2 * NETLINK Kernel-user communication protocol.
4 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
5 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
6 * Patrick McHardy <kaber@trash.net>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
13 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
14 * added netlink_proto_exit
15 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
16 * use nlk_sk, as sk->protinfo is on a diet 8)
17 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
18 * - inc module use count of module that owns
19 * the kernel socket in case userspace opens
20 * socket of same protocol
21 * - remove all module support, since netlink is
22 * mandatory if CONFIG_NET=y these days
25 #include <linux/module.h>
27 #include <linux/capability.h>
28 #include <linux/kernel.h>
29 #include <linux/init.h>
30 #include <linux/signal.h>
31 #include <linux/sched.h>
32 #include <linux/errno.h>
33 #include <linux/string.h>
34 #include <linux/stat.h>
35 #include <linux/socket.h>
37 #include <linux/fcntl.h>
38 #include <linux/termios.h>
39 #include <linux/sockios.h>
40 #include <linux/net.h>
42 #include <linux/slab.h>
43 #include <linux/uaccess.h>
44 #include <linux/skbuff.h>
45 #include <linux/netdevice.h>
46 #include <linux/rtnetlink.h>
47 #include <linux/proc_fs.h>
48 #include <linux/seq_file.h>
49 #include <linux/notifier.h>
50 #include <linux/security.h>
51 #include <linux/jhash.h>
52 #include <linux/jiffies.h>
53 #include <linux/random.h>
54 #include <linux/bitops.h>
56 #include <linux/types.h>
57 #include <linux/audit.h>
58 #include <linux/mutex.h>
59 #include <linux/vmalloc.h>
60 #include <linux/if_arp.h>
61 #include <linux/rhashtable.h>
62 #include <asm/cacheflush.h>
63 #include <linux/hash.h>
64 #include <linux/genetlink.h>
65 #include <linux/net_namespace.h>
67 #include <net/net_namespace.h>
70 #include <net/netlink.h>
72 #include "af_netlink.h"
76 unsigned long masks
[0];
80 #define NETLINK_S_CONGESTED 0x0
82 static inline int netlink_is_kernel(struct sock
*sk
)
84 return nlk_sk(sk
)->flags
& NETLINK_F_KERNEL_SOCKET
;
87 struct netlink_table
*nl_table __read_mostly
;
88 EXPORT_SYMBOL_GPL(nl_table
);
90 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait
);
92 static struct lock_class_key nlk_cb_mutex_keys
[MAX_LINKS
];
94 static const char *const nlk_cb_mutex_key_strings
[MAX_LINKS
+ 1] = {
97 "nlk_cb_mutex-USERSOCK",
98 "nlk_cb_mutex-FIREWALL",
99 "nlk_cb_mutex-SOCK_DIAG",
100 "nlk_cb_mutex-NFLOG",
102 "nlk_cb_mutex-SELINUX",
103 "nlk_cb_mutex-ISCSI",
104 "nlk_cb_mutex-AUDIT",
105 "nlk_cb_mutex-FIB_LOOKUP",
106 "nlk_cb_mutex-CONNECTOR",
107 "nlk_cb_mutex-NETFILTER",
108 "nlk_cb_mutex-IP6_FW",
109 "nlk_cb_mutex-DNRTMSG",
110 "nlk_cb_mutex-KOBJECT_UEVENT",
111 "nlk_cb_mutex-GENERIC",
113 "nlk_cb_mutex-SCSITRANSPORT",
114 "nlk_cb_mutex-ECRYPTFS",
116 "nlk_cb_mutex-CRYPTO",
127 "nlk_cb_mutex-MAX_LINKS"
130 static int netlink_dump(struct sock
*sk
);
132 /* nl_table locking explained:
133 * Lookup and traversal are protected with an RCU read-side lock. Insertion
134 * and removal are protected with per bucket lock while using RCU list
135 * modification primitives and may run in parallel to RCU protected lookups.
136 * Destruction of the Netlink socket may only occur *after* nl_table_lock has
137 * been acquired * either during or after the socket has been removed from
138 * the list and after an RCU grace period.
140 DEFINE_RWLOCK(nl_table_lock
);
141 EXPORT_SYMBOL_GPL(nl_table_lock
);
142 static atomic_t nl_table_users
= ATOMIC_INIT(0);
144 #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
146 static BLOCKING_NOTIFIER_HEAD(netlink_chain
);
148 static DEFINE_SPINLOCK(netlink_tap_lock
);
149 static struct list_head netlink_tap_all __read_mostly
;
151 static const struct rhashtable_params netlink_rhashtable_params
;
153 static inline u32
netlink_group_mask(u32 group
)
155 return group
? 1 << (group
- 1) : 0;
158 static struct sk_buff
*netlink_to_full_skb(const struct sk_buff
*skb
,
161 unsigned int len
= skb_end_offset(skb
);
164 new = alloc_skb(len
, gfp_mask
);
168 NETLINK_CB(new).portid
= NETLINK_CB(skb
).portid
;
169 NETLINK_CB(new).dst_group
= NETLINK_CB(skb
).dst_group
;
170 NETLINK_CB(new).creds
= NETLINK_CB(skb
).creds
;
172 skb_put_data(new, skb
->data
, len
);
176 int netlink_add_tap(struct netlink_tap
*nt
)
178 if (unlikely(nt
->dev
->type
!= ARPHRD_NETLINK
))
181 spin_lock(&netlink_tap_lock
);
182 list_add_rcu(&nt
->list
, &netlink_tap_all
);
183 spin_unlock(&netlink_tap_lock
);
185 __module_get(nt
->module
);
189 EXPORT_SYMBOL_GPL(netlink_add_tap
);
191 static int __netlink_remove_tap(struct netlink_tap
*nt
)
194 struct netlink_tap
*tmp
;
196 spin_lock(&netlink_tap_lock
);
198 list_for_each_entry(tmp
, &netlink_tap_all
, list
) {
200 list_del_rcu(&nt
->list
);
206 pr_warn("__netlink_remove_tap: %p not found\n", nt
);
208 spin_unlock(&netlink_tap_lock
);
211 module_put(nt
->module
);
213 return found
? 0 : -ENODEV
;
216 int netlink_remove_tap(struct netlink_tap
*nt
)
220 ret
= __netlink_remove_tap(nt
);
225 EXPORT_SYMBOL_GPL(netlink_remove_tap
);
227 static bool netlink_filter_tap(const struct sk_buff
*skb
)
229 struct sock
*sk
= skb
->sk
;
231 /* We take the more conservative approach and
232 * whitelist socket protocols that may pass.
234 switch (sk
->sk_protocol
) {
236 case NETLINK_USERSOCK
:
237 case NETLINK_SOCK_DIAG
:
240 case NETLINK_FIB_LOOKUP
:
241 case NETLINK_NETFILTER
:
242 case NETLINK_GENERIC
:
249 static int __netlink_deliver_tap_skb(struct sk_buff
*skb
,
250 struct net_device
*dev
)
252 struct sk_buff
*nskb
;
253 struct sock
*sk
= skb
->sk
;
256 if (!net_eq(dev_net(dev
), sock_net(sk
)))
261 if (is_vmalloc_addr(skb
->head
))
262 nskb
= netlink_to_full_skb(skb
, GFP_ATOMIC
);
264 nskb
= skb_clone(skb
, GFP_ATOMIC
);
267 nskb
->protocol
= htons((u16
) sk
->sk_protocol
);
268 nskb
->pkt_type
= netlink_is_kernel(sk
) ?
