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 struct listeners_rcu_head
{
87 struct rcu_head rcu_head
;
91 #define NETLINK_KERNEL_SOCKET 0x1
92 #define NETLINK_RECV_PKTINFO 0x2
93 #define NETLINK_BROADCAST_SEND_ERROR 0x4
94 #define NETLINK_RECV_NO_ENOBUFS 0x8
96 static inline struct netlink_sock
*nlk_sk(struct sock
*sk
)
98 return container_of(sk
, struct netlink_sock
, sk
);
101 static inline int netlink_is_kernel(struct sock
*sk
)
103 return nlk_sk(sk
)->flags
& NETLINK_KERNEL_SOCKET
;
107 struct hlist_head
*table
;
108 unsigned long rehash_time
;
113 unsigned int entries
;
114 unsigned int max_shift
;
119 struct netlink_table
{
120 struct nl_pid_hash hash
;
121 struct hlist_head mc_list
;
122 unsigned long *listeners
;
123 unsigned int nl_nonroot
;
125 struct mutex
*cb_mutex
;
126 struct module
*module
;
130 static struct netlink_table
*nl_table
;
132 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait
);
134 static int netlink_dump(struct sock
*sk
);
135 static void netlink_destroy_callback(struct netlink_callback
*cb
);
137 static DEFINE_RWLOCK(nl_table_lock
);
138 static atomic_t nl_table_users
= ATOMIC_INIT(0);
140 static ATOMIC_NOTIFIER_HEAD(netlink_chain
);
142 static u32
netlink_group_mask(u32 group
)
144 return group
? 1 << (group
- 1) : 0;
147 static struct hlist_head
*nl_pid_hashfn(struct nl_pid_hash
*hash
, u32 pid
)
149 return &hash
->table
[jhash_1word(pid
, hash
->rnd
) & hash
->mask
];
152 static void netlink_sock_destruct(struct sock
*sk
)
154 struct netlink_sock
*nlk
= nlk_sk(sk
);
158 nlk
->cb
->done(nlk
->cb
);
159 netlink_destroy_callback(nlk
->cb
);
162 skb_queue_purge(&sk
->sk_receive_queue
);
164 if (!sock_flag(sk
, SOCK_DEAD
)) {
165 printk(KERN_ERR
"Freeing alive netlink socket %p\n", sk
);
169 WARN_ON(atomic_read(&sk
->sk_rmem_alloc
));
170 WARN_ON(atomic_read(&sk
->sk_wmem_alloc
));
171 WARN_ON(nlk_sk(sk
)->groups
);
174 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
175 * SMP. Look, when several writers sleep and reader wakes them up, all but one
176 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
177 * this, _but_ remember, it adds useless work on UP machines.
180 void netlink_table_grab(void)
181 __acquires(nl_table_lock
)
185 write_lock_irq(&nl_table_lock
);
187 if (atomic_read(&nl_table_users
)) {
188 DECLARE_WAITQUEUE(wait
, current
);
190 add_wait_queue_exclusive(&nl_table_wait
, &wait
);
192 set_current_state(TASK_UNINTERRUPTIBLE
);
193 if (atomic_read(&nl_table_users
) == 0)
195 write_unlock_irq(&nl_table_lock
);
197 write_lock_irq(&nl_table_lock
);
200 __set_current_state(TASK_RUNNING
);
201 remove_wait_queue(&nl_table_wait
, &wait
);
205 void netlink_table_ungrab(void)
206 __releases(nl_table_lock
)
208 write_unlock_irq(&nl_table_lock
);
209 wake_up(&nl_table_wait
);
213 netlink_lock_table(void)
215 /* read_lock() synchronizes us to netlink_table_grab */
217 read_lock(&nl_table_lock
);
218 atomic_inc(&nl_table_users
);
219 read_unlock(&nl_table_lock
);
223 netlink_unlock_table(void)
225 if (atomic_dec_and_test(&nl_table_users
))
226 wake_up(&nl_table_wait
);
229 static inline struct sock
*netlink_lookup(struct net
*net
, int protocol
,
232 struct nl_pid_hash
*hash
= &nl_table
[protocol
].hash
;
233 struct hlist_head
*head
;
235 struct hlist_node
*node
;
237 read_lock(&nl_table_lock
);
238 head
= nl_pid_hashfn(hash
, pid
);
239 sk_for_each(sk
, node
, head
) {
240 if (net_eq(sock_net(sk
), net
) && (nlk_sk(sk
)->pid
== pid
)) {
247 read_unlock(&nl_table_lock
);
251 static inline struct hlist_head
*nl_pid_hash_zalloc(size_t size
)
253 if (size
<= PAGE_SIZE
)
254 return kzalloc(size
, GFP_ATOMIC
);
256 return (struct hlist_head
*)
257 __get_free_pages(GFP_ATOMIC
| __GFP_ZERO
,
261 static inline void nl_pid_hash_free(struct hlist_head
*table
, size_t size
)
263 if (size
<= PAGE_SIZE
)
266 free_pages((unsigned long)table
, get_order(size
));
269 static int nl_pid_hash_rehash(struct nl_pid_hash
*hash
, int grow
)
271 unsigned int omask
, mask
, shift
;
273 struct hlist_head
*otable
, *table
;
276 omask
= mask
= hash
->mask
;
277 osize
= size
= (mask
+ 1) * sizeof(*table
);
281 if (++shift
> hash
->max_shift
)
287 table
= nl_pid_hash_zalloc(size
);
291 otable
= hash
->table
;
295 get_random_bytes(&hash
->rnd
, sizeof(hash
->rnd
));
297 for (i
= 0; i
<= omask
; i
++) {
299 struct hlist_node
*node
, *tmp
;
301 sk_for_each_safe(sk
, node
, tmp
, &otable
[i
])
302 __sk_add_node(sk
, nl_pid_hashfn(hash
, nlk_sk(sk
)->pid
));
305 nl_pid_hash_free(otable
, osize
);
306 hash
->rehash_time
= jiffies
+ 10 * 60 * HZ
;
310 static inline int nl_pid_hash_dilute(struct nl_pid_hash
*hash
, int len
)
312 int avg
= hash
->entries
>> hash
->shift
;
314 if (unlikely(avg
> 1) && nl_pid_hash_rehash(hash
, 1))
317 if (unlikely(len
> avg
) && time_after(jiffies
, hash
->rehash_time
)) {
318 nl_pid_hash_rehash(hash
, 0);
325 static const struct proto_ops netlink_ops
;
328 netlink_update_listeners(struct sock
*sk
)
330 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
331 struct hlist_node
*node
;
335 for (i
= 0; i
< NLGRPLONGS(tbl
->groups
); i
++) {
337 sk_for_each_bound(sk
, node
, &tbl
->mc_list
) {
338 if (i
< NLGRPLONGS(nlk_sk(sk
)->ngroups
))
339 mask
|= nlk_sk(sk
)->groups
[i
];
341 tbl
->listeners
[i
] = mask
;
343 /* this function is only called with the netlink table "grabbed", which
344 * makes sure updates are visible before bind or setsockopt return. */
347 static int netlink_insert(struct sock
*sk
, struct net
*net
, u32 pid
)
349 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
350 struct hlist_head
*head
;
351 int err
= -EADDRINUSE
;
353 struct hlist_node
*node
;
