct sock
*sk
, unsigned int subscriptions
)
570 struct netlink_sock
*nlk
= nlk_sk(sk
);
572 if (nlk
->subscriptions
&& !subscriptions
)
573 __sk_del_bind_node(sk
);
574 else if (!nlk
->subscriptions
&& subscriptions
)
575 sk_add_bind_node(sk
, &nl_table
[sk
->sk_protocol
].mc_list
);
576 nlk
->subscriptions
= subscriptions
;
579 static int netlink_realloc_groups(struct sock
*sk
)
581 struct netlink_sock
*nlk
= nlk_sk(sk
);
583 unsigned long *new_groups
;
586 netlink_table_grab();
588 groups
= nl_table
[sk
->sk_protocol
].groups
;
589 if (!nl_table
[sk
->sk_protocol
].registered
) {
594 if (nlk
->ngroups
>= groups
)
597 new_groups
= krealloc(nlk
->groups
, NLGRPSZ(groups
), GFP_ATOMIC
);
598 if (new_groups
== NULL
) {
602 memset((char *)new_groups
+ NLGRPSZ(nlk
->ngroups
), 0,
603 NLGRPSZ(groups
) - NLGRPSZ(nlk
->ngroups
));
605 nlk
->groups
= new_groups
;
606 nlk
->ngroups
= groups
;
608 netlink_table_ungrab();
612 static int netlink_bind(struct socket
*sock
, struct sockaddr
*addr
,
615 struct sock
*sk
= sock
->sk
;
616 struct net
*net
= sock_net(sk
);
617 struct netlink_sock
*nlk
= nlk_sk(sk
);
618 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
621 if (nladdr
->nl_family
!= AF_NETLINK
)
624 /* Only superuser is allowed to listen multicasts */
625 if (nladdr
->nl_groups
) {
626 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
628 err
= netlink_realloc_groups(sk
);
634 if (nladdr
->nl_pid
!= nlk
->pid
)
637 err
= nladdr
->nl_pid
?
638 netlink_insert(sk
, net
, nladdr
->nl_pid
) :
639 netlink_autobind(sock
);
644 if (!nladdr
->nl_groups
&& (nlk
->groups
== NULL
|| !(u32
)nlk
->groups
[0]))
647 netlink_table_grab();
648 netlink_update_subscriptions(sk
, nlk
->subscriptions
+
649 hweight32(nladdr
->nl_groups
) -
650 hweight32(nlk
->groups
[0]));
651 nlk
->groups
[0] = (nlk
->groups
[0] & ~0xffffffffUL
) | nladdr
->nl_groups
;
652 netlink_update_listeners(sk
);
653 netlink_table_ungrab();
658 static int netlink_connect(struct socket
*sock
, struct sockaddr
*addr
,
662 struct sock
*sk
= sock
->sk
;
663 struct netlink_sock
*nlk
= nlk_sk(sk
);
664 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
666 if (addr
->sa_family
== AF_UNSPEC
) {
667 sk
->sk_state
= NETLINK_UNCONNECTED
;
672 if (addr
->sa_family
!= AF_NETLINK
)
675 /* Only superuser is allowed to send multicasts */
676 if (nladdr
->nl_groups
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
680 err
= netlink_autobind(sock
);
683 sk
->sk_state
= NETLINK_CONNECTED
;
684 nlk
->dst_pid
= nladdr
->nl_pid
;
685 nlk
->dst_group
= ffs(nladdr
->nl_groups
);
691 static int netlink_getname(struct socket
*sock
, struct sockaddr
*addr
,
692 int *addr_len
, int peer
)
694 struct sock
*sk
= sock
->sk
;
695 struct netlink_sock
*nlk
= nlk_sk(sk
);
696 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
698 nladdr
->nl_family
= AF_NETLINK
;
700 *addr_len
= sizeof(*nladdr
);
703 nladdr
->nl_pid
= nlk
->dst_pid
;
704 nladdr
->nl_groups
= netlink_group_mask(nlk
->dst_group
);
706 nladdr
->nl_pid
= nlk
->pid
;
707 nladdr
->nl_groups
= nlk
->groups
? nlk
->groups
[0] : 0;
712 static void netlink_overrun(struct sock
*sk
)
714 if (!test_and_set_bit(0, &nlk_sk(sk
)->state
)) {
715 sk
->sk_err
= ENOBUFS
;
716 sk
->sk_error_report(sk
);
720 static struct sock
*netlink_getsockbypid(struct sock
*ssk
, u32 pid
)
723 struct netlink_sock
*nlk
;
725 sock
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, pid
);
727 return ERR_PTR(-ECONNREFUSED
);
729 /* Don't bother queuing skb if kernel socket has no input function */
731 if (sock
->sk_state
== NETLINK_CONNECTED
&&
732 nlk
->dst_pid
!= nlk_sk(ssk
)->pid
) {
734 return ERR_PTR(-ECONNREFUSED
);
739 struct sock
*netlink_getsockbyfilp(struct file
*filp
)
741 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
744 if (!S_ISSOCK(inode
->i_mode
))
745 return ERR_PTR(-ENOTSOCK
);
747 sock
= SOCKET_I(inode
)->sk
;
748 if (sock
->sk_family
!= AF_NETLINK
)
749 return ERR_PTR(-EINVAL
);
756 * Attach a skb to a netlink socket.