269 PACKET_KERNEL
: PACKET_USER
;
270 skb_reset_network_header(nskb
);
271 ret
= dev_queue_xmit(nskb
);
272 if (unlikely(ret
> 0))
273 ret
= net_xmit_errno(ret
);
280 static void __netlink_deliver_tap(struct sk_buff
*skb
)
283 struct netlink_tap
*tmp
;
285 if (!netlink_filter_tap(skb
))
288 list_for_each_entry_rcu(tmp
, &netlink_tap_all
, list
) {
289 ret
= __netlink_deliver_tap_skb(skb
, tmp
->dev
);
295 static void netlink_deliver_tap(struct sk_buff
*skb
)
299 if (unlikely(!list_empty(&netlink_tap_all
)))
300 __netlink_deliver_tap(skb
);
305 static void netlink_deliver_tap_kernel(struct sock
*dst
, struct sock
*src
,
308 if (!(netlink_is_kernel(dst
) && netlink_is_kernel(src
)))
309 netlink_deliver_tap(skb
);
312 static void netlink_overrun(struct sock
*sk
)
314 struct netlink_sock
*nlk
= nlk_sk(sk
);
316 if (!(nlk
->flags
& NETLINK_F_RECV_NO_ENOBUFS
)) {
317 if (!test_and_set_bit(NETLINK_S_CONGESTED
,
318 &nlk_sk(sk
)->state
)) {
319 sk
->sk_err
= ENOBUFS
;
320 sk
->sk_error_report(sk
);
323 atomic_inc(&sk
->sk_drops
);
326 static void netlink_rcv_wake(struct sock
*sk
)
328 struct netlink_sock
*nlk
= nlk_sk(sk
);
330 if (skb_queue_empty(&sk
->sk_receive_queue
))
331 clear_bit(NETLINK_S_CONGESTED
, &nlk
->state
);
332 if (!test_bit(NETLINK_S_CONGESTED
, &nlk
->state
))
333 wake_up_interruptible(&nlk
->wait
);
336 static void netlink_skb_destructor(struct sk_buff
*skb
)
338 if (is_vmalloc_addr(skb
->head
)) {
340 !atomic_dec_return(&(skb_shinfo(skb
)->dataref
)))
349 static void netlink_skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
351 WARN_ON(skb
->sk
!= NULL
);
353 skb
->destructor
= netlink_skb_destructor
;
354 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
355 sk_mem_charge(sk
, skb
->truesize
);
358 static void netlink_sock_destruct(struct sock
*sk
)
360 struct netlink_sock
*nlk
= nlk_sk(sk
);
362 if (nlk
->cb_running
) {
364 nlk
->cb
.done(&nlk
->cb
);
365 module_put(nlk
->cb
.module
);
366 kfree_skb(nlk
->cb
.skb
);
369 skb_queue_purge(&sk
->sk_receive_queue
);
371 if (!sock_flag(sk
, SOCK_DEAD
)) {
372 printk(KERN_ERR
"Freeing alive netlink socket %p\n", sk
);
376 WARN_ON(atomic_read(&sk
->sk_rmem_alloc
));
377 WARN_ON(refcount_read(&sk
->sk_wmem_alloc
));
378 WARN_ON(nlk_sk(sk
)->groups
);
381 static void netlink_sock_destruct_work(struct work_struct
*work
)
383 struct netlink_sock
*nlk
= container_of(work
, struct netlink_sock
,
389 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
390 * SMP. Look, when several writers sleep and reader wakes them up, all but one
391 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
392 * this, _but_ remember, it adds useless work on UP machines.
395 void netlink_table_grab(void)
396 __acquires(nl_table_lock
)
400 write_lock_irq(&nl_table_lock
);
402 if (atomic_read(&nl_table_users
)) {
403 DECLARE_WAITQUEUE(wait
, current
);
405 add_wait_queue_exclusive(&nl_table_wait
, &wait
);
407 set_current_state(TASK_UNINTERRUPTIBLE
);
408 if (atomic_read(&nl_table_users
) == 0)
410 write_unlock_irq(&nl_table_lock
);
412 write_lock_irq(&nl_table_lock
);
415 __set_current_state(TASK_RUNNING
);
416 remove_wait_queue(&nl_table_wait
, &wait
);
420 void netlink_table_ungrab(void)
421 __releases(nl_table_lock
)
423 write_unlock_irq(&nl_table_lock
);
424 wake_up(&nl_table_wait
);
428 netlink_lock_table(void)
430 /* read_lock() synchronizes us to netlink_table_grab */
432 read_lock(&nl_table_lock
);
433 atomic_inc(&nl_table_users
);
434 read_unlock(&nl_table_lock
);
438 netlink_unlock_table(void)
440 if (atomic_dec_and_test(&nl_table_users
))
441 wake_up(&nl_table_wait
);
444 struct netlink_compare_arg
450 /* Doing sizeof directly may yield 4 extra bytes on 64-bit. */
451 #define netlink_compare_arg_len \
452 (offsetof(struct netlink_compare_arg, portid) + sizeof(u32))
454 static inline int netlink_compare(struct rhashtable_compare_arg
*arg
,
457 const struct netlink_compare_arg
*x
= arg
->key
;
458 const struct netlink_sock
*nlk
= ptr
;
460 return nlk
->portid
!= x
->portid
||
461 !net_eq(sock_net(&nlk
->sk
), read_pnet(&x
->pnet
));
464 static void netlink_compare_arg_init(struct netlink_compare_arg
*arg
,
465 struct net
*net
, u32 portid
)
467 memset(arg
, 0, sizeof(*arg
));
468 write_pnet(&arg
->pnet
, net
);
469 arg
->portid
= portid
;
472 static struct sock
*__netlink_lookup(struct netlink_table
*table
, u32 portid
,
475 struct netlink_compare_arg arg
;
477 netlink_compare_arg_init(&arg
, net
, portid
);
478 return rhashtable_lookup_fast(&table
->hash
, &arg
,
479 netlink_rhashtable_params
);
482 static int __netlink_insert(struct netlink_table
*table
, struct sock
*sk
)
484 struct netlink_compare_arg arg
;
486 netlink_compare_arg_init(&arg
, sock_net(sk
), nlk_sk(sk
)->portid
);
487 return rhashtable_lookup_insert_key(&table
->hash
, &arg
,
489 netlink_rhashtable_params
);
492 static struct sock
*netlink_lookup(struct net
*net
, int protocol
, u32 portid
)
494 struct netlink_table
*table
= &nl_table
[protocol
];
498 sk
= __netlink_lookup(table
, portid
, net
);
506 static const struct proto_ops netlink_ops
;
509 netlink_update_listeners(struct sock
*sk
)
511 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
514 struct listeners
*listeners
;
516 listeners
= nl_deref_protected(tbl
->listeners
);
520 for (i
= 0; i
< NLGRPLONGS(tbl
->groups
); i
++) {
522 sk_for_each_bound(sk
, &tbl
->mc_list
) {
523 if (i
< NLGRPLONGS(nlk_sk(sk
)->ngroups
))
524 mask
|= nlk_sk(sk
)->groups
[i
];
526 listeners
->masks
[i
] = mask
;
528 /* this function is only called with the netlink table "grabbed", which
529 * makes sure updates are visible before bind or setsockopt return. */
532 static int netlink_insert(struct sock
*sk
, u32 portid
)
534 struct netlink_table
*table
= &nl_table
[sk
->sk_protocol
];
539 err
= nlk_sk(sk
)->portid
== portid
? 0 : -EBUSY
;
540 if (nlk_sk(sk
)->bound
)
544 if (BITS_PER_LONG
> 32 &&
545 unlikely(atomic_read(&table
->hash
.nelems
) >= UINT_MAX
))
548 nlk_sk(sk
)->portid
= portid
;
551 err
= __netlink_insert(table
, sk
);
553 /* In case the hashtable backend returns with -EBUSY
554 * from here, it must not escape to the caller.