356 netlink_table_grab();
357 head
= nl_pid_hashfn(hash
, pid
);
359 sk_for_each(osk
, node
, head
) {
360 if (net_eq(sock_net(osk
), net
) && (nlk_sk(osk
)->pid
== pid
))
372 if (BITS_PER_LONG
> 32 && unlikely(hash
->entries
>= UINT_MAX
))
375 if (len
&& nl_pid_hash_dilute(hash
, len
))
376 head
= nl_pid_hashfn(hash
, pid
);
378 nlk_sk(sk
)->pid
= pid
;
379 sk_add_node(sk
, head
);
383 netlink_table_ungrab();
387 static void netlink_remove(struct sock
*sk
)
389 netlink_table_grab();
390 if (sk_del_node_init(sk
))
391 nl_table
[sk
->sk_protocol
].hash
.entries
--;
392 if (nlk_sk(sk
)->subscriptions
)
393 __sk_del_bind_node(sk
);
394 netlink_table_ungrab();
397 static struct proto netlink_proto
= {
399 .owner
= THIS_MODULE
,
400 .obj_size
= sizeof(struct netlink_sock
),
403 static int __netlink_create(struct net
*net
, struct socket
*sock
,
404 struct mutex
*cb_mutex
, int protocol
)
407 struct netlink_sock
*nlk
;
409 sock
->ops
= &netlink_ops
;
411 sk
= sk_alloc(net
, PF_NETLINK
, GFP_KERNEL
, &netlink_proto
);
415 sock_init_data(sock
, sk
);
419 nlk
->cb_mutex
= cb_mutex
;
421 nlk
->cb_mutex
= &nlk
->cb_def_mutex
;
422 mutex_init(nlk
->cb_mutex
);
424 init_waitqueue_head(&nlk
->wait
);
426 sk
->sk_destruct
= netlink_sock_destruct
;
427 sk
->sk_protocol
= protocol
;
431 static int netlink_create(struct net
*net
, struct socket
*sock
, int protocol
,
434 struct module
*module
= NULL
;
435 struct mutex
*cb_mutex
;
436 struct netlink_sock
*nlk
;
439 sock
->state
= SS_UNCONNECTED
;
441 if (sock
->type
!= SOCK_RAW
&& sock
->type
!= SOCK_DGRAM
)
442 return -ESOCKTNOSUPPORT
;
444 if (protocol
< 0 || protocol
>= MAX_LINKS
)
445 return -EPROTONOSUPPORT
;
447 netlink_lock_table();
448 #ifdef CONFIG_MODULES
449 if (!nl_table
[protocol
].registered
) {
450 netlink_unlock_table();
451 request_module("net-pf-%d-proto-%d", PF_NETLINK
, protocol
);
452 netlink_lock_table();
455 if (nl_table
[protocol
].registered
&&
456 try_module_get(nl_table
[protocol
].module
))
457 module
= nl_table
[protocol
].module
;
459 err
= -EPROTONOSUPPORT
;
460 cb_mutex
= nl_table
[protocol
].cb_mutex
;
461 netlink_unlock_table();
466 err
= __netlink_create(net
, sock
, cb_mutex
, protocol
);
471 sock_prot_inuse_add(net
, &netlink_proto
, 1);
474 nlk
= nlk_sk(sock
->sk
);
475 nlk
->module
= module
;
484 static int netlink_release(struct socket
*sock
)
486 struct sock
*sk
= sock
->sk
;
487 struct netlink_sock
*nlk
;
497 * OK. Socket is unlinked, any packets that arrive now
502 wake_up_interruptible_all(&nlk
->wait
);
504 skb_queue_purge(&sk
->sk_write_queue
);
507 struct netlink_notify n
= {
509 .protocol
= sk
->sk_protocol
,
512 atomic_notifier_call_chain(&netlink_chain
,
513 NETLINK_URELEASE
, &n
);
516 module_put(nlk
->module
);
518 netlink_table_grab();
519 if (netlink_is_kernel(sk
)) {
520 BUG_ON(nl_table
[sk
->sk_protocol
].registered
== 0);
521 if (--nl_table
[sk
->sk_protocol
].registered
== 0) {
522 kfree(nl_table
[sk
->sk_protocol
].listeners
);
523 nl_table
[sk
->sk_protocol
].module
= NULL
;
524 nl_table
[sk
->sk_protocol
].registered
= 0;
526 } else if (nlk
->subscriptions
)
527 netlink_update_listeners(sk
);
528 netlink_table_ungrab();
534 sock_prot_inuse_add(sock_net(sk
), &netlink_proto
, -1);
540 static int netlink_autobind(struct socket
*sock
)
542 struct sock
*sk
= sock
->sk
;
543 struct net
*net
= sock_net(sk
);
544 struct nl_pid_hash
*hash
= &nl_table
[sk
->sk_protocol
].hash
;
545 struct hlist_head
*head
;
547 struct hlist_node
*node
;
548 s32 pid
= task_tgid_vnr(current
);
550 static s32 rover
= -4097;
554 netlink_table_grab();
555 head
= nl_pid_hashfn(hash
, pid
);
556 sk_for_each(osk
, node
, head
) {
557 if (!net_eq(sock_net(osk
), net
))
559 if (nlk_sk(osk
)->pid
== pid
) {
560 /* Bind collision, search negative pid values. */
564 netlink_table_ungrab();
568 netlink_table_ungrab();
570 err
= netlink_insert(sk
, net
, pid
);
571 if (err
== -EADDRINUSE
)
574 /* If 2 threads race to autobind, that is fine. */
581 static inline int netlink_capable(struct socket
*sock
, unsigned int flag
)
583 return (nl_table
[sock
->sk
->sk_protocol
].nl_nonroot
& flag
) ||
584 capable(CAP_NET_ADMIN
);
588 netlink_update_subscriptions(struct sock
*sk
, unsigned int subscriptions
)
590 struct netlink_sock
*nlk
= nlk_sk(sk
);
592 if (nlk
->subscriptions
&& !subscriptions
)
593 __sk_del_bind_node(sk
);
594 else if (!nlk
->subscriptions
&& subscriptions
)
595 sk_add_bind_node(sk
, &nl_table
[sk
->sk_protocol
].mc_list
);
596 nlk
->subscriptions
= subscriptions
;
599 static int netlink_realloc_groups(struct sock
*sk
)
601 struct netlink_sock
*nlk
= nlk_sk(sk
);
603 unsigned long *new_groups
;
606 netlink_table_grab();
608 groups
= nl_table
[sk
->sk_protocol
].groups
;
609 if (!nl_table
[sk
->sk_protocol
].registered
) {
614 if (nlk
->ngroups
>= groups
)
617 new_groups
= krealloc(nlk
->groups
, NLGRPSZ(groups
), GFP_ATOMIC
);
618 if (new_groups
== NULL
) {
622 memset((char *)new_groups
+ NLGRPSZ(nlk
->ngroups
), 0,
623 NLGRPSZ(groups
) - NLGRPSZ(nlk
->ngroups
));
625 nlk
->groups
= new_groups
;
626 nlk
->ngroups
= groups
;
628 netlink_table_ungrab();
632 static int netlink_bind(struct socket
*sock
, struct sockaddr
*addr
,
635 struct sock
*sk
= sock
->sk
;
636 struct net
*net
= sock_net(sk
);
637 struct netlink_sock
*nlk
= nlk_sk(sk
);
638 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
641 if (nladdr
->nl_family
!= AF_NETLINK
)
644 /* Only superuser is allowed to listen multicasts */
645 if (nladdr
->nl_groups
) {
646 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
648 err
= netlink_realloc_groups(sk
);
654 if (nladdr
->nl_pid
!= nlk
->pid
)
657 err
= nladdr
->nl_pid
?