757 * The caller must hold a reference to the destination socket. On error, the
758 * reference is dropped. The skb is not send to the destination, just all
759 * all error checks are performed and memory in the queue is reserved.
761 * < 0: error. skb freed, reference to sock dropped.
763 * 1: repeat lookup - reference dropped while waiting for socket memory.
765 int netlink_attachskb(struct sock
*sk
, struct sk_buff
*skb
,
766 long *timeo
, struct sock
*ssk
)
768 struct netlink_sock
*nlk
;
772 if (atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
773 test_bit(0, &nlk
->state
)) {
774 DECLARE_WAITQUEUE(wait
, current
);
776 if (!ssk
|| netlink_is_kernel(ssk
))
783 __set_current_state(TASK_INTERRUPTIBLE
);
784 add_wait_queue(&nlk
->wait
, &wait
);
786 if ((atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
787 test_bit(0, &nlk
->state
)) &&
788 !sock_flag(sk
, SOCK_DEAD
))
789 *timeo
= schedule_timeout(*timeo
);
791 __set_current_state(TASK_RUNNING
);
792 remove_wait_queue(&nlk
->wait
, &wait
);
795 if (signal_pending(current
)) {
797 return sock_intr_errno(*timeo
);
801 skb_set_owner_r(skb
, sk
);
805 int netlink_sendskb(struct sock
*sk
, struct sk_buff
*skb
)
809 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
810 sk
->sk_data_ready(sk
, len
);
815 void netlink_detachskb(struct sock
*sk
, struct sk_buff
*skb
)
821 static inline struct sk_buff
*netlink_trim(struct sk_buff
*skb
,
828 delta
= skb
->end
- skb
->tail
;
829 if (delta
* 2 < skb
->truesize
)
832 if (skb_shared(skb
)) {
833 struct sk_buff
*nskb
= skb_clone(skb
, allocation
);
840 if (!pskb_expand_head(skb
, 0, -delta
, allocation
))
841 skb
->truesize
-= delta
;
846 static inline void netlink_rcv_wake(struct sock
*sk
)
848 struct netlink_sock
*nlk
= nlk_sk(sk
);
850 if (skb_queue_empty(&sk
->sk_receive_queue
))
851 clear_bit(0, &nlk
->state
);
852 if (!test_bit(0, &nlk
->state
))
853 wake_up_interruptible(&nlk
->wait
);
856 static inline int netlink_unicast_kernel(struct sock
*sk
, struct sk_buff
*skb
)
859 struct netlink_sock
*nlk
= nlk_sk(sk
);
862 if (nlk
->netlink_rcv
!= NULL
) {
864 skb_set_owner_r(skb
, sk
);
865 nlk
->netlink_rcv(skb
);
872 int netlink_unicast(struct sock
*ssk
, struct sk_buff
*skb
,
873 u32 pid
, int nonblock
)
879 skb
= netlink_trim(skb
, gfp_any());
881 timeo
= sock_sndtimeo(ssk
, nonblock
);
883 sk
= netlink_getsockbypid(ssk
, pid
);
888 if (netlink_is_kernel(sk
))
889 return netlink_unicast_kernel(sk
, skb
);
891 if (sk_filter(sk
, skb
)) {
898 err
= netlink_attachskb(sk
, skb
, &timeo
, ssk
);
904 return netlink_sendskb(sk
, skb
);
906 EXPORT_SYMBOL(netlink_unicast
);
908 int netlink_has_listeners(struct sock
*sk
, unsigned int group
)
911 unsigned long *listeners
;
913 BUG_ON(!netlink_is_kernel(sk
));
916 listeners
= rcu_dereference(nl_table
[sk
->sk_protocol
].listeners
);
918 if (group
- 1 < nl_table
[sk
->sk_protocol
].groups
)
919 res
= test_bit(group
- 1, listeners
);
925 EXPORT_SYMBOL_GPL(netlink_has_listeners
);
927 static inline int netlink_broadcast_deliver(struct sock
*sk
,
930 struct netlink_sock
*nlk
= nlk_sk(sk
);
932 if (atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
&&
933 !test_bit(0, &nlk
->state
)) {
934 skb_set_owner_r(skb
, sk
);
935 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
936 sk
->sk_data_ready(sk
, skb
->len
);
937 return atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
;
942 struct netlink_broadcast_data
{
943 struct sock
*exclude_sk
;
951 struct sk_buff
*skb
, *skb2
;
954 static inline int do_one_broadcast(struct sock
*sk
,
955 struct netlink_broadcast_data
*p
)
957 struct netlink_sock
*nlk
= nlk_sk(sk
);
960 if (p
->exclude_sk
== sk
)
963 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
964 !test_bit(p
->group
- 1, nlk
->groups
))
967 if (!net_eq(sock_net(sk
), p
->net
))
976 if (p
->skb2
== NULL
) {
977 if (skb_shared(p
->skb
)) {
978 p
->skb2
= skb_clone(p
->skb
, p
->allocation
);
980 p
->skb2
= skb_get(p
->skb
);
982 * skb ownership may have been set when
983 * delivered to a previous socket.