556 if (unlikely(err
== -EBUSY
))
564 /* We need to ensure that the socket is hashed and visible. */
566 nlk_sk(sk
)->bound
= portid
;
573 static void netlink_remove(struct sock
*sk
)
575 struct netlink_table
*table
;
577 table
= &nl_table
[sk
->sk_protocol
];
578 if (!rhashtable_remove_fast(&table
->hash
, &nlk_sk(sk
)->node
,
579 netlink_rhashtable_params
)) {
580 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
584 netlink_table_grab();
585 if (nlk_sk(sk
)->subscriptions
) {
586 __sk_del_bind_node(sk
);
587 netlink_update_listeners(sk
);
589 if (sk
->sk_protocol
== NETLINK_GENERIC
)
590 atomic_inc(&genl_sk_destructing_cnt
);
591 netlink_table_ungrab();
594 static struct proto netlink_proto
= {
596 .owner
= THIS_MODULE
,
597 .obj_size
= sizeof(struct netlink_sock
),
600 static int __netlink_create(struct net
*net
, struct socket
*sock
,
601 struct mutex
*cb_mutex
, int protocol
,
605 struct netlink_sock
*nlk
;
607 sock
->ops
= &netlink_ops
;
609 sk
= sk_alloc(net
, PF_NETLINK
, GFP_KERNEL
, &netlink_proto
, kern
);
613 sock_init_data(sock
, sk
);
617 nlk
->cb_mutex
= cb_mutex
;
619 nlk
->cb_mutex
= &nlk
->cb_def_mutex
;
620 mutex_init(nlk
->cb_mutex
);
621 lockdep_set_class_and_name(nlk
->cb_mutex
,
622 nlk_cb_mutex_keys
+ protocol
,
623 nlk_cb_mutex_key_strings
[protocol
]);
625 init_waitqueue_head(&nlk
->wait
);
627 sk
->sk_destruct
= netlink_sock_destruct
;
628 sk
->sk_protocol
= protocol
;
632 static int netlink_create(struct net
*net
, struct socket
*sock
, int protocol
,
635 struct module
*module
= NULL
;
636 struct mutex
*cb_mutex
;
637 struct netlink_sock
*nlk
;
638 int (*bind
)(struct net
*net
, int group
);
639 void (*unbind
)(struct net
*net
, int group
);
642 sock
->state
= SS_UNCONNECTED
;
644 if (sock
->type
!= SOCK_RAW
&& sock
->type
!= SOCK_DGRAM
)
645 return -ESOCKTNOSUPPORT
;
647 if (protocol
< 0 || protocol
>= MAX_LINKS
)
648 return -EPROTONOSUPPORT
;
650 netlink_lock_table();
651 #ifdef CONFIG_MODULES
652 if (!nl_table
[protocol
].registered
) {
653 netlink_unlock_table();
654 request_module("net-pf-%d-proto-%d", PF_NETLINK
, protocol
);
655 netlink_lock_table();
658 if (nl_table
[protocol
].registered
&&
659 try_module_get(nl_table
[protocol
].module
))
660 module
= nl_table
[protocol
].module
;
662 err
= -EPROTONOSUPPORT
;
663 cb_mutex
= nl_table
[protocol
].cb_mutex
;
664 bind
= nl_table
[protocol
].bind
;
665 unbind
= nl_table
[protocol
].unbind
;
666 netlink_unlock_table();
671 err
= __netlink_create(net
, sock
, cb_mutex
, protocol
, kern
);
676 sock_prot_inuse_add(net
, &netlink_proto
, 1);
679 nlk
= nlk_sk(sock
->sk
);
680 nlk
->module
= module
;
681 nlk
->netlink_bind
= bind
;
682 nlk
->netlink_unbind
= unbind
;
691 static void deferred_put_nlk_sk(struct rcu_head
*head
)
693 struct netlink_sock
*nlk
= container_of(head
, struct netlink_sock
, rcu
);
694 struct sock
*sk
= &nlk
->sk
;
699 if (!refcount_dec_and_test(&sk
->sk_refcnt
))
702 if (nlk
->cb_running
&& nlk
->cb
.done
) {
703 INIT_WORK(&nlk
->work
, netlink_sock_destruct_work
);
704 schedule_work(&nlk
->work
);
711 static int netlink_release(struct socket
*sock
)
713 struct sock
*sk
= sock
->sk
;
714 struct netlink_sock
*nlk
;
724 * OK. Socket is unlinked, any packets that arrive now
728 /* must not acquire netlink_table_lock in any way again before unbind
729 * and notifying genetlink is done as otherwise it might deadlock
731 if (nlk
->netlink_unbind
) {
734 for (i
= 0; i
< nlk
->ngroups
; i
++)
735 if (test_bit(i
, nlk
->groups
))
736 nlk
->netlink_unbind(sock_net(sk
), i
+ 1);
738 if (sk
->sk_protocol
== NETLINK_GENERIC
&&
739 atomic_dec_return(&genl_sk_destructing_cnt
) == 0)
740 wake_up(&genl_sk_destructing_waitq
);
743 wake_up_interruptible_all(&nlk
->wait
);
745 skb_queue_purge(&sk
->sk_write_queue
);
747 if (nlk
->portid
&& nlk
->bound
) {
748 struct netlink_notify n
= {
750 .protocol
= sk
->sk_protocol
,
751 .portid
= nlk
->portid
,
753 blocking_notifier_call_chain(&netlink_chain
,
754 NETLINK_URELEASE
, &n
);
757 module_put(nlk
->module
);
759 if (netlink_is_kernel(sk
)) {
760 netlink_table_grab();
761 BUG_ON(nl_table
[sk
->sk_protocol
].registered
== 0);
762 if (--nl_table
[sk
->sk_protocol
].registered
== 0) {
763 struct listeners
*old
;
765 old
= nl_deref_protected(nl_table
[sk
->sk_protocol
].listeners
);
766 RCU_INIT_POINTER(nl_table
[sk
->sk_protocol
].listeners
, NULL
);
768 nl_table
[sk
->sk_protocol
].module
= NULL
;
769 nl_table
[sk
->sk_protocol
].bind
= NULL
;
770 nl_table
[sk
->sk_protocol
].unbind
= NULL
;
771 nl_table
[sk
->sk_protocol
].flags
= 0;
772 nl_table
[sk
->sk_protocol
].registered
= 0;
774 netlink_table_ungrab();
778 sock_prot_inuse_add(sock_net(sk
), &netlink_proto
, -1);
780 call_rcu(&nlk
->rcu
, deferred_put_nlk_sk
);
784 static int netlink_autobind(struct socket
*sock
)
786 struct sock
*sk
= sock
->sk
;
787 struct net
*net
= sock_net(sk
);
788 struct netlink_table
*table
= &nl_table
[sk
->sk_protocol
];
789 s32 portid
= task_tgid_vnr(current
);
797 ok
= !__netlink_lookup(table
, portid
, net
);
800 /* Bind collision, search negative portid values. */
802 /* rover will be in range [S32_MIN, -4097] */
803 rover
= S32_MIN
+ prandom_u32_max(-4096 - S32_MIN
);
804 else if (rover
>= -4096)
810 err
= netlink_insert(sk
, portid
);
811 if (err
== -EADDRINUSE
)
814 /* If 2 threads race to autobind, that is fine. */
822 * __netlink_ns_capable - General netlink message capability test
823 * @nsp: NETLINK_CB of the socket buffer holding a netlink command from userspace.
824 * @user_ns: The user namespace of the capability to use
825 * @cap: The capability to use
827 * Test to see if the opener of the socket we received the message
828 * from had when the netlink socket was created and the sender of the
829 * message has has the capability @cap in the user namespace @user_ns.
831 bool __netlink_ns_capable(const struct netlink_skb_parms
*nsp
,
832 struct user_namespace
*user_ns
, int cap
)
834 return ((nsp
->flags
& NETLINK_SKB_DST
) ||
835 file_ns_capable(nsp
->sk
->sk_socket
->file
, user_ns
, cap
)) &&
836 ns_capable(user_ns
, cap
);
838 EXPORT_SYMBOL(__netlink_ns_capable
);
841 * netlink_ns_capable - General netlink message capability test
842 * @skb: socket buffer holding a netlink command from userspace
843 * @user_ns: The user namespace of the capability to use
844 * @cap: The capability to use
846 * Test to see if the opener of the socket we received the message
847 * from had when the netlink socket was created and the sender of the
848 * message has has the capability @cap in the user namespace @user_ns.
850 bool netlink_ns_capable(const struct sk_buff
*skb
,
851 struct user_namespace
*user_ns
, int cap
)
853 return __netlink_ns_capable(&NETLINK_CB(skb
), user_ns
, cap
);
855 EXPORT_SYMBOL(netlink_ns_capable
);
858 * netlink_capable - Netlink global message capability test
859 * @skb: socket buffer holding a netlink command from userspace
860 * @cap: The capability to use
862 * Test to see if the opener of the socket we received the message
863 * from had when the netlink socket was created and the sender of the
864 * message has has the capability @cap in all user namespaces.
866 bool netlink_capable(const struct sk_buff
*skb
, int cap
)
868 return netlink_ns_capable(skb
, &init_user_ns
, cap
);
870 EXPORT_SYMBOL(netlink_capable
);
873 * netlink_net_capable - Netlink network namespace message capability test
874 * @skb: socket buffer holding a netlink command from userspace
875 * @cap: The capability to use
877 * Test to see if the opener of the socket we received the message
878 * from had when the netlink socket was created and the sender of the
879 * message has has the capability @cap over the network namespace of
880 * the socket we received the message from.