658 netlink_insert(sk
, net
, nladdr
->nl_pid
) :
659 netlink_autobind(sock
);
664 if (!nladdr
->nl_groups
&& (nlk
->groups
== NULL
|| !(u32
)nlk
->groups
[0]))
667 netlink_table_grab();
668 netlink_update_subscriptions(sk
, nlk
->subscriptions
+
669 hweight32(nladdr
->nl_groups
) -
670 hweight32(nlk
->groups
[0]));
671 nlk
->groups
[0] = (nlk
->groups
[0] & ~0xffffffffUL
) | nladdr
->nl_groups
;
672 netlink_update_listeners(sk
);
673 netlink_table_ungrab();
678 static int netlink_connect(struct socket
*sock
, struct sockaddr
*addr
,
682 struct sock
*sk
= sock
->sk
;
683 struct netlink_sock
*nlk
= nlk_sk(sk
);
684 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
686 if (alen
< sizeof(addr
->sa_family
))
689 if (addr
->sa_family
== AF_UNSPEC
) {
690 sk
->sk_state
= NETLINK_UNCONNECTED
;
695 if (addr
->sa_family
!= AF_NETLINK
)
698 /* Only superuser is allowed to send multicasts */
699 if (nladdr
->nl_groups
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
703 err
= netlink_autobind(sock
);
706 sk
->sk_state
= NETLINK_CONNECTED
;
707 nlk
->dst_pid
= nladdr
->nl_pid
;
708 nlk
->dst_group
= ffs(nladdr
->nl_groups
);
714 static int netlink_getname(struct socket
*sock
, struct sockaddr
*addr
,
715 int *addr_len
, int peer
)
717 struct sock
*sk
= sock
->sk
;
718 struct netlink_sock
*nlk
= nlk_sk(sk
);
719 DECLARE_SOCKADDR(struct sockaddr_nl
*, nladdr
, addr
);
721 nladdr
->nl_family
= AF_NETLINK
;
723 *addr_len
= sizeof(*nladdr
);
726 nladdr
->nl_pid
= nlk
->dst_pid
;
727 nladdr
->nl_groups
= netlink_group_mask(nlk
->dst_group
);
729 nladdr
->nl_pid
= nlk
->pid
;
730 nladdr
->nl_groups
= nlk
->groups
? nlk
->groups
[0] : 0;
735 static void netlink_overrun(struct sock
*sk
)
737 struct netlink_sock
*nlk
= nlk_sk(sk
);
739 if (!(nlk
->flags
& NETLINK_RECV_NO_ENOBUFS
)) {
740 if (!test_and_set_bit(0, &nlk_sk(sk
)->state
)) {
741 sk
->sk_err
= ENOBUFS
;
742 sk
->sk_error_report(sk
);
745 atomic_inc(&sk
->sk_drops
);
748 static struct sock
*netlink_getsockbypid(struct sock
*ssk
, u32 pid
)
751 struct netlink_sock
*nlk
;
753 sock
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, pid
);
755 return ERR_PTR(-ECONNREFUSED
);
757 /* Don't bother queuing skb if kernel socket has no input function */
759 if (sock
->sk_state
== NETLINK_CONNECTED
&&
760 nlk
->dst_pid
!= nlk_sk(ssk
)->pid
) {
762 return ERR_PTR(-ECONNREFUSED
);
767 struct sock
*netlink_getsockbyfilp(struct file
*filp
)
769 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
772 if (!S_ISSOCK(inode
->i_mode
))
773 return ERR_PTR(-ENOTSOCK
);
775 sock
= SOCKET_I(inode
)->sk
;
776 if (sock
->sk_family
!= AF_NETLINK
)
777 return ERR_PTR(-EINVAL
);
784 * Attach a skb to a netlink socket.
785 * The caller must hold a reference to the destination socket. On error, the
786 * reference is dropped. The skb is not send to the destination, just all
787 * all error checks are performed and memory in the queue is reserved.
789 * < 0: error. skb freed, reference to sock dropped.
791 * 1: repeat lookup - reference dropped while waiting for socket memory.