988 if (p
->skb2
== NULL
) {
990 /* Clone failed. Notify ALL listeners. */
992 } else if (sk_filter(sk
, p
->skb2
)) {
995 } else if ((val
= netlink_broadcast_deliver(sk
, p
->skb2
)) < 0) {
1008 int netlink_broadcast(struct sock
*ssk
, struct sk_buff
*skb
, u32 pid
,
1009 u32 group
, gfp_t allocation
)
1011 struct net
*net
= sock_net(ssk
);
1012 struct netlink_broadcast_data info
;
1013 struct hlist_node
*node
;
1016 skb
= netlink_trim(skb
, allocation
);
1018 info
.exclude_sk
= ssk
;
1025 info
.allocation
= allocation
;
1029 /* While we sleep in clone, do not allow to change socket list */
1031 netlink_lock_table();
1033 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1034 do_one_broadcast(sk
, &info
);
1038 netlink_unlock_table();
1041 kfree_skb(info
.skb2
);
1043 if (info
.delivered
) {
1044 if (info
.congested
&& (allocation
& __GFP_WAIT
))
1052 EXPORT_SYMBOL(netlink_broadcast
);
1054 struct netlink_set_err_data
{
1055 struct sock
*exclude_sk
;
1061 static inline int do_one_set_err(struct sock
*sk
,
1062 struct netlink_set_err_data
*p
)
1064 struct netlink_sock
*nlk
= nlk_sk(sk
);
1066 if (sk
== p
->exclude_sk
)
1069 if (sock_net(sk
) != sock_net(p
->exclude_sk
))
1072 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
1073 !test_bit(p
->group
- 1, nlk
->groups
))
1076 sk
->sk_err
= p
->code
;
1077 sk
->sk_error_report(sk
);
1082 void netlink_set_err(struct sock
*ssk
, u32 pid
, u32 group
, int code
)
1084 struct netlink_set_err_data info
;
1085 struct hlist_node
*node
;
1088 info
.exclude_sk
= ssk
;
1093 read_lock(&nl_table_lock
);
1095 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1096 do_one_set_err(sk
, &info
);
1098 read_unlock(&nl_table_lock
);
1101 /* must be called with netlink table grabbed */
1102 static void netlink_update_socket_mc(struct netlink_sock
*nlk
,
1106 int old
, new = !!is_new
, subscriptions
;
1108 old
= test_bit(group
- 1, nlk
->groups
);
1109 subscriptions
= nlk
->subscriptions
- old
+ new;
1111 __set_bit(group
- 1, nlk
->groups
);
1113 __clear_bit(group
- 1, nlk
->groups
);
1114 netlink_update_subscriptions(&nlk
->sk
, subscriptions
);
1115 netlink_update_listeners(&nlk
->sk
);
1118 static int netlink_setsockopt(struct socket
*sock
, int level
, int optname
,
1119 char __user
*optval
, int optlen
)
1121 struct sock
*sk
= sock
->sk
;
1122 struct netlink_sock
*nlk
= nlk_sk(sk
);
1123 unsigned int val
= 0;
1126 if (level
!= SOL_NETLINK
)
1127 return -ENOPROTOOPT
;
1129 if (optlen
>= sizeof(int) &&
1130 get_user(val
, (unsigned int __user
*)optval
))
1134 case NETLINK_PKTINFO
:
1136 nlk
->flags
|= NETLINK_RECV_PKTINFO
;
1138 nlk
->flags
&= ~NETLINK_RECV_PKTINFO
;
1141 case NETLINK_ADD_MEMBERSHIP
:
1142 case NETLINK_DROP_MEMBERSHIP
: {
1143 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
1145 err
= netlink_realloc_groups(sk
);
1148 if (!val
|| val
- 1 >= nlk
->ngroups
)
1150 netlink_table_grab();
1151 netlink_update_socket_mc(nlk
, val
,
1152 optname
== NETLINK_ADD_MEMBERSHIP
);
1153 netlink_table_ungrab();
1163 static int netlink_getsockopt(struct socket
*sock
, int level
, int optname
,
1164 char __user
*optval
, int __user
*optlen
)
1166 struct sock
*sk
= sock
->sk
;
1167 struct netlink_sock
*nlk
= nlk_sk(sk
);
1170 if (level
!= SOL_NETLINK