882 bool netlink_net_capable(const struct sk_buff
*skb
, int cap
)
884 return netlink_ns_capable(skb
, sock_net(skb
->sk
)->user_ns
, cap
);
886 EXPORT_SYMBOL(netlink_net_capable
);
888 static inline int netlink_allowed(const struct socket
*sock
, unsigned int flag
)
890 return (nl_table
[sock
->sk
->sk_protocol
].flags
& flag
) ||
891 ns_capable(sock_net(sock
->sk
)->user_ns
, CAP_NET_ADMIN
);
895 netlink_update_subscriptions(struct sock
*sk
, unsigned int subscriptions
)
897 struct netlink_sock
*nlk
= nlk_sk(sk
);
899 if (nlk
->subscriptions
&& !subscriptions
)
900 __sk_del_bind_node(sk
);
901 else if (!nlk
->subscriptions
&& subscriptions
)
902 sk_add_bind_node(sk
, &nl_table
[sk
->sk_protocol
].mc_list
);
903 nlk
->subscriptions
= subscriptions
;
906 static int netlink_realloc_groups(struct sock
*sk
)
908 struct netlink_sock
*nlk
= nlk_sk(sk
);
910 unsigned long *new_groups
;
913 netlink_table_grab();
915 groups
= nl_table
[sk
->sk_protocol
].groups
;
916 if (!nl_table
[sk
->sk_protocol
].registered
) {
921 if (nlk
->ngroups
>= groups
)
924 new_groups
= krealloc(nlk
->groups
, NLGRPSZ(groups
), GFP_ATOMIC
);
925 if (new_groups
== NULL
) {
929 memset((char *)new_groups
+ NLGRPSZ(nlk
->ngroups
), 0,
930 NLGRPSZ(groups
) - NLGRPSZ(nlk
->ngroups
));
932 nlk
->groups
= new_groups
;
933 nlk
->ngroups
= groups
;
935 netlink_table_ungrab();
939 static void netlink_undo_bind(int group
, long unsigned int groups
,
942 struct netlink_sock
*nlk
= nlk_sk(sk
);
945 if (!nlk
->netlink_unbind
)
948 for (undo
= 0; undo
< group
; undo
++)
949 if (test_bit(undo
, &groups
))
950 nlk
->netlink_unbind(sock_net(sk
), undo
+ 1);
953 static int netlink_bind(struct socket
*sock
, struct sockaddr
*addr
,
956 struct sock
*sk
= sock
->sk
;
957 struct net
*net
= sock_net(sk
);
958 struct netlink_sock
*nlk
= nlk_sk(sk
);
959 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
961 long unsigned int groups
= nladdr
->nl_groups
;
964 if (addr_len
< sizeof(struct sockaddr_nl
))
967 if (nladdr
->nl_family
!= AF_NETLINK
)
970 /* Only superuser is allowed to listen multicasts */
972 if (!netlink_allowed(sock
, NL_CFG_F_NONROOT_RECV
))
974 err
= netlink_realloc_groups(sk
);
981 /* Ensure nlk->portid is up-to-date. */
984 if (nladdr
->nl_pid
!= nlk
->portid
)
988 netlink_lock_table();
989 if (nlk
->netlink_bind
&& groups
) {
992 for (group
= 0; group
< nlk
->ngroups
; group
++) {
993 if (!test_bit(group
, &groups
))
995 err
= nlk
->netlink_bind(net
, group
+ 1);
998 netlink_undo_bind(group
, groups
, sk
);
1003 /* No need for barriers here as we return to user-space without
1004 * using any of the bound attributes.
1007 err
= nladdr
->nl_pid
?
1008 netlink_insert(sk
, nladdr
->nl_pid
) :
1009 netlink_autobind(sock
);
1011 netlink_undo_bind(nlk
->ngroups
, groups
, sk
);
1016 if (!groups
&& (nlk
->groups
== NULL
|| !(u32
)nlk
->groups
[0]))
1018 netlink_unlock_table();
1020 netlink_table_grab();
1021 netlink_update_subscriptions(sk
, nlk
->subscriptions
+
1023 hweight32(nlk
->groups
[0]));
1024 nlk
->groups
[0] = (nlk
->groups
[0] & ~0xffffffffUL
) | groups
;
1025 netlink_update_listeners(sk
);
1026 netlink_table_ungrab();
1031 netlink_unlock_table();
1035 static int netlink_connect(struct socket
*sock
, struct sockaddr
*addr
,
1036 int alen
, int flags
)
1039 struct sock
*sk
= sock
->sk
;
1040 struct netlink_sock
*nlk
= nlk_sk(sk
);
1041 struct sockaddr_nl
*nladdr
= (struct sockaddr_nl
*)addr
;
1043 if (alen
< sizeof(addr
->sa_family
))
1046 if (addr
->sa_family
== AF_UNSPEC
) {
1047 sk
->sk_state
= NETLINK_UNCONNECTED
;
1048 nlk
->dst_portid
= 0;
1052 if (addr
->sa_family
!= AF_NETLINK
)
1055 if (alen
< sizeof(struct sockaddr_nl
))
1058 if ((nladdr
->nl_groups
|| nladdr
->nl_pid
) &&
1059 !netlink_allowed(sock
, NL_CFG_F_NONROOT_SEND
))
1062 /* No need for barriers here as we return to user-space without
1063 * using any of the bound attributes.
1066 err
= netlink_autobind(sock
);
1069 sk
->sk_state
= NETLINK_CONNECTED
;
1070 nlk
->dst_portid
= nladdr
->nl_pid
;
1071 nlk
->dst_group
= ffs(nladdr
->nl_groups
);
1077 static int netlink_getname(struct socket
*sock
, struct sockaddr
*addr
,
1078 int *addr_len
, int peer
)
1080 struct sock
*sk
= sock
->sk
;
1081 struct netlink_sock
*nlk
= nlk_sk(sk
);
1082 DECLARE_SOCKADDR(struct sockaddr_nl
*, nladdr
, addr
);
1084 nladdr
->nl_family
= AF_NETLINK
;
1086 *addr_len
= sizeof(*nladdr
);
1089 nladdr
->nl_pid
= nlk
->dst_portid
;
1090 nladdr
->nl_groups
= netlink_group_mask(nlk
->dst_group
);
1092 nladdr
->nl_pid
= nlk
->portid
;
1093 netlink_lock_table();
1094 nladdr
->nl_groups
= nlk
->groups
? nlk
->groups
[0] : 0;
1095 netlink_unlock_table();
1100 static int netlink_ioctl(struct socket
*sock
, unsigned int cmd
,
1103 /* try to hand this ioctl down to the NIC drivers.
1105 return -ENOIOCTLCMD
;
1108 static struct sock
*netlink_getsockbyportid(struct sock
*ssk
, u32 portid
)
1111 struct netlink_sock
*nlk
;
1113 sock
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, portid
);
1115 return ERR_PTR(-ECONNREFUSED
);
1117 /* Don't bother queuing skb if kernel socket has no input function */
1119 if (sock
->sk_state
== NETLINK_CONNECTED
&&
1120 nlk
->dst_portid
!= nlk_sk(ssk
)->portid
) {
1122 return ERR_PTR(-ECONNREFUSED
);
1127 struct sock
*netlink_getsockbyfilp(struct file
*filp
)
1129 struct inode
*inode
= file_inode(filp
);
1132 if (!S_ISSOCK(inode
->i_mode
))
1133 return ERR_PTR(-ENOTSOCK
);
1135 sock
= SOCKET_I(inode
)->sk
;
1136 if (sock
->sk_family
!= AF_NETLINK
)
1137 return ERR_PTR(-EINVAL
);
1143 static struct sk_buff
*netlink_alloc_large_skb(unsigned int size
,
1146 struct sk_buff
*skb
;
1149 if (size
<= NLMSG_GOODSIZE
|| broadcast
)
1150 return alloc_skb(size
, GFP_KERNEL
);
1152 size
= SKB_DATA_ALIGN(size
) +
1153 SKB_DATA_ALIGN(sizeof(struct skb_shared_info
));
1155 data
= vmalloc(size
);
1159 skb
= __build_skb(data
, size
);
1163 skb
->destructor
= netlink_skb_destructor
;
1169 * Attach a skb to a netlink socket.
1170 * The caller must hold a reference to the destination socket. On error, the
1171 * reference is dropped. The skb is not send to the destination, just all
1172 * all error checks are performed and memory in the queue is reserved.
1174 * < 0: error. skb freed, reference to sock dropped.
1176 * 1: repeat lookup - reference dropped while waiting for socket memory.