793 int netlink_attachskb(struct sock
*sk
, struct sk_buff
*skb
,
794 long *timeo
, struct sock
*ssk
)
796 struct netlink_sock
*nlk
;
800 if (atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
801 test_bit(0, &nlk
->state
)) {
802 DECLARE_WAITQUEUE(wait
, current
);
804 if (!ssk
|| netlink_is_kernel(ssk
))
811 __set_current_state(TASK_INTERRUPTIBLE
);
812 add_wait_queue(&nlk
->wait
, &wait
);
814 if ((atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
815 test_bit(0, &nlk
->state
)) &&
816 !sock_flag(sk
, SOCK_DEAD
))
817 *timeo
= schedule_timeout(*timeo
);
819 __set_current_state(TASK_RUNNING
);
820 remove_wait_queue(&nlk
->wait
, &wait
);
823 if (signal_pending(current
)) {
825 return sock_intr_errno(*timeo
);
829 skb_set_owner_r(skb
, sk
);
833 int netlink_sendskb(struct sock
*sk
, struct sk_buff
*skb
)
837 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
838 sk
->sk_data_ready(sk
, len
);
843 void netlink_detachskb(struct sock
*sk
, struct sk_buff
*skb
)
849 static inline struct sk_buff
*netlink_trim(struct sk_buff
*skb
,
856 delta
= skb
->end
- skb
->tail
;
857 if (delta
* 2 < skb
->truesize
)
860 if (skb_shared(skb
)) {
861 struct sk_buff
*nskb
= skb_clone(skb
, allocation
);
868 if (!pskb_expand_head(skb
, 0, -delta
, allocation
))
869 skb
->truesize
-= delta
;
874 static inline void netlink_rcv_wake(struct sock
*sk
)
876 struct netlink_sock
*nlk
= nlk_sk(sk
);
878 if (skb_queue_empty(&sk
->sk_receive_queue
))
879 clear_bit(0, &nlk
->state
);
880 if (!test_bit(0, &nlk
->state
))
881 wake_up_interruptible(&nlk
->wait
);
884 static inline int netlink_unicast_kernel(struct sock
*sk
, struct sk_buff
*skb
)
887 struct netlink_sock
*nlk
= nlk_sk(sk
);
890 if (nlk
->netlink_rcv
!= NULL
) {
892 skb_set_owner_r(skb
, sk
);
893 nlk
->netlink_rcv(skb
);
900 int netlink_unicast(struct sock
*ssk
, struct sk_buff
*skb
,
901 u32 pid
, int nonblock
)
907 skb
= netlink_trim(skb
, gfp_any());
909 timeo
= sock_sndtimeo(ssk
, nonblock
);
911 sk
= netlink_getsockbypid(ssk
, pid
);
916 if (netlink_is_kernel(sk
))
917 return netlink_unicast_kernel(sk
, skb
);
919 if (sk_filter(sk
, skb
)) {
926 err
= netlink_attachskb(sk
, skb
, &timeo
, ssk
);
932 return netlink_sendskb(sk
, skb
);
934 EXPORT_SYMBOL(netlink_unicast
);
936 int netlink_has_listeners(struct sock
*sk
, unsigned int group
)
939 unsigned long *listeners
;
941 BUG_ON(!netlink_is_kernel(sk
));
944 listeners
= rcu_dereference(nl_table
[sk
->sk_protocol
].listeners
);
946 if (group
- 1 < nl_table
[sk
->sk_protocol
].groups
)
947 res
= test_bit(group
- 1, listeners
);
953 EXPORT_SYMBOL_GPL(netlink_has_listeners
);
955 static inline int netlink_broadcast_deliver(struct sock
*sk
,
958 struct netlink_sock
*nlk
= nlk_sk(sk
);
960 if (atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
&&
961 !test_bit(0, &nlk
->state
)) {
962 skb_set_owner_r(skb
, sk
);
963 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
964 sk
->sk_data_ready(sk
, skb
->len
);
965 return atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
;
970 struct netlink_broadcast_data
{
971 struct sock
*exclude_sk
;
976 int delivery_failure
;
980 struct sk_buff
*skb
, *skb2
;
981 int (*tx_filter
)(struct sock
*dsk
, struct sk_buff
*skb
, void *data
);
985 static inline int do_one_broadcast(struct sock
*sk
,
986 struct netlink_broadcast_data
*p
)
988 struct netlink_sock
*nlk
= nlk_sk(sk
);
991 if (p
->exclude_sk
== sk
)
994 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
995 !test_bit(p
->group
- 1, nlk
->groups
))
998 if (!net_eq(sock_net(sk
), p
->net
))
1002 netlink_overrun(sk
);
1007 if (p
->skb2
== NULL
) {
1008 if (skb_shared(p
->skb
)) {
1009 p
->skb2
= skb_clone(p
->skb
, p
->allocation
);
1011 p
->skb2
= skb_get(p
->skb
);
1013 * skb ownership may have been set when
1014 * delivered to a previous socket.
1016 skb_orphan(p
->skb2
);
1019 if (p
->skb2
== NULL
) {
1020 netlink_overrun(sk
);
1021 /* Clone failed. Notify ALL listeners. */
1023 if (nlk
->flags
& NETLINK_BROADCAST_SEND_ERROR
)
1024 p
->delivery_failure
= 1;
1025 } else if (p
->tx_filter
&& p
->tx_filter(sk
, p
->skb2
, p
->tx_data
)) {
1028 } else if (sk_filter(sk
, p
->skb2
)) {
1031 } else if ((val
= netlink_broadcast_deliver(sk
, p
->skb2
)) < 0) {
1032 netlink_overrun(sk
);
1033 if (nlk
->flags
& NETLINK_BROADCAST_SEND_ERROR
)
1034 p
->delivery_failure
= 1;
1036 p
->congested
|= val
;
1046 int netlink_broadcast_filtered(struct sock
*ssk
, struct sk_buff
*skb
, u32 pid
,
1047 u32 group
, gfp_t allocation
,
1048 int (*filter
)(struct sock
*dsk
, struct sk_buff
*skb
, void *data
),
1051 struct net
*net
= sock_net(ssk
);
1052 struct netlink_broadcast_data info
;
1053 struct hlist_node
*node
;
1056 skb
= netlink_trim(skb
, allocation
);
1058 info
.exclude_sk
= ssk
;
1063 info
.delivery_failure
= 0;
1066 info
.allocation
= allocation
;
1069 info
.tx_filter
= filter
;
1070 info
.tx_data
= filter_data
;
1072 /* While we sleep in clone, do not allow to change socket list */
1074 netlink_lock_table();
1076 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1077 do_one_broadcast(sk
, &info
);
1081 netlink_unlock_table();
1083 kfree_skb(info
.skb2
);
1085 if (info
.delivery_failure
)
1088 if (info
.delivered
) {
1089 if (info
.congested
&& (allocation
& __GFP_WAIT
))
1095 EXPORT_SYMBOL(netlink_broadcast_filtered
);
1097 int netlink_broadcast(struct sock
*ssk
, struct sk_buff
*skb
, u32 pid
,
1098 u32 group
, gfp_t allocation
)
1100 return netlink_broadcast_filtered(ssk
, skb
, pid
, group
, allocation
,
1103 EXPORT_SYMBOL(netlink_broadcast
);
1105 struct netlink_set_err_data
{
1106 struct sock
*exclude_sk
;
1112 static inline int do_one_set_err(struct sock
*sk
,
1113 struct netlink_set_err_data
*p
)
1115 struct netlink_sock
*nlk
= nlk_sk(sk
);
1118 if (sk
== p
->exclude_sk
)
1121 if (!net_eq(sock_net(sk
), sock_net(p
->exclude_sk
)))
1124 if (nlk
->pid
== p
->pid
|| p
->group
- 1 >= nlk
->ngroups
||
1125 !test_bit(p
->group
- 1, nlk
->groups
))
1128 if (p
->code
== ENOBUFS
&& nlk
->flags
& NETLINK_RECV_NO_ENOBUFS
) {
1133 sk
->sk_err
= p
->code
;
1134 sk
->sk_error_report(sk
);
1140 * netlink_set_err - report error to broadcast listeners
1141 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1142 * @pid: the PID of a process that we want to skip (if any)
1143 * @groups: the broadcast group that will notice the error
1144 * @code: error code, must be negative (as usual in kernelspace)
1146 * This function returns the number of broadcast listeners that have set the
1147 * NETLINK_RECV_NO_ENOBUFS socket option.