)
1171 return -ENOPROTOOPT
;
1173 if (get_user(len
, optlen
))
1179 case NETLINK_PKTINFO
:
1180 if (len
< sizeof(int))
1183 val
= nlk
->flags
& NETLINK_RECV_PKTINFO
? 1 : 0;
1184 if (put_user(len
, optlen
) ||
1185 put_user(val
, optval
))
1195 static void netlink_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
1197 struct nl_pktinfo info
;
1199 info
.group
= NETLINK_CB(skb
).dst_group
;
1200 put_cmsg(msg
, SOL_NETLINK
, NETLINK_PKTINFO
, sizeof(info
), &info
);
1203 static int netlink_sendmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1204 struct msghdr
*msg
, size_t len
)
1206 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1207 struct sock
*sk
= sock
->sk
;
1208 struct netlink_sock
*nlk
= nlk_sk(sk
);
1209 struct sockaddr_nl
*addr
= msg
->msg_name
;
1212 struct sk_buff
*skb
;
1214 struct scm_cookie scm
;
1216 if (msg
->msg_flags
&MSG_OOB
)
1219 if (NULL
== siocb
->scm
)
1221 err
= scm_send(sock
, msg
, siocb
->scm
);
1225 if (msg
->msg_namelen
) {
1226 if (addr
->nl_family
!= AF_NETLINK
)
1228 dst_pid
= addr
->nl_pid
;
1229 dst_group
= ffs(addr
->nl_groups
);
1230 if (dst_group
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
1233 dst_pid
= nlk
->dst_pid
;
1234 dst_group
= nlk
->dst_group
;
1238 err
= netlink_autobind(sock
);
1244 if (len
> sk
->sk_sndbuf
- 32)
1247 skb
= alloc_skb(len
, GFP_KERNEL
);
1251 NETLINK_CB(skb
).pid
= nlk
->pid
;
1252 NETLINK_CB(skb
).dst_group
= dst_group
;
1253 NETLINK_CB(skb
).loginuid
= audit_get_loginuid(current
);
1254 NETLINK_CB(skb
).sessionid
= audit_get_sessionid(current
);
1255 security_task_getsecid(current
, &(NETLINK_CB(skb
).sid
));
1256 memcpy(NETLINK_CREDS(skb
), &siocb
->scm
->creds
, sizeof(struct ucred
));
1258 /* What can I do? Netlink is asynchronous, so that
1259 we will have to save current capabilities to
1260 check them, when this message will be delivered
1261 to corresponding kernel module. --ANK (980802)
1265 if (memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
)) {
1270 err
= security_netlink_send(sk
, skb
);
1277 atomic_inc(&skb
->users
);
1278 netlink_broadcast(sk
, skb
, dst_pid
, dst_group
, GFP_KERNEL
);
1280 err
= netlink_unicast(sk
, skb
, dst_pid
, msg
->msg_flags
&MSG_DONTWAIT
);
1286 static int netlink_recvmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1287 struct msghdr
*msg
, size_t len
,
1290 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1291 struct scm_cookie scm
;
1292 struct sock
*sk
= sock
->sk
;
1293 struct netlink_sock
*nlk
= nlk_sk(sk
);
1294 int noblock
= flags
&MSG_DONTWAIT
;
1296 struct sk_buff
*skb
;
1304 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
1308 msg
->msg_namelen
= 0;
1312 msg
->msg_flags
|= MSG_TRUNC
;
1316 skb_reset_transport_header(skb
);
1317 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1319 if (msg
->msg_name
) {
1320 struct sockaddr_nl
*addr
= (struct sockaddr_nl
*)msg
->msg_name
;
1321 addr
->nl_family
= AF_NETLINK
;
1323 addr
->nl_pid
= NETLINK_CB(skb
).pid
;
1324 addr
->nl_groups
= netlink_group_mask(NETLINK_CB(skb
).dst_group
);
1325 msg
->msg_namelen
= sizeof(*addr
);
1328 if (nlk
->flags
& NETLINK_RECV_PKTINFO
)
1329 netlink_cmsg_recv_pktinfo(msg
, skb
);
1331 if (NULL
== siocb
->scm
) {
1332 memset(&scm