1178 int netlink_attachskb(struct sock
*sk
, struct sk_buff
*skb
,
1179 long *timeo
, struct sock
*ssk
)
1181 struct netlink_sock
*nlk
;
1185 if ((atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
1186 test_bit(NETLINK_S_CONGESTED
, &nlk
->state
))) {
1187 DECLARE_WAITQUEUE(wait
, current
);
1189 if (!ssk
|| netlink_is_kernel(ssk
))
1190 netlink_overrun(sk
);
1196 __set_current_state(TASK_INTERRUPTIBLE
);
1197 add_wait_queue(&nlk
->wait
, &wait
);
1199 if ((atomic_read(&sk
->sk_rmem_alloc
) > sk
->sk_rcvbuf
||
1200 test_bit(NETLINK_S_CONGESTED
, &nlk
->state
)) &&
1201 !sock_flag(sk
, SOCK_DEAD
))
1202 *timeo
= schedule_timeout(*timeo
);
1204 __set_current_state(TASK_RUNNING
);
1205 remove_wait_queue(&nlk
->wait
, &wait
);
1208 if (signal_pending(current
)) {
1210 return sock_intr_errno(*timeo
);
1214 netlink_skb_set_owner_r(skb
, sk
);
1218 static int __netlink_sendskb(struct sock
*sk
, struct sk_buff
*skb
)
1222 netlink_deliver_tap(skb
);
1224 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1225 sk
->sk_data_ready(sk
);
1229 int netlink_sendskb(struct sock
*sk
, struct sk_buff
*skb
)
1231 int len
= __netlink_sendskb(sk
, skb
);
1237 void netlink_detachskb(struct sock
*sk
, struct sk_buff
*skb
)
1243 static struct sk_buff
*netlink_trim(struct sk_buff
*skb
, gfp_t allocation
)
1247 WARN_ON(skb
->sk
!= NULL
);
1248 delta
= skb
->end
- skb
->tail
;
1249 if (is_vmalloc_addr(skb
->head
) || delta
* 2 < skb
->truesize
)
1252 if (skb_shared(skb
)) {
1253 struct sk_buff
*nskb
= skb_clone(skb
, allocation
);
1260 pskb_expand_head(skb
, 0, -delta
,
1261 (allocation
& ~__GFP_DIRECT_RECLAIM
) |
1262 __GFP_NOWARN
| __GFP_NORETRY
);
1266 static int netlink_unicast_kernel(struct sock
*sk
, struct sk_buff
*skb
,
1270 struct netlink_sock
*nlk
= nlk_sk(sk
);
1272 ret
= -ECONNREFUSED
;
1273 if (nlk
->netlink_rcv
!= NULL
) {
1275 netlink_skb_set_owner_r(skb
, sk
);
1276 NETLINK_CB(skb
).sk
= ssk
;
1277 netlink_deliver_tap_kernel(sk
, ssk
, skb
);
1278 nlk
->netlink_rcv(skb
);
1287 int netlink_unicast(struct sock
*ssk
, struct sk_buff
*skb
,
1288 u32 portid
, int nonblock
)
1294 skb
= netlink_trim(skb
, gfp_any());
1296 timeo
= sock_sndtimeo(ssk
, nonblock
);
1298 sk
= netlink_getsockbyportid(ssk
, portid
);
1303 if (netlink_is_kernel(sk
))
1304 return netlink_unicast_kernel(sk
, skb
, ssk
);
1306 if (sk_filter(sk
, skb
)) {
1313 err
= netlink_attachskb(sk
, skb
, &timeo
, ssk
);
1319 return netlink_sendskb(sk
, skb
);
1321 EXPORT_SYMBOL(netlink_unicast
);
1323 int netlink_has_listeners(struct sock
*sk
, unsigned int group
)
1326 struct listeners
*listeners
;
1328 BUG_ON(!netlink_is_kernel(sk
));
1331 listeners
= rcu_dereference(nl_table
[sk
->sk_protocol
].listeners
);
1333 if (listeners
&& group
- 1 < nl_table
[sk
->sk_protocol
].groups
)
1334 res
= test_bit(group
- 1, listeners
->masks
);
1340 EXPORT_SYMBOL_GPL(netlink_has_listeners
);
1342 static int netlink_broadcast_deliver(struct sock
*sk
, struct sk_buff
*skb
)
1344 struct netlink_sock
*nlk
= nlk_sk(sk
);
1346 if (atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
&&
1347 !test_bit(NETLINK_S_CONGESTED
, &nlk
->state
)) {
1348 netlink_skb_set_owner_r(skb
, sk
);
1349 __netlink_sendskb(sk
, skb
);
1350 return atomic_read(&sk
->sk_rmem_alloc
) > (sk
->sk_rcvbuf
>> 1);
1355 struct netlink_broadcast_data
{
1356 struct sock
*exclude_sk
;
1361 int delivery_failure
;
1365 struct sk_buff
*skb
, *skb2
;
1366 int (*tx_filter
)(struct sock
*dsk
, struct sk_buff
*skb
, void *data
);
1370 static void do_one_broadcast(struct sock
*sk
,
1371 struct netlink_broadcast_data
*p
)
1373 struct netlink_sock
*nlk
= nlk_sk(sk
);
1376 if (p
->exclude_sk
== sk
)
1379 if (nlk
->portid
== p
->portid
|| p
->group
- 1 >= nlk
->ngroups
||
1380 !test_bit(p
->group
- 1, nlk
->groups
))
1383 if (!net_eq(sock_net(sk
), p
->net
)) {
1384 if (!(nlk
->flags
& NETLINK_F_LISTEN_ALL_NSID
))
1387 if (!peernet_has_id(sock_net(sk
), p
->net
))
1390 if (!file_ns_capable(sk
->sk_socket
->file
, p
->net
->user_ns
,
1396 netlink_overrun(sk
);
1401 if (p
->skb2
== NULL
) {
1402 if (skb_shared(p
->skb
)) {
1403 p
->skb2
= skb_clone(p
->skb
, p
->allocation
);
1405 p
->skb2
= skb_get(p
->skb
);
1407 * skb ownership may have been set when
1408 * delivered to a previous socket.
1410 skb_orphan(p
->skb2
);
1413 if (p
->skb2
== NULL
) {
1414 netlink_overrun(sk
);
1415 /* Clone failed. Notify ALL listeners. */
1417 if (nlk
->flags
& NETLINK_F_BROADCAST_SEND_ERROR
)
1418 p
->delivery_failure
= 1;
1421 if (p
->tx_filter
&& p
->tx_filter(sk
, p
->skb2
, p
->tx_data
)) {
1426 if (sk_filter(sk
, p
->skb2
)) {
1431 NETLINK_CB(p
->skb2
).nsid
= peernet2id(sock_net(sk
), p
->net
);
1432 if (NETLINK_CB(p
->skb2
).nsid
!= NETNSA_NSID_NOT_ASSIGNED
)
1433 NETLINK_CB(p
->skb2
).nsid_is_set
= true;
1434 val
= netlink_broadcast_deliver(sk
, p
->skb2
);
1436 netlink_overrun(sk
);
1437 if (nlk
->flags
& NETLINK_F_BROADCAST_SEND_ERROR
)
1438 p
->delivery_failure
= 1;
1440 p
->congested
|= val
;
1448 int netlink_broadcast_filtered(struct sock
*ssk
, struct sk_buff
*skb
, u32 portid
,
1449 u32 group
, gfp_t allocation
,
1450 int (*filter
)(struct sock
*dsk
, struct sk_buff
*skb
, void *data
),
1453 struct net
*net
= sock_net(ssk
);
1454 struct netlink_broadcast_data info
;
1457 skb
= netlink_trim(skb
, allocation
);
1459 info
.exclude_sk
= ssk
;
1461 info
.portid
= portid
;
1464 info
.delivery_failure
= 0;
1467 info
.allocation
= allocation
;
1470 info
.tx_filter
= filter
;
1471 info
.tx_data
= filter_data
;
1473 /* While we sleep in clone, do not allow to change socket list */
1475 netlink_lock_table();
1477 sk_for_each_bound(sk
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1478 do_one_broadcast(sk
, &info
);
1482 netlink_unlock_table();
1484 if (info
.delivery_failure
) {
1485 kfree_skb(info
.skb2
);
1488 consume_skb(info
.skb2
);
1490 if (info
.delivered
) {
1491 if (info
.congested
&& gfpflags_allow_blocking(allocation
))
1497 EXPORT_SYMBOL(netlink_broadcast_filtered
);
1499 int netlink_broadcast(struct sock
*ssk
, struct sk_buff
*skb
, u32 portid
,
1500 u32 group
, gfp_t allocation
)
1502 return netlink_broadcast_filtered(ssk
, skb
, portid
, group
, allocation
,
1505 EXPORT_SYMBOL(netlink_broadcast
);
1507 struct netlink_set_err_data
{
1508 struct sock
*exclude_sk
;
1514 static int do_one_set_err(struct sock
*sk
, struct netlink_set_err_data
*p
)
1516 struct netlink_sock
*nlk
= nlk_sk(sk
);
1519 if (sk
== p
->exclude_sk
)
1522 if (!net_eq(sock_net(sk
), sock_net(p
->exclude_sk
)))
1525 if (nlk
->portid
== p
->portid
|| p
->group
- 1 >= nlk
->ngroups
||
1526 !test_bit(p
->group
- 1, nlk
->groups
))
1529 if (p
->code
== ENOBUFS
&& nlk
->flags
& NETLINK_F_RECV_NO_ENOBUFS
) {
1534 sk
->sk_err
= p
->code
;
1535 sk
->sk_error_report(sk
);
1541 * netlink_set_err - report error to broadcast listeners
1542 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1543 * @portid: the PORTID of a process that we want to skip (if any)
1544 * @group: the broadcast group that will notice the error
1545 * @code: error code, must be negative (as usual in kernelspace)
1547 * This function returns the number of broadcast listeners that have set the
1548 * NETLINK_NO_ENOBUFS socket option.