1149 int netlink_set_err(struct sock
*ssk
, u32 pid
, u32 group
, int code
)
1151 struct netlink_set_err_data info
;
1152 struct hlist_node
*node
;
1156 info
.exclude_sk
= ssk
;
1159 /* sk->sk_err wants a positive error value */
1162 read_lock(&nl_table_lock
);
1164 sk_for_each_bound(sk
, node
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1165 ret
+= do_one_set_err(sk
, &info
);
1167 read_unlock(&nl_table_lock
);
1170 EXPORT_SYMBOL(netlink_set_err
);
1172 /* must be called with netlink table grabbed */
1173 static void netlink_update_socket_mc(struct netlink_sock
*nlk
,
1177 int old
, new = !!is_new
, subscriptions
;
1179 old
= test_bit(group
- 1, nlk
->groups
);
1180 subscriptions
= nlk
->subscriptions
- old
+ new;
1182 __set_bit(group
- 1, nlk
->groups
);
1184 __clear_bit(group
- 1, nlk
->groups
);
1185 netlink_update_subscriptions(&nlk
->sk
, subscriptions
);
1186 netlink_update_listeners(&nlk
->sk
);
1189 static int netlink_setsockopt(struct socket
*sock
, int level
, int optname
,
1190 char __user
*optval
, unsigned int optlen
)
1192 struct sock
*sk
= sock
->sk
;
1193 struct netlink_sock
*nlk
= nlk_sk(sk
);
1194 unsigned int val
= 0;
1197 if (level
!= SOL_NETLINK
)
1198 return -ENOPROTOOPT
;
1200 if (optlen
>= sizeof(int) &&
1201 get_user(val
, (unsigned int __user
*)optval
))
1205 case NETLINK_PKTINFO
:
1207 nlk
->flags
|= NETLINK_RECV_PKTINFO
;
1209 nlk
->flags
&= ~NETLINK_RECV_PKTINFO
;
1212 case NETLINK_ADD_MEMBERSHIP
:
1213 case NETLINK_DROP_MEMBERSHIP
: {
1214 if (!netlink_capable(sock
, NL_NONROOT_RECV
))
1216 err
= netlink_realloc_groups(sk
);
1219 if (!val
|| val
- 1 >= nlk
->ngroups
)
1221 netlink_table_grab();
1222 netlink_update_socket_mc(nlk
, val
,
1223 optname
== NETLINK_ADD_MEMBERSHIP
);
1224 netlink_table_ungrab();
1228 case NETLINK_BROADCAST_ERROR
:
1230 nlk
->flags
|= NETLINK_BROADCAST_SEND_ERROR
;
1232 nlk
->flags
&= ~NETLINK_BROADCAST_SEND_ERROR
;
1235 case NETLINK_NO_ENOBUFS
:
1237 nlk
->flags
|= NETLINK_RECV_NO_ENOBUFS
;
1238 clear_bit(0, &nlk
->state
);
1239 wake_up_interruptible(&nlk
->wait
);
1241 nlk
->flags
&= ~NETLINK_RECV_NO_ENOBUFS
;
1250 static int netlink_getsockopt(struct socket
*sock
, int level
, int optname
,
1251 char __user
*optval
, int __user
*optlen
)
1253 struct sock
*sk
= sock
->sk
;
1254 struct netlink_sock
*nlk
= nlk_sk(sk
);
1257 if (level
!= SOL_NETLINK
)
1258 return -ENOPROTOOPT
;
1260 if (get_user(len
, optlen
))
1266 case NETLINK_PKTINFO
:
1267 if (len
< sizeof(int))
1270 val
= nlk
->flags
& NETLINK_RECV_PKTINFO
? 1 : 0;
1271 if (put_user(len
, optlen
) ||
1272 put_user(val
, optval
))
1276 case NETLINK_BROADCAST_ERROR
:
1277 if (len
< sizeof(int))
1280 val
= nlk
->flags
& NETLINK_BROADCAST_SEND_ERROR
? 1 : 0;
1281 if (put_user(len
, optlen
) ||
1282 put_user(val
, optval
))
1286 case NETLINK_NO_ENOBUFS
:
1287 if (len
< sizeof(int))
1290 val
= nlk
->flags
& NETLINK_RECV_NO_ENOBUFS
? 1 : 0;
1291 if (put_user(len
, optlen
) ||
1292 put_user(val
, optval
))
1302 static void netlink_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
1304 struct nl_pktinfo info
;
1306 info
.group
= NETLINK_CB(skb
).dst_group
;
1307 put_cmsg(msg
, SOL_NETLINK
, NETLINK_PKTINFO
, sizeof(info
), &info
);
1310 static int netlink_sendmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1311 struct msghdr
*msg
, size_t len
)
1313 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1314 struct sock
*sk
= sock
->sk
;
1315 struct netlink_sock
*nlk
= nlk_sk(sk
);
1316 struct sockaddr_nl
*addr
= msg
->msg_name
;
1319 struct sk_buff
*skb
;
1321 struct scm_cookie scm
;
1323 if (msg
->msg_flags
&MSG_OOB
)
1326 if (NULL
== siocb
->scm
) {
1328 memset(&scm
, 0, sizeof(scm
));
1330 err
= scm_send(sock
, msg
, siocb
->scm
);
1334 if (msg
->msg_namelen
) {
1336 if (addr
->nl_family
!= AF_NETLINK
)
1338 dst_pid
= addr
->nl_pid
;
1339 dst_group
= ffs(addr
->nl_groups
);
1341 if (dst_group
&& !netlink_capable(sock
, NL_NONROOT_SEND
))
1344 dst_pid
= nlk
->dst_pid
;
1345 dst_group
= nlk
->dst_group
;
1349 err
= netlink_autobind(sock
);
1355 if (len
> sk
->sk_sndbuf
- 32)
1358 skb
= alloc_skb(len
, GFP_KERNEL
);
1362 NETLINK_CB(skb
).pid
= nlk
->pid
;
1363 NETLINK_CB(skb
).dst_group
= dst_group
;
1364 NETLINK_CB(skb
).loginuid
= audit_get_loginuid(current
);
1365 NETLINK_CB(skb
).sessionid
= audit_get_sessionid(current
);
1366 security_task_getsecid(current
, &(NETLINK_CB(skb
).sid
));
1367 memcpy(NETLINK_CREDS(skb
), &siocb
->scm
->creds
, sizeof(struct ucred
));
1369 /* What can I do? Netlink is asynchronous, so that
1370 we will have to save current capabilities to
1371 check them, when this message will be delivered
1372 to corresponding kernel module. --ANK (980802)
1376 if (memcpy_fromiovec(skb_put(skb
, len
), msg
->msg_iov
, len
)) {
1381 err
= security_netlink_send(sk
, skb
);
1388 atomic_inc(&skb
->users
);
1389 netlink_broadcast(sk
, skb
, dst_pid
, dst_group
, GFP_KERNEL
);
1391 err
= netlink_unicast(sk
, skb
, dst_pid
, msg
->msg_flags
&MSG_DONTWAIT
);
1394 scm_destroy(siocb
->scm
);
1398 static int netlink_recvmsg(struct kiocb
*kiocb
, struct socket
*sock
,
1399 struct msghdr
*msg
, size_t len
,
1402 struct sock_iocb
*siocb
= kiocb_to_siocb(kiocb
);
1403 struct scm_cookie scm
;
1404 struct sock
*sk
= sock
->sk
;
1405 struct netlink_sock
*nlk
= nlk_sk(sk
);
1406 int noblock
= flags
&MSG_DONTWAIT
;
1408 struct sk_buff
*skb
, *frag __maybe_unused
= NULL
;
1416 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
1420 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1421 if (unlikely(skb_shinfo(skb
)->frag_list
)) {
1422 bool need_compat
= !!(flags
& MSG_CMSG_COMPAT
);
1425 * If this skb has a frag_list, then here that means that
1426 * we will have to use the frag_list skb for compat tasks
1427 * and the regular skb for non-compat tasks.