, 0, sizeof(scm
));
1335 siocb
->scm
->creds
= *NETLINK_CREDS(skb
);
1336 if (flags
& MSG_TRUNC
)
1338 skb_free_datagram(sk
, skb
);
1340 if (nlk
->cb
&& atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
/ 2)
1343 scm_recv(sock
, msg
, siocb
->scm
, flags
);
1345 netlink_rcv_wake(sk
);
1346 return err
? : copied
;
1349 static void netlink_data_ready(struct sock
*sk
, int len
)
1355 * We export these functions to other modules. They provide a
1356 * complete set of kernel non-blocking support for message
1361 netlink_kernel_create(struct net
*net
, int unit
, unsigned int groups
,
1362 void (*input
)(struct sk_buff
*skb
),
1363 struct mutex
*cb_mutex
, struct module
*module
)
1365 struct socket
*sock
;
1367 struct netlink_sock
*nlk
;
1368 unsigned long *listeners
= NULL
;
1372 if (unit
< 0 || unit
>= MAX_LINKS
)
1375 if (sock_create_lite(PF_NETLINK
, SOCK_DGRAM
, unit
, &sock
))
1379 * We have to just have a reference on the net from sk, but don't
1380 * get_net it. Besides, we cannot get and then put the net here.
1381 * So we create one inside init_net and the move it to net.
1384 if (__netlink_create(&init_net
, sock
, cb_mutex
, unit
) < 0)
1385 goto out_sock_release_nosk
;
1388 sk_change_net(sk
, net
);
1393 listeners
= kzalloc(NLGRPSZ(groups
), GFP_KERNEL
);
1395 goto out_sock_release
;
1397 sk
->sk_data_ready
= netlink_data_ready
;
1399 nlk_sk(sk
)->netlink_rcv
= input
;
1401 if (netlink_insert(sk
, net
, 0))
1402 goto out_sock_release
;
1405 nlk
->flags
|= NETLINK_KERNEL_SOCKET
;
1407 netlink_table_grab();
1408 if (!nl_table
[unit
].registered
) {
1409 nl_table
[unit
].groups
= groups
;
1410 nl_table
[unit
].listeners
= listeners
;
1411 nl_table
[unit
].cb_mutex
= cb_mutex
;
1412 nl_table
[unit
].module
= module
;
1413 nl_table
[unit
].registered
= 1;
1416 nl_table
[unit
].registered
++;
1418 netlink_table_ungrab();
1423 netlink_kernel_release(sk
);
1426 out_sock_release_nosk
:
1430 EXPORT_SYMBOL(netlink_kernel_create
);
1434 netlink_kernel_release(struct sock
*sk
)
1436 sk_release_kernel(sk
);
1438 EXPORT_SYMBOL(netlink_kernel_release
);
1442 * netlink_change_ngroups - change number of multicast groups
1444 * This changes the number of multicast groups that are available
1445 * on a certain netlink family. Note that it is not possible to
1446 * change the number of groups to below 32. Also note that it does
1447 * not implicitly call netlink_clear_multicast_users() when the
1448 * number of groups is reduced.
1450 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1451 * @groups: The new number of groups.
1453 int netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
1455 unsigned long *listeners
, *old
= NULL
;
1456 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
1462 netlink_table_grab();
1463 if (NLGRPSZ(tbl
->groups
) < NLGRPSZ(groups
)) {
1464 listeners
= kzalloc(NLGRPSZ(groups
), GFP_ATOMIC
);
1469 old
= tbl
->listeners
;
1470 memcpy(listeners
, old
, NLGRPSZ(tbl
->groups
));
1471 rcu_assign_pointer(tbl
->listeners
, listeners
);
1473 tbl
->groups
= groups
;
1476 netlink_table_ungrab();
1481 EXPORT_SYMBOL(netlink_change_ngroups
);
1484 * netlink_clear_multicast_users - kick off multicast listeners
1486 * This function removes all listeners from the given group.