1550 int netlink_set_err(struct sock
*ssk
, u32 portid
, u32 group
, int code
)
1552 struct netlink_set_err_data info
;
1556 info
.exclude_sk
= ssk
;
1557 info
.portid
= portid
;
1559 /* sk->sk_err wants a positive error value */
1562 read_lock(&nl_table_lock
);
1564 sk_for_each_bound(sk
, &nl_table
[ssk
->sk_protocol
].mc_list
)
1565 ret
+= do_one_set_err(sk
, &info
);
1567 read_unlock(&nl_table_lock
);
1570 EXPORT_SYMBOL(netlink_set_err
);
1572 /* must be called with netlink table grabbed */
1573 static void netlink_update_socket_mc(struct netlink_sock
*nlk
,
1577 int old
, new = !!is_new
, subscriptions
;
1579 old
= test_bit(group
- 1, nlk
->groups
);
1580 subscriptions
= nlk
->subscriptions
- old
+ new;
1582 __set_bit(group
- 1, nlk
->groups
);
1584 __clear_bit(group
- 1, nlk
->groups
);
1585 netlink_update_subscriptions(&nlk
->sk
, subscriptions
);
1586 netlink_update_listeners(&nlk
->sk
);
1589 static int netlink_setsockopt(struct socket
*sock
, int level
, int optname
,
1590 char __user
*optval
, unsigned int optlen
)
1592 struct sock
*sk
= sock
->sk
;
1593 struct netlink_sock
*nlk
= nlk_sk(sk
);
1594 unsigned int val
= 0;
1597 if (level
!= SOL_NETLINK
)
1598 return -ENOPROTOOPT
;
1600 if (optlen
>= sizeof(int) &&
1601 get_user(val
, (unsigned int __user
*)optval
))
1605 case NETLINK_PKTINFO
:
1607 nlk
->flags
|= NETLINK_F_RECV_PKTINFO
;
1609 nlk
->flags
&= ~NETLINK_F_RECV_PKTINFO
;
1612 case NETLINK_ADD_MEMBERSHIP
:
1613 case NETLINK_DROP_MEMBERSHIP
: {
1614 if (!netlink_allowed(sock
, NL_CFG_F_NONROOT_RECV
))
1616 err
= netlink_realloc_groups(sk
);
1619 if (!val
|| val
- 1 >= nlk
->ngroups
)
1621 if (optname
== NETLINK_ADD_MEMBERSHIP
&& nlk
->netlink_bind
) {
1622 err
= nlk
->netlink_bind(sock_net(sk
), val
);
1626 netlink_table_grab();
1627 netlink_update_socket_mc(nlk
, val
,
1628 optname
== NETLINK_ADD_MEMBERSHIP
);
1629 netlink_table_ungrab();
1630 if (optname
== NETLINK_DROP_MEMBERSHIP
&& nlk
->netlink_unbind
)
1631 nlk
->netlink_unbind(sock_net(sk
), val
);
1636 case NETLINK_BROADCAST_ERROR
:
1638 nlk
->flags
|= NETLINK_F_BROADCAST_SEND_ERROR
;
1640 nlk
->flags
&= ~NETLINK_F_BROADCAST_SEND_ERROR
;
1643 case NETLINK_NO_ENOBUFS
:
1645 nlk
->flags
|= NETLINK_F_RECV_NO_ENOBUFS
;
1646 clear_bit(NETLINK_S_CONGESTED
, &nlk
->state
);
1647 wake_up_interruptible(&nlk
->wait
);
1649 nlk
->flags
&= ~NETLINK_F_RECV_NO_ENOBUFS
;
1653 case NETLINK_LISTEN_ALL_NSID
:
1654 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_BROADCAST
))
1658 nlk
->flags
|= NETLINK_F_LISTEN_ALL_NSID
;
1660 nlk
->flags
&= ~NETLINK_F_LISTEN_ALL_NSID
;
1663 case NETLINK_CAP_ACK
:
1665 nlk
->flags
|= NETLINK_F_CAP_ACK
;
1667 nlk
->flags
&= ~NETLINK_F_CAP_ACK
;
1670 case NETLINK_EXT_ACK
:
1672 nlk
->flags
|= NETLINK_F_EXT_ACK
;
1674 nlk
->flags
&= ~NETLINK_F_EXT_ACK
;
1683 static int netlink_getsockopt(struct socket
*sock
, int level
, int optname
,
1684 char __user
*optval
, int __user
*optlen
)
1686 struct sock
*sk
= sock
->sk
;
1687 struct netlink_sock
*nlk
= nlk_sk(sk
);
1690 if (level
!= SOL_NETLINK
)
1691 return -ENOPROTOOPT
;
1693 if (get_user(len
, optlen
))
1699 case NETLINK_PKTINFO
:
1700 if (len
< sizeof(int))
1703 val
= nlk
->flags
& NETLINK_F_RECV_PKTINFO
? 1 : 0;
1704 if (put_user(len
, optlen
) ||
1705 put_user(val
, optval
))
1709 case NETLINK_BROADCAST_ERROR
:
1710 if (len
< sizeof(int))
1713 val
= nlk
->flags
& NETLINK_F_BROADCAST_SEND_ERROR
? 1 : 0;
1714 if (put_user(len
, optlen
) ||
1715 put_user(val
, optval
))
1719 case NETLINK_NO_ENOBUFS
:
1720 if (len
< sizeof(int))
1723 val
= nlk
->flags
& NETLINK_F_RECV_NO_ENOBUFS
? 1 : 0;
1724 if (put_user(len
, optlen
) ||
1725 put_user(val
, optval
))
1729 case NETLINK_LIST_MEMBERSHIPS
: {
1730 int pos
, idx
, shift
;
1733 netlink_lock_table();
1734 for (pos
= 0; pos
* 8 < nlk
->ngroups
; pos
+= sizeof(u32
)) {
1735 if (len
- pos
< sizeof(u32
))
1738 idx
= pos
/ sizeof(unsigned long);
1739 shift
= (pos
% sizeof(unsigned long)) * 8;
1740 if (put_user((u32
)(nlk
->groups
[idx
] >> shift
),
1741 (u32 __user
*)(optval
+ pos
))) {
1746 if (put_user(ALIGN(nlk
->ngroups
/ 8, sizeof(u32
)), optlen
))
1748 netlink_unlock_table();
1751 case NETLINK_CAP_ACK
:
1752 if (len
< sizeof(int))
1755 val
= nlk
->flags
& NETLINK_F_CAP_ACK
? 1 : 0;
1756 if (put_user(len
, optlen
) ||
1757 put_user(val
, optval
))
1761 case NETLINK_EXT_ACK
:
1762 if (len
< sizeof(int))
1765 val
= nlk
->flags
& NETLINK_F_EXT_ACK
? 1 : 0;
1766 if (put_user(len
, optlen
) || put_user(val
, optval
))
1776 static void netlink_cmsg_recv_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
1778 struct nl_pktinfo info
;
1780 info
.group
= NETLINK_CB(skb
).dst_group
;
1781 put_cmsg(msg
, SOL_NETLINK
, NETLINK_PKTINFO
, sizeof(info
), &info
);
1784 static void netlink_cmsg_listen_all_nsid(struct sock
*sk
, struct msghdr
*msg
,
1785 struct sk_buff
*skb
)
1787 if (!NETLINK_CB(skb
).nsid_is_set
)
1790 put_cmsg(msg
, SOL_NETLINK
, NETLINK_LISTEN_ALL_NSID
, sizeof(int),
1791 &NETLINK_CB(skb
).nsid
);
1794 static int netlink_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t len
)
1796 struct sock
*sk
= sock
->sk
;
1797 struct netlink_sock
*nlk
= nlk_sk(sk
);
1798 DECLARE_SOCKADDR(struct sockaddr_nl
*, addr
, msg
->msg_name
);
1801 struct sk_buff
*skb
;
1803 struct scm_cookie scm
;
1804 u32 netlink_skb_flags
= 0;
1806 if (msg
->msg_flags
&MSG_OOB
)
1809 err
= scm_send(sock
, msg
, &scm
, true);
1813 if (msg
->msg_namelen
) {
1815 if (msg
->msg_namelen
< sizeof(struct sockaddr_nl
))
1817 if (addr
->nl_family
!= AF_NETLINK
)
1819 dst_portid
= addr
->nl_pid
;
1820 dst_group
= ffs(addr
->nl_groups
);
1822 if ((dst_group
|| dst_portid
) &&
1823 !netlink_allowed(sock
, NL_CFG_F_NONROOT_SEND
))
1825 netlink_skb_flags
|= NETLINK_SKB_DST
;
1827 dst_portid
= nlk
->dst_portid
;
1828 dst_group
= nlk
->dst_group
;
1832 err
= netlink_autobind(sock
);
1836 /* Ensure nlk is hashed and visible. */
1841 if (len
> sk
->sk_sndbuf
- 32)
1844 skb
= netlink_alloc_large_skb(len
, dst_group
);
1848 NETLINK_CB(skb
).portid
= nlk
->portid
;
1849 NETLINK_CB(skb
).dst_group
= dst_group
;
1850 NETLINK_CB(skb
).creds
= scm
.creds
;
1851 NETLINK_CB(skb
).flags
= netlink_skb_flags
;
1854 if (memcpy_from_msg(skb_put(skb
, len
), msg
, len
)) {
1859 err
= security_netlink_send(sk
, skb
);
1866 refcount_inc(&skb
->users
);
1867 netlink_broadcast(sk
, skb
, dst_portid
, dst_group
, GFP_KERNEL
);
1869 err
= netlink_unicast(sk
, skb
, dst_portid
, msg
->msg_flags
&MSG_DONTWAIT
);
1876 static int netlink_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t len
,
1879 struct scm_cookie scm
;
1880 struct sock
*sk
= sock
->sk
;
1881 struct netlink_sock
*nlk
= nlk_sk(sk
);
1882 int noblock
= flags
&MSG_DONTWAIT
;
1884 struct sk_buff
*skb
, *data_skb
;
1892 skb
= skb_recv_datagram(sk
, flags
, noblock
, &err
);
1898 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1899 if (unlikely(skb_shinfo(skb
)->frag_list
)) {
1901 * If this skb has a frag_list, then here that means that we
1902 * will have to use the frag_list skb's data for compat tasks
1903 * and the regular skb's data for normal (non-compat) tasks.