1429 * The skb might (and likely will) be cloned, so we can't
1430 * just reset frag_list and go on with things -- we need to
1431 * keep that. For the compat case that's easy -- simply get
1432 * a reference to the compat skb and free the regular one
1433 * including the frag. For the non-compat case, we need to
1434 * avoid sending the frag to the user -- so assign NULL but
1435 * restore it below before freeing the skb.
1438 struct sk_buff
*compskb
= skb_shinfo(skb
)->frag_list
;
1443 frag
= skb_shinfo(skb
)->frag_list
;
1444 skb_shinfo(skb
)->frag_list
= NULL
;
1449 msg
->msg_namelen
= 0;
1453 msg
->msg_flags
|= MSG_TRUNC
;
1457 skb_reset_transport_header(skb
);
1458 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
1460 if (msg
->msg_name
) {
1461 struct sockaddr_nl
*addr
= (struct sockaddr_nl
*)msg
->msg_name
;
1462 addr
->nl_family
= AF_NETLINK
;
1464 addr
->nl_pid
= NETLINK_CB(skb
).pid
;
1465 addr
->nl_groups
= netlink_group_mask(NETLINK_CB(skb
).dst_group
);
1466 msg
->msg_namelen
= sizeof(*addr
);
1469 if (nlk
->flags
& NETLINK_RECV_PKTINFO
)
1470 netlink_cmsg_recv_pktinfo(msg
, skb
);
1472 if (NULL
== siocb
->scm
) {
1473 memset(&scm
, 0, sizeof(scm
));
1476 siocb
->scm
->creds
= *NETLINK_CREDS(skb
);
1477 if (flags
& MSG_TRUNC
)
1480 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1481 skb_shinfo(skb
)->frag_list
= frag
;
1484 skb_free_datagram(sk
, skb
);
1486 if (nlk
->cb
&& atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
/ 2)
1489 scm_recv(sock
, msg
, siocb
->scm
, flags
);
1491 netlink_rcv_wake(sk
);
1492 return err
? : copied
;
1495 static void netlink_data_ready(struct sock
*sk
, int len
)
1501 * We export these functions to other modules. They provide a
1502 * complete set of kernel non-blocking support for message
1507 netlink_kernel_create(struct net
*net
, int unit
, unsigned int groups
,
1508 void (*input
)(struct sk_buff
*skb
),
1509 struct mutex
*cb_mutex
, struct module
*module
)
1511 struct socket
*sock
;
1513 struct netlink_sock
*nlk
;
1514 unsigned long *listeners
= NULL
;
1518 if (unit
< 0 || unit
>= MAX_LINKS
)
1521 if (sock_create_lite(PF_NETLINK
, SOCK_DGRAM
, unit
, &sock
))
1525 * We have to just have a reference on the net from sk, but don't
1526 * get_net it. Besides, we cannot get and then put the net here.
1527 * So we create one inside init_net and the move it to net.
1530 if (__netlink_create(&init_net
, sock
, cb_mutex
, unit
) < 0)
1531 goto out_sock_release_nosk
;
1534 sk_change_net(sk
, net
);
1539 listeners
= kzalloc(NLGRPSZ(groups
) + sizeof(struct listeners_rcu_head
),
1542 goto out_sock_release
;
1544 sk
->sk_data_ready
= netlink_data_ready
;
1546 nlk_sk(sk
)->netlink_rcv
= input
;
1548 if (netlink_insert(sk
, net
, 0))
1549 goto out_sock_release
;
1552 nlk
->flags
|= NETLINK_KERNEL_SOCKET
;
1554 netlink_table_grab();
1555 if (!nl_table
[unit
].registered
) {
1556 nl_table
[unit
].groups
= groups
;
1557 nl_table
[unit
].listeners
= listeners
;
1558 nl_table
[unit
].cb_mutex
= cb_mutex
;
1559 nl_table
[unit
].module
= module
;
1560 nl_table
[unit
].registered
= 1;
1563 nl_table
[unit
].registered
++;
1565 netlink_table_ungrab();
1570 netlink_kernel_release(sk
);
1573 out_sock_release_nosk
:
1577 EXPORT_SYMBOL(netlink_kernel_create
);
1581 netlink_kernel_release(struct sock
*sk
)
1583 sk_release_kernel(sk
);
1585 EXPORT_SYMBOL(netlink_kernel_release
);
1588 static void netlink_free_old_listeners(struct rcu_head
*rcu_head
)
1590 struct listeners_rcu_head
*lrh
;
1592 lrh
= container_of(rcu_head
, struct listeners_rcu_head
, rcu_head
);
1596 int __netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
1598 unsigned long *listeners
, *old
= NULL
;
1599 struct listeners_rcu_head
*old_rcu_head
;
1600 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
1605 if (NLGRPSZ(tbl
->groups
) < NLGRPSZ(groups
)) {
1606 listeners
= kzalloc(NLGRPSZ(groups
) +
1607 sizeof(struct listeners_rcu_head
),
1611 old
= tbl
->listeners
;
1612 memcpy(listeners
, old
, NLGRPSZ(tbl
->groups
));
1613 rcu_assign_pointer(tbl
->listeners
, listeners
);
1615 * Free the old memory after an RCU grace period so we
1616 * don't leak it. We use call_rcu() here in order to be
1617 * able to call this function from atomic contexts. The
1618 * allocation of this memory will have reserved enough
1619 * space for struct listeners_rcu_head at the end.
1621 old_rcu_head
= (void *)(tbl
->listeners
+
1622 NLGRPLONGS(tbl
->groups
));
1623 old_rcu_head
->ptr
= old
;
1624 call_rcu(&old_rcu_head
->rcu_head
, netlink_free_old_listeners
);
1626 tbl
->groups
= groups
;
1632 * netlink_change_ngroups - change number of multicast groups
1634 * This changes the number of multicast groups that are available
1635 * on a certain netlink family. Note that it is not possible to
1636 * change the number of groups to below 32. Also note that it does
1637 * not implicitly call netlink_clear_multicast_users() when the
1638 * number of groups is reduced.
1640 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1641 * @groups: The new number of groups.