1487 * @ksk: The kernel netlink socket, as returned by
1488 * netlink_kernel_create().
1489 * @group: The multicast group to clear.
1491 void netlink_clear_multicast_users(struct sock
*ksk
, unsigned int group
)
1494 struct hlist_node
*node
;
1495 struct netlink_table
*tbl
= &nl_table
[ksk
->sk_protocol
];
1497 netlink_table_grab();
1499 sk_for_each_bound(sk
, node
, &tbl
->mc_list
)
1500 netlink_update_socket_mc(nlk_sk(sk
), group
, 0);
1502 netlink_table_ungrab();
1504 EXPORT_SYMBOL(netlink_clear_multicast_users
);
1506 void netlink_set_nonroot(int protocol
, unsigned int flags
)
1508 if ((unsigned int)protocol
< MAX_LINKS
)
1509 nl_table
[protocol
].nl_nonroot
= flags
;
1511 EXPORT_SYMBOL(netlink_set_nonroot
);
1513 static void netlink_destroy_callback(struct netlink_callback
*cb
)
1521 * It looks a bit ugly.
1522 * It would be better to create kernel thread.
1525 static int netlink_dump(struct sock
*sk
)
1527 struct netlink_sock
*nlk
= nlk_sk(sk
);
1528 struct netlink_callback
*cb
;
1529 struct sk_buff
*skb
;
1530 struct nlmsghdr
*nlh
;
1531 int len
, err
= -ENOBUFS
;
1533 skb
= sock_rmalloc(sk
, NLMSG_GOODSIZE
, 0, GFP_KERNEL
);
1537 mutex_lock(nlk
->cb_mutex
);
1545 len
= cb
->dump(skb
, cb
);
1548 mutex_unlock(nlk
->cb_mutex
);
1550 if (sk_filter(sk
, skb
))
1553 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1554 sk
->sk_data_ready(sk
, skb
->len
);
1559 nlh
= nlmsg_put_answer(skb
, cb
, NLMSG_DONE
, sizeof(len
), NLM_F_MULTI
);
1563 memcpy(nlmsg_data(nlh
), &len
, sizeof(len
));
1565 if (sk_filter(sk
, skb
))
1568 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1569 sk
->sk_data_ready(sk
, skb
->len
);
1575 mutex_unlock(nlk
->cb_mutex
);
1577 netlink_destroy_callback(cb
);
1581 mutex_unlock(nlk
->cb_mutex
);
1587 int netlink_dump_start(struct sock
*ssk
, struct sk_buff
*skb
,
1588 struct nlmsghdr
*nlh
,
1589 int (*dump
)(struct sk_buff
*skb
,
1590 struct netlink_callback
*),
1591 int (*done
)(struct netlink_callback
*))
1593 struct netlink_callback
*cb
;
1595 struct netlink_sock
*nlk
;
1597 cb
= kzalloc(sizeof(*cb
), GFP_KERNEL
);
1604 atomic_inc(&skb
->users
);
1607 sk
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, NETLINK_CB(skb
).pid
);
1609 netlink_destroy_callback(cb
);
1610 return -ECONNREFUSED
;
1613 /* A dump is in progress... */
1614 mutex_lock(nlk
->cb_mutex
);
1616 mutex_unlock(nlk
->cb_mutex
);
1617 netlink_destroy_callback(cb
);
1622 mutex_unlock(nlk
->cb_mutex
);
1627 /* We successfully started a dump, by returning -EINTR we
1628 * signal not to send ACK even if it was requested.