1905 * If we need to send the compat skb, assign it to the
1906 * 'data_skb' variable so that it will be used below for data
1907 * copying. We keep 'skb' for everything else, including
1908 * freeing both later.
1910 if (flags
& MSG_CMSG_COMPAT
)
1911 data_skb
= skb_shinfo(skb
)->frag_list
;
1915 /* Record the max length of recvmsg() calls for future allocations */
1916 nlk
->max_recvmsg_len
= max(nlk
->max_recvmsg_len
, len
);
1917 nlk
->max_recvmsg_len
= min_t(size_t, nlk
->max_recvmsg_len
,
1918 SKB_WITH_OVERHEAD(32768));
1920 copied
= data_skb
->len
;
1922 msg
->msg_flags
|= MSG_TRUNC
;
1926 skb_reset_transport_header(data_skb
);
1927 err
= skb_copy_datagram_msg(data_skb
, 0, msg
, copied
);
1929 if (msg
->msg_name
) {
1930 DECLARE_SOCKADDR(struct sockaddr_nl
*, addr
, msg
->msg_name
);
1931 addr
->nl_family
= AF_NETLINK
;
1933 addr
->nl_pid
= NETLINK_CB(skb
).portid
;
1934 addr
->nl_groups
= netlink_group_mask(NETLINK_CB(skb
).dst_group
);
1935 msg
->msg_namelen
= sizeof(*addr
);
1938 if (nlk
->flags
& NETLINK_F_RECV_PKTINFO
)
1939 netlink_cmsg_recv_pktinfo(msg
, skb
);
1940 if (nlk
->flags
& NETLINK_F_LISTEN_ALL_NSID
)
1941 netlink_cmsg_listen_all_nsid(sk
, msg
, skb
);
1943 memset(&scm
, 0, sizeof(scm
));
1944 scm
.creds
= *NETLINK_CREDS(skb
);
1945 if (flags
& MSG_TRUNC
)
1946 copied
= data_skb
->len
;
1948 skb_free_datagram(sk
, skb
);
1950 if (nlk
->cb_running
&&
1951 atomic_read(&sk
->sk_rmem_alloc
) <= sk
->sk_rcvbuf
/ 2) {
1952 ret
= netlink_dump(sk
);
1955 sk
->sk_error_report(sk
);
1959 scm_recv(sock
, msg
, &scm
, flags
);
1961 netlink_rcv_wake(sk
);
1962 return err
? : copied
;
1965 static void netlink_data_ready(struct sock
*sk
)
1971 * We export these functions to other modules. They provide a
1972 * complete set of kernel non-blocking support for message
1977 __netlink_kernel_create(struct net
*net
, int unit
, struct module
*module
,
1978 struct netlink_kernel_cfg
*cfg
)
1980 struct socket
*sock
;
1982 struct netlink_sock
*nlk
;
1983 struct listeners
*listeners
= NULL
;
1984 struct mutex
*cb_mutex
= cfg
? cfg
->cb_mutex
: NULL
;
1985 unsigned int groups
;
1989 if (unit
< 0 || unit
>= MAX_LINKS
)
1992 if (sock_create_lite(PF_NETLINK
, SOCK_DGRAM
, unit
, &sock
))
1995 if (__netlink_create(net
, sock
, cb_mutex
, unit
, 1) < 0)
1996 goto out_sock_release_nosk
;
2000 if (!cfg
|| cfg
->groups
< 32)
2003 groups
= cfg
->groups
;
2005 listeners
= kzalloc(sizeof(*listeners
) + NLGRPSZ(groups
), GFP_KERNEL
);
2007 goto out_sock_release
;
2009 sk
->sk_data_ready
= netlink_data_ready
;
2010 if (cfg
&& cfg
->input
)
2011 nlk_sk(sk
)->netlink_rcv
= cfg
->input
;
2013 if (netlink_insert(sk
, 0))
2014 goto out_sock_release
;
2017 nlk
->flags
|= NETLINK_F_KERNEL_SOCKET
;
2019 netlink_table_grab();
2020 if (!nl_table
[unit
].registered
) {
2021 nl_table
[unit
].groups
= groups
;
2022 rcu_assign_pointer(nl_table
[unit
].listeners
, listeners
);
2023 nl_table
[unit
].cb_mutex
= cb_mutex
;
2024 nl_table
[unit
].module
= module
;
2026 nl_table
[unit
].bind
= cfg
->bind
;
2027 nl_table
[unit
].unbind
= cfg
->unbind
;
2028 nl_table
[unit
].flags
= cfg
->flags
;
2030 nl_table
[unit
].compare
= cfg
->compare
;
2032 nl_table
[unit
].registered
= 1;
2035 nl_table
[unit
].registered
++;
2037 netlink_table_ungrab();
2042 netlink_kernel_release(sk
);
2045 out_sock_release_nosk
:
2049 EXPORT_SYMBOL(__netlink_kernel_create
);
2052 netlink_kernel_release(struct sock
*sk
)
2054 if (sk
== NULL
|| sk
->sk_socket
== NULL
)
2057 sock_release(sk
->sk_socket
);
2059 EXPORT_SYMBOL(netlink_kernel_release
);
2061 int __netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
2063 struct listeners
*new, *old
;
2064 struct netlink_table
*tbl
= &nl_table
[sk
->sk_protocol
];
2069 if (NLGRPSZ(tbl
->groups
) < NLGRPSZ(groups
)) {
2070 new = kzalloc(sizeof(*new) + NLGRPSZ(groups
), GFP_ATOMIC
);
2073 old
= nl_deref_protected(tbl
->listeners
);
2074 memcpy(new->masks
, old
->masks
, NLGRPSZ(tbl
->groups
));
2075 rcu_assign_pointer(tbl
->listeners
, new);
2077 kfree_rcu(old
, rcu
);
2079 tbl
->groups
= groups
;
2085 * netlink_change_ngroups - change number of multicast groups
2087 * This changes the number of multicast groups that are available
2088 * on a certain netlink family. Note that it is not possible to
2089 * change the number of groups to below 32. Also note that it does
2090 * not implicitly call netlink_clear_multicast_users() when the
2091 * number of groups is reduced.
2093 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
2094 * @groups: The new number of groups.
2096 int netlink_change_ngroups(struct sock
*sk
, unsigned int groups
)
2100 netlink_table_grab();
2101 err
= __netlink_change_ngroups(sk
, groups
);
2102 netlink_table_ungrab();
2107 void __netlink_clear_multicast_users(struct sock
*ksk
, unsigned int group
)
2110 struct netlink_table
*tbl
= &nl_table
[ksk
->sk_protocol
];
2112 sk_for_each_bound(sk
, &tbl
->mc_list
)
2113 netlink_update_socket_mc(nlk_sk(sk
), group
, 0);
2117 __nlmsg_put(struct sk_buff
*skb
, u32 portid
, u32 seq
, int type
, int len
, int flags
)
2119 struct nlmsghdr
*nlh
;
2120 int size
= nlmsg_msg_size(len
);
2122 nlh
= skb_put(skb
, NLMSG_ALIGN(size
));
2123 nlh
->nlmsg_type
= type
;
2124 nlh
->nlmsg_len
= size
;
2125 nlh
->nlmsg_flags
= flags
;
2126 nlh
->nlmsg_pid
= portid
;
2127 nlh
->nlmsg_seq
= seq
;
2128 if (!__builtin_constant_p(size
) || NLMSG_ALIGN(size
) - size
!= 0)
2129 memset(nlmsg_data(nlh
) + len
, 0, NLMSG_ALIGN(size
) - size
);
2132 EXPORT_SYMBOL(__nlmsg_put
);
2135 * It looks a bit ugly.
2136 * It would be better to create kernel thread.