1643 int netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
1647 netlink_table_grab();
1648 err
= __netlink_change_ngroups(sk
, groups
);
1649 netlink_table_ungrab();
1654 void __netlink_clear_multicast_users(struct sock
*ksk
, unsigned int group
)
1657 struct hlist_node
*node
;
1658 struct netlink_table
*tbl
= &nl_table
[ksk
->sk_protocol
];
1660 sk_for_each_bound(sk
, node
, &tbl
->mc_list
)
1661 netlink_update_socket_mc(nlk_sk(sk
), group
, 0);
1665 * netlink_clear_multicast_users - kick off multicast listeners
1667 * This function removes all listeners from the given group.
1668 * @ksk: The kernel netlink socket, as returned by
1669 * netlink_kernel_create().
1670 * @group: The multicast group to clear.
1672 void netlink_clear_multicast_users(struct sock
*ksk
, unsigned int group
)
1674 netlink_table_grab();
1675 __netlink_clear_multicast_users(ksk
, group
);
1676 netlink_table_ungrab();
1679 void netlink_set_nonroot(int protocol
, unsigned int flags
)
1681 if ((unsigned int)protocol
< MAX_LINKS
)
1682 nl_table
[protocol
].nl_nonroot
= flags
;
1684 EXPORT_SYMBOL(netlink_set_nonroot
);
1686 static void netlink_destroy_callback(struct netlink_callback
*cb
)
1693 * It looks a bit ugly.
1694 * It would be better to create kernel thread.
1697 static int netlink_dump(struct sock
*sk
)
1699 struct netlink_sock
*nlk
= nlk_sk(sk
);
1700 struct netlink_callback
*cb
;
1701 struct sk_buff
*skb
;
1702 struct nlmsghdr
*nlh
;
1703 int len
, err
= -ENOBUFS
;
1705 skb
= sock_rmalloc(sk
, NLMSG_GOODSIZE
, 0, GFP_KERNEL
);
1709 mutex_lock(nlk
->cb_mutex
);
1717 len
= cb
->dump(skb
, cb
);
1720 mutex_unlock(nlk
->cb_mutex
);
1722 if (sk_filter(sk
, skb
))
1725 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1726 sk
->sk_data_ready(sk
, skb
->len
);
1731 nlh
= nlmsg_put_answer(skb
, cb
, NLMSG_DONE
, sizeof(len
), NLM_F_MULTI
);
1735 memcpy(nlmsg_data(nlh
), &len
, sizeof(len
));
1737 if (sk_filter(sk
, skb
))
1740 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1741 sk
->sk_data_ready(sk
, skb
->len
);
1747 mutex_unlock(nlk
->cb_mutex
);
1749 netlink_destroy_callback(cb
);
1753 mutex_unlock(nlk
->cb_mutex
);
1759 int netlink_dump_start(struct sock
*ssk
, struct sk_buff
*skb
,
1760 const struct nlmsghdr
*nlh
,
1761 int (*dump
)(struct sk_buff
*skb
,
1762 struct netlink_callback
*),
1763 int (*done
)(struct netlink_callback
*))
1765 struct netlink_callback
*cb
;
1767 struct netlink_sock
*nlk
;
1769 cb
= kzalloc(sizeof(*cb
), GFP_KERNEL
);
1776 atomic_inc(&skb
->users
);
1779 sk
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, NETLINK_CB(skb
).pid
);
1781 netlink_destroy_callback(cb
);
1782 return -ECONNREFUSED
;
1785 /* A dump is in progress... */
1786 mutex_lock(nlk
->cb_mutex
);
1788 mutex_unlock(nlk
->cb_mutex
);
1789 netlink_destroy_callback(cb
);
1794 mutex_unlock(nlk
->cb_mutex
);
1799 /* We successfully started a dump, by returning -EINTR we
1800 * signal not to send ACK even if it was requested.
1804 EXPORT_SYMBOL(netlink_dump_start
);
1806 void netlink_ack(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, int err
)
1808 struct sk_buff
*skb
;
1809 struct nlmsghdr
*rep
;
1810 struct nlmsgerr
*errmsg
;
1811 size_t payload
= sizeof(*errmsg
);
1813 /* error messages get the original request appened */
1815 payload
+= nlmsg_len(nlh
);
1817 skb
= nlmsg_new(payload
, GFP_KERNEL
);
1821 sk
= netlink_lookup(sock_net(in_skb
->sk
),
1822 in_skb
->sk
->sk_protocol
,
1823 NETLINK_CB(in_skb
).pid
);
1825 sk
->sk_err
= ENOBUFS
;
1826 sk
->sk_error_report(sk
);
1832 rep
= __nlmsg_put(skb
, NETLINK_CB(in_skb
).pid
, nlh
->nlmsg_seq
,
1833 NLMSG_ERROR
, payload
, 0);
1834 errmsg
= nlmsg_data(rep
);
1835 errmsg
->error
= err
;
1836 memcpy(&errmsg
->msg
, nlh
, err
? nlh
->nlmsg_len
: sizeof(*nlh
));
1837 netlink_unicast(in_skb
->sk
, skb
, NETLINK_CB(in_skb
).pid
, MSG_DONTWAIT
);
1839 EXPORT_SYMBOL(netlink_ack
);
1841 int netlink_rcv_skb(struct sk_buff
*skb
, int (*cb
)(struct sk_buff
*,
1844 struct nlmsghdr
*nlh
;
1847 while (skb
->len
>= nlmsg_total_size(0)) {
1850 nlh
= nlmsg_hdr(skb
);
1853 if (nlh
->nlmsg_len
< NLMSG_HDRLEN
|| skb
->len
< nlh
->nlmsg_len
)
1856 /* Only requests are handled by the kernel */
1857 if (!(nlh
->nlmsg_flags
& NLM_F_REQUEST
))
1860 /* Skip control messages */
1861 if (nlh
->nlmsg_type
< NLMSG_MIN_TYPE
)
1869 if (nlh
->nlmsg_flags
& NLM_F_ACK
|| err
)
1870 netlink_ack(skb
, nlh
, err
);
1873 msglen
= NLMSG_ALIGN(nlh
->nlmsg_len
);
1874 if (msglen
> skb
->len
)
1876 skb_pull(skb
, msglen
);
1881 EXPORT_SYMBOL(netlink_rcv_skb
);
1884 * nlmsg_notify - send a notification netlink message
1885 * @sk: netlink socket to use
1886 * @skb: notification message
1887 * @pid: destination netlink pid for reports or 0
1888 * @group: destination multicast group or 0
1889 * @report: 1 to report back, 0 to disable
1890 * @flags: allocation flags
1892 int nlmsg_notify(struct sock
*sk
, struct sk_buff
*skb
, u32 pid
,
1893 unsigned int group
, int report
, gfp_t flags
)
1898 int exclude_pid
= 0;
1901 atomic_inc(&skb
->users
);
1905 /* errors reported via destination sk->sk_err, but propagate
1906 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
1907 err
= nlmsg_multicast(sk
, skb
, exclude_pid
, group
, flags
);
1913 err2
= nlmsg_unicast(sk
, skb
, pid