1632 EXPORT_SYMBOL(netlink_dump_start
);
1634 void netlink_ack(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, int err
)
1636 struct sk_buff
*skb
;
1637 struct nlmsghdr
*rep
;
1638 struct nlmsgerr
*errmsg
;
1639 size_t payload
= sizeof(*errmsg
);
1641 /* error messages get the original request appened */
1643 payload
+= nlmsg_len(nlh
);
1645 skb
= nlmsg_new(payload
, GFP_KERNEL
);
1649 sk
= netlink_lookup(sock_net(in_skb
->sk
),
1650 in_skb
->sk
->sk_protocol
,
1651 NETLINK_CB(in_skb
).pid
);
1653 sk
->sk_err
= ENOBUFS
;
1654 sk
->sk_error_report(sk
);
1660 rep
= __nlmsg_put(skb
, NETLINK_CB(in_skb
).pid
, nlh
->nlmsg_seq
,
1661 NLMSG_ERROR
, sizeof(struct nlmsgerr
), 0);
1662 errmsg
= nlmsg_data(rep
);
1663 errmsg
->error
= err
;
1664 memcpy(&errmsg
->msg
, nlh
, err
? nlh
->nlmsg_len
: sizeof(*nlh
));
1665 netlink_unicast(in_skb
->sk
, skb
, NETLINK_CB(in_skb
).pid
, MSG_DONTWAIT
);
1667 EXPORT_SYMBOL(netlink_ack
);
1669 int netlink_rcv_skb(struct sk_buff
*skb
, int (*cb
)(struct sk_buff
*,
1672 struct nlmsghdr
*nlh
;
1675 while (skb
->len
>= nlmsg_total_size(0)) {
1678 nlh
= nlmsg_hdr(skb
);
1681 if (nlh
->nlmsg_len
< NLMSG_HDRLEN
|| skb
->len
< nlh
->nlmsg_len
)
1684 /* Only requests are handled by the kernel */
1685 if (!(nlh
->nlmsg_flags
& NLM_F_REQUEST
))
1688 /* Skip control messages */
1689 if (nlh
->nlmsg_type
< NLMSG_MIN_TYPE
)
1697 if (nlh
->nlmsg_flags
& NLM_F_ACK
|| err
)
1698 netlink_ack(skb
, nlh
, err
);
1701 msglen
= NLMSG_ALIGN(nlh
->nlmsg_len
);
1702 if (msglen
> skb
->len
)
1704 skb_pull(skb
, msglen
);
1709 EXPORT_SYMBOL(netlink_rcv_skb
);
1712 * nlmsg_notify - send a notification netlink message
1713 * @sk: netlink socket to use
1714 * @skb: notification message
1715 * @pid: destination netlink pid for reports or 0
1716 * @group: destination multicast group or 0
1717 * @report: 1 to report back, 0 to disable
1718 * @flags: allocation flags
1720 int nlmsg_notify(struct sock
*sk
, struct sk_buff
*skb
, u32 pid
,
1721 unsigned int group
, int report
, gfp_t flags
)
1726 int exclude_pid
= 0;
1729 atomic_inc(&skb
->users
);
1733 /* errors reported via destination sk->sk_err */
1734 nlmsg_multicast(sk
, skb
, exclude_pid
, group
, flags
);
1738 err
= nlmsg_unicast(sk
, skb
, pid
);
1742 EXPORT_SYMBOL(nlmsg_notify
);
1744 #ifdef CONFIG_PROC_FS
1745 struct nl_seq_iter
{
1746 struct seq_net_private p
;
1751 static struct sock
*netlink_seq_socket_idx(struct seq_file
*seq
, loff_t pos
)
1753 struct nl_seq_iter
*iter
= seq
->private;
1756 struct hlist_node
*node
;
1759 for (i
= 0; i
< MAX_LINKS
; i
++) {
1760 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1762 for (j
= 0; j
<= hash
->mask
; j
++) {
1763 sk_for_each(s
, node
, &hash
->table
[j
]) {
1764 if (sock_net(s
) != seq_file_net(seq
))
1778 static void *netlink_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1779 __acquires(nl_table_lock
)
1781 read_lock(&nl_table_lock
);
1782 return *pos
? netlink_seq_socket_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1785 static void *netlink_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1788 struct nl_seq_iter
*iter
;
1793 if (v
== SEQ_START_TOKEN
)
1794 return netlink_seq_socket_idx(seq
, 0);
1796 iter
= seq
->private;
1800 } while (s
&& sock_net(s
) != seq_file_net(seq
));
1805 j
= iter
->hash_idx
+ 1;
1808 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1810 for (; j
<= hash
->mask
; j
++) {
1811 s
= sk_head(&hash
->table
[j
]);
1812 while (s
&& sock_net(s
) != seq_file_net(seq
))
1822 } while (++i
< MAX_LINKS
);
1827 static void netlink_seq_stop(struct seq_file
*seq
, void *v
)
1828 __releases(nl_table_lock
)
1830 read_unlock(&nl_table_lock
);
1834 static int netlink_seq_show(struct seq_file
*seq