);
1914 if (!err
|| err
== -ESRCH
)
1920 EXPORT_SYMBOL(nlmsg_notify
);
1922 #ifdef CONFIG_PROC_FS
1923 struct nl_seq_iter
{
1924 struct seq_net_private p
;
1929 static struct sock
*netlink_seq_socket_idx(struct seq_file
*seq
, loff_t pos
)
1931 struct nl_seq_iter
*iter
= seq
->private;
1934 struct hlist_node
*node
;
1937 for (i
= 0; i
< MAX_LINKS
; i
++) {
1938 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1940 for (j
= 0; j
<= hash
->mask
; j
++) {
1941 sk_for_each(s
, node
, &hash
->table
[j
]) {
1942 if (sock_net(s
) != seq_file_net(seq
))
1956 static void *netlink_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1957 __acquires(nl_table_lock
)
1959 read_lock(&nl_table_lock
);
1960 return *pos
? netlink_seq_socket_idx(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1963 static void *netlink_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1966 struct nl_seq_iter
*iter
;
1971 if (v
== SEQ_START_TOKEN
)
1972 return netlink_seq_socket_idx(seq
, 0);
1974 iter
= seq
->private;
1978 } while (s
&& sock_net(s
) != seq_file_net(seq
));
1983 j
= iter
->hash_idx
+ 1;
1986 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
1988 for (; j
<= hash
->mask
; j
++) {
1989 s
= sk_head(&hash
->table
[j
]);
1990 while (s
&& sock_net(s
) != seq_file_net(seq
))
2000 } while (++i
< MAX_LINKS
);
2005 static void netlink_seq_stop(struct seq_file
*seq
, void *v
)
2006 __releases(nl_table_lock
)
2008 read_unlock(&nl_table_lock
);
2012 static int netlink_seq_show(struct seq_file
*seq
, void *v
)
2014 if (v
== SEQ_START_TOKEN
)
2016 "sk Eth Pid Groups "
2017 "Rmem Wmem Dump Locks Drops Inode\n");
2020 struct netlink_sock
*nlk
= nlk_sk(s
);
2022 seq_printf(seq
, "%p %-3d %-6d %08x %-8d %-8d %p %-8d %-8d %-8lu\n",
2026 nlk
->groups
? (u32
)nlk
->groups
[0] : 0,
2027 sk_rmem_alloc_get(s
),
2028 sk_wmem_alloc_get(s
),
2030 atomic_read(&s
->sk_refcnt
),
2031 atomic_read(&s
->sk_drops
),
2039 static const struct seq_operations netlink_seq_ops
= {
2040 .start
= netlink_seq_start
,
2041 .next
= netlink_seq_next
,
2042 .stop
= netlink_seq_stop
,
2043 .show
= netlink_seq_show
,
2047 static int netlink_seq_open(struct inode
*inode
, struct file
*file
)
2049 return seq_open_net(inode
, file
, &netlink_seq_ops
,
2050 sizeof(struct nl_seq_iter
));
2053 static const struct file_operations netlink_seq_fops
= {
2054 .owner
= THIS_MODULE
,
2055 .open
= netlink_seq_open
,
2057 .llseek
= seq_lseek
,
2058 .release
= seq_release_net
,
2063 int netlink_register_notifier(struct notifier_block
*nb
)
2065 return atomic_notifier_chain_register(&netlink_chain
, nb
);
2067 EXPORT_SYMBOL(netlink_register_notifier
);
2069 int netlink_unregister_notifier(struct notifier_block
*nb
)
2071 return atomic_notifier_chain_unregister(&netlink_chain
, nb
);
2073 EXPORT_SYMBOL(netlink_unregister_notifier
);
2075 static const struct proto_ops netlink_ops
= {
2076 .family
= PF_NETLINK
,
2077 .owner
= THIS_MODULE
,
2078 .release
= netlink_release
,
2079 .bind
= netlink_bind
,
2080 .connect
= netlink_connect
,
2081 .socketpair
= sock_no_socketpair
,
2082 .accept
= sock_no_accept
,
2083 .getname
= netlink_getname
,
2084 .poll
= datagram_poll
,
2085 .ioctl
= sock_no_ioctl
,
2086 .listen
= sock_no_listen
,
2087 .shutdown
= sock_no_shutdown
,
2088 .setsockopt
= netlink_setsockopt
,
2089 .getsockopt
= netlink_getsockopt
,
2090 .sendmsg
= netlink_sendmsg
,
2091 .recvmsg
= netlink_recvmsg
,
2092 .mmap
= sock_no_mmap
,
2093 .sendpage
= sock_no_sendpage
,
2096 static const struct net_proto_family netlink_family_ops
= {
2097 .family
= PF_NETLINK
,
2098 .create
= netlink_create
,
2099 .owner
= THIS_MODULE
, /* for consistency 8) */
2102 static int __net_init
netlink_net_init(struct net
*net
)
2104 #ifdef CONFIG_PROC_FS
2105 if (!proc_net_fops_create(net
, "netlink", 0, &netlink_seq_fops
))
2111 static void __net_exit
netlink_net_exit(struct net
*net
)
2113 #ifdef CONFIG_PROC_FS
2114 proc_net_remove(net
, "netlink");
2118 static struct pernet_operations __net_initdata netlink_net_ops
= {
2119 .init
= netlink_net_init
,
2120 .exit
= netlink_net_exit
,
2123 static int __init
netlink_proto_init(void)
2125 struct sk_buff
*dummy_skb
;
2127 unsigned long limit
;
2129 int err
= proto_register(&netlink_proto
, 0);
2134 BUILD_BUG_ON(sizeof(struct netlink_skb_parms
) > sizeof(dummy_skb
->cb
));
2136 nl_table
= kcalloc(MAX_LINKS
, sizeof(*nl_table
), GFP_KERNEL
);
2140 if (totalram_pages
>= (128 * 1024))
2141 limit
= totalram_pages
>> (21 - PAGE_SHIFT
);
2143 limit
= totalram_pages
>> (23 - PAGE_SHIFT
);
2145 order
= get_bitmask_order(limit
) - 1 + PAGE_SHIFT
;
2146 limit
= (1UL << order
) / sizeof(struct hlist_head
);
2147 order
= get_bitmask_order(min(limit
, (unsigned long)UINT_MAX
)) - 1;
2149 for (i
= 0; i
< MAX_LINKS
; i
++) {
2150 struct nl_pid_hash
*hash
= &nl_table
[i
].hash
;
2152 hash
->table
= nl_pid_hash_zalloc(1 * sizeof(*hash
->table
));
2155 nl_pid_hash_free(nl_table
[i
].hash
.table
,
2156 1 * sizeof(*hash
->table
));
2160 hash
->max_shift
= order
;
2163 hash
->rehash_time
= jiffies
;
2166 sock_register(&netlink_family_ops
);
2167 register_pernet_subsys(&netlink_net_ops
);
2168 /* The netlink device handler may be needed early. */
2173 panic("netlink_init: Cannot allocate nl_table\n");
2176 core_initcall(netlink_proto_init
);