, void *v
)
1836 if (v
== SEQ_START_TOKEN
)
1838 "sk Eth Pid Groups "
1839 "Rmem Wmem Dump Locks\n");
1842 struct netlink_sock
*nlk
= nlk_sk(s
);
1844 seq_printf(seq
, "%p %-3d %-6d %08x %-8d %-8d %p %d\n",
1848 nlk
->groups
? (u32
)nlk
->groups
[0] : 0,
1849 atomic_read(&s
->sk_rmem_alloc
),
1850 atomic_read(&s
->sk_wmem_alloc
),
1852 atomic_read(&s
->sk_refcnt
)
1859 static const struct seq_operations netlink_seq_ops
= {
1860 .start
= netlink_seq_start
,
1861 .next
= netlink_seq_next
,
1862 .stop
= netlink_seq_stop
,
1863 .show
= netlink_seq_show
,
1867 static int netlink_seq_open(struct inode
*inode
, struct file
*file
)
1869 return seq_open_net(inode
, file
, &netlink_seq_ops
,
1870 sizeof(struct nl_seq_iter
));
1873 static const struct file_operations netlink_seq_fops
= {
1874 .owner
= THIS_MODULE
,
1875 .open
= netlink_seq_open
,
1877 .llseek
= seq_lseek
,
1878 .release
= seq_release_net
,
1883 int netlink_register_notifier(struct notifier_block
*nb
)
1885 return atomic_notifier_chain_register(&netlink_chain
, nb
);
1887 EXPORT_SYMBOL(netlink_register_notifier
);
1889 int netlink_unregister_notifier(struct notifier_block
*nb
)
1891 return atomic_notifier_chain_unregister(&netlink_chain
, nb
);
1893 EXPORT_SYMBOL(netlink_unregister_notifier
);
1895 static const struct proto_ops netlink_ops
= {
1896 .family
= PF_NETLINK
,
1897 .owner
= THIS_MODULE
,
1898 .release
= netlink_release
,
1899 .bind
= netlink_bind
,
1900 .connect
= netlink_connect
,
1901 .socketpair
= sock_no_socketpair
,
1902 .accept
= sock_no_accept
,
1903 .getname
= netlink_getname
,
1904 .poll
= datagram_poll
,
1905 .ioctl
= sock_no_ioctl
,
1906 .listen
= sock_no_listen
,
1907 .shutdown
= sock_no_shutdown
,
1908 .setsockopt
= netlink_setsockopt
,
1909 .getsockopt
= netlink_getsockopt
,
1910 .sendmsg
= netlink_sendmsg
,
1911 .recvmsg
= netlink_recvmsg
,
1912 .mmap
= sock_no_mmap
,
1913 .sendpage
= sock_no_sendpage
,
1916 static struct net_proto_family netlink_family_ops
= {
1917 .family
= PF_NETLINK
,
1918 .create
= netlink_create
,
1919 .owner
= THIS_MODULE
, /* for consistency 8) */
1922 static int __net_init
netlink_net_init(struct net
*net
)
1924 #ifdef CONFIG_PROC_FS
1925 if (!proc_net_fops_create(net
, "netlink", 0, &netlink_seq_fops
))
1931 static void __net_exit
netlink_net_exit(struct net
*net
)
1933 #ifdef CONFIG_PROC_FS
1934 proc_net_remove(net
, "netlink");
1938 static struct pernet_operations __net_initdata netlink_net_ops
= {
1939 .init
= netlink_net_init
,
1940 .exit
= netlink_net_exit
,
1943 static int __init
netlink_proto_init(void)
1945 struct sk_buff
*dummy_skb
;
1947 unsigned long limit
;
1949 int err
= proto_register(&netlink_proto
, 0);
1954 BUILD_BUG_ON(sizeof(struct netlink_skb_parms
) > sizeof(dummy_skb
->cb
));
1956 nl_table
= kcalloc(MAX_LINKS
, sizeof(*nl_table
), GFP_KERNEL
);
1960 if (num_physpages
>= (128 * 1024))
1961 limit
= num_physpages
>> (21 - PAGE_SHIFT
);
1963 limit
= num_physpages
>> (23 - PAGE_SHIFT
);
1965 order
= get_bitmask_order(limit
) - 1 + PAGE_SHIFT
;
1966 limit
= (1UL << order
) / sizeof(struct hlist_head
);
1967 order
= get_bitmask_order(min(limit
, (unsigned long)UINT_MAX
)) - 1;
1969 for (i
= 0; i
< MAX_LINKS
; i
++) {
1970 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1972 hash
->table
= nl_pid_hash_zalloc(1 * sizeof(*hash
->table
));
1975 nl_pid_hash_free(nl_table
[i
].hash
.table
,
1976 1 * sizeof(*hash
->table
));
1980 hash
->max_shift
= order
;
1983 hash
->rehash_time
= jiffies
;
1986 sock_register(&netlink_family_ops
);
1987 register_pernet_subsys(&netlink_net_ops
);
1988 /* The netlink device handler may be needed early. */
1993 panic("netlink_init: Cannot allocate nl_table\n");
1996 core_initcall(netlink_proto_init
);