2139 static int netlink_dump(struct sock
*sk
)
2141 struct netlink_sock
*nlk
= nlk_sk(sk
);
2142 struct netlink_callback
*cb
;
2143 struct sk_buff
*skb
= NULL
;
2144 struct nlmsghdr
*nlh
;
2145 struct module
*module
;
2150 mutex_lock(nlk
->cb_mutex
);
2151 if (!nlk
->cb_running
) {
2156 if (atomic_read(&sk
->sk_rmem_alloc
) >= sk
->sk_rcvbuf
)
2159 /* NLMSG_GOODSIZE is small to avoid high order allocations being
2160 * required, but it makes sense to _attempt_ a 16K bytes allocation
2161 * to reduce number of system calls on dump operations, if user
2162 * ever provided a big enough buffer.
2165 alloc_min_size
= max_t(int, cb
->min_dump_alloc
, NLMSG_GOODSIZE
);
2167 if (alloc_min_size
< nlk
->max_recvmsg_len
) {
2168 alloc_size
= nlk
->max_recvmsg_len
;
2169 skb
= alloc_skb(alloc_size
,
2170 (GFP_KERNEL
& ~__GFP_DIRECT_RECLAIM
) |
2171 __GFP_NOWARN
| __GFP_NORETRY
);
2174 alloc_size
= alloc_min_size
;
2175 skb
= alloc_skb(alloc_size
, GFP_KERNEL
);
2180 /* Trim skb to allocated size. User is expected to provide buffer as
2181 * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at
2182 * netlink_recvmsg())). dump will pack as many smaller messages as
2183 * could fit within the allocated skb. skb is typically allocated
2184 * with larger space than required (could be as much as near 2x the
2185 * requested size with align to next power of 2 approach). Allowing
2186 * dump to use the excess space makes it difficult for a user to have a
2187 * reasonable static buffer based on the expected largest dump of a
2188 * single netdev. The outcome is MSG_TRUNC error.
2190 skb_reserve(skb
, skb_tailroom(skb
) - alloc_size
);
2191 netlink_skb_set_owner_r(skb
, sk
);
2193 if (nlk
->dump_done_errno
> 0)
2194 nlk
->dump_done_errno
= cb
->dump(skb
, cb
);
2196 if (nlk
->dump_done_errno
> 0 ||
2197 skb_tailroom(skb
) < nlmsg_total_size(sizeof(nlk
->dump_done_errno
))) {
2198 mutex_unlock(nlk
->cb_mutex
);
2200 if (sk_filter(sk
, skb
))
2203 __netlink_sendskb(sk
, skb
);
2207 nlh
= nlmsg_put_answer(skb
, cb
, NLMSG_DONE
,
2208 sizeof(nlk
->dump_done_errno
), NLM_F_MULTI
);
2212 nl_dump_check_consistent(cb
, nlh
);
2214 memcpy(nlmsg_data(nlh
), &nlk
->dump_done_errno
,
2215 sizeof(nlk
->dump_done_errno
));
2217 if (sk_filter(sk
, skb
))
2220 __netlink_sendskb(sk
, skb
);
2225 nlk
->cb_running
= false;
2226 module
= cb
->module
;
2228 mutex_unlock(nlk
->cb_mutex
);
2234 mutex_unlock(nlk
->cb_mutex
);
2239 int __netlink_dump_start(struct sock
*ssk
, struct sk_buff
*skb
,
2240 const struct nlmsghdr
*nlh
,
2241 struct netlink_dump_control
*control
)
2243 struct netlink_callback
*cb
;
2245 struct netlink_sock
*nlk
;
2248 refcount_inc(&skb
->users
);
2250 sk
= netlink_lookup(sock_net(ssk
), ssk
->sk_protocol
, NETLINK_CB(skb
).portid
);
2252 ret
= -ECONNREFUSED
;
2257 mutex_lock(nlk
->cb_mutex
);
2258 /* A dump is in progress... */
2259 if (nlk
->cb_running
) {
2263 /* add reference of module which cb->dump belongs to */
2264 if (!try_module_get(control
->module
)) {
2265 ret
= -EPROTONOSUPPORT
;
2270 memset(cb
, 0, sizeof(*cb
));
2271 cb
->start
= control
->start
;
2272 cb
->dump
= control
->dump
;
2273 cb
->done
= control
->done
;
2275 cb
->data
= control
->data
;
2276 cb
->module
= control
->module
;
2277 cb
->min_dump_alloc
= control
->min_dump_alloc
;
2281 ret
= cb
->start(cb
);
2286 nlk
->cb_running
= true;
2287 nlk
->dump_done_errno
= INT_MAX
;
2289 mutex_unlock(nlk
->cb_mutex
);
2291 ret
= netlink_dump(sk
);
2298 /* We successfully started a dump, by returning -EINTR we
2299 * signal not to send ACK even if it was requested.
2304 module_put(control
->module
);
2307 mutex_unlock(nlk
->cb_mutex
);
2312 EXPORT_SYMBOL(__netlink_dump_start
);
2314 void netlink_ack(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
, int err
,
2315 const struct netlink_ext_ack
*extack
)
2317 struct sk_buff
*skb
;
2318 struct nlmsghdr
*rep
;
2319 struct nlmsgerr
*errmsg
;
2320 size_t payload
= sizeof(*errmsg
);
2322 struct netlink_sock
*nlk
= nlk_sk(NETLINK_CB(in_skb
).sk
);
2323 unsigned int flags
= 0;
2324 bool nlk_has_extack
= nlk
->flags
& NETLINK_F_EXT_ACK
;
2326 /* Error messages get the original request appened, unless the user
2327 * requests to cap the error message, and get extra error data if
2330 if (nlk_has_extack
&& extack
&& extack
->_msg
)
2331 tlvlen
+= nla_total_size(strlen(extack
->_msg
) + 1);
2334 if (!(nlk
->flags
& NETLINK_F_CAP_ACK
))
2335 payload
+= nlmsg_len(nlh
);
2337 flags
|= NLM_F_CAPPED
;
2338 if (nlk_has_extack
&& extack
&& extack
->bad_attr
)
2339 tlvlen
+= nla_total_size(sizeof(u32
));
2341 flags
|= NLM_F_CAPPED
;
2343 if (nlk_has_extack
&& extack
&& extack
->cookie_len
)
2344 tlvlen
+= nla_total_size(extack
->cookie_len
);
2348 flags
|= NLM_F_ACK_TLVS
;
2350 skb
= nlmsg_new(payload
+ tlvlen
, GFP_KERNEL
);
2352 NETLINK_CB(in_skb
).sk
->sk_err
= ENOBUFS
;
2353 NETLINK_CB(in_skb
).sk
->sk_error_report(NETLINK_CB(in_skb
).sk
);
2357 rep
= __nlmsg_put(skb
, NETLINK_CB(in_skb
).portid
, nlh
->nlmsg_seq
,
2358 NLMSG_ERROR
, payload
, flags
);
2359 errmsg
= nlmsg_data(rep
);
2360 errmsg
->error
= err
;
2361 memcpy(&errmsg
->msg
, nlh
, payload
> sizeof(*errmsg
) ? nlh
->nlmsg_len
: sizeof(*nlh
));
2363 if (nlk_has_extack
&& extack
) {
2365 WARN_ON(nla_put_string(skb
, NLMSGERR_ATTR_MSG
,
2369 if (extack
->bad_attr
&&
2370 !WARN_ON((u8
*)extack
->bad_attr
< in_skb
->data
||
2371 (u8
*)extack
->bad_attr
>= in_skb
->data
+
2373 WARN_ON(nla_put_u32(skb
, NLMSGERR_ATTR_OFFS
,
2374 (u8
*)extack
->bad_attr
-
2377 if (extack
->cookie_len
)
2378 WARN_ON(nla_put(skb
, NLMSGERR_ATTR_COOKIE
,
2384 nlmsg_end(skb
, rep
);
2386 netlink_unicast(in_skb
->sk
, skb
, NETLINK_CB(in_skb
).portid
, MSG_DONTWAIT
);
2388 EXPORT_SYMBOL(netlink_ack
);
2390 int netlink_rcv_skb(struct sk_buff
*skb
, int (*cb
)(struct sk_buff
*,
2392 struct netlink_ext_ack
*))
2394 struct netlink_ext_ack extack
;
2395 struct nlmsghdr
*nlh
;
2398 while (skb
->len
>= nlmsg_total_size(0)) {
2401 memset(&extack
, 0, sizeof(extack
));
2402 nlh
= nlmsg_hdr(skb
);
2405 if (nlh
->nlmsg_len
< NLMSG_HDRLEN
|| skb
->len
< nlh
->nlmsg_len
)
2408 /* Only requests are handled by the kernel */
2409 if (!(nlh
->nlmsg_flags
& NLM_F_REQUEST
))
2412 /* Skip control messages */
2413 if (nlh
->nlmsg_type
< NLMSG_MIN_TYPE
)
2416 err
= cb(skb
, nlh
, &extack
);
2421 if (nlh
->nlmsg_flags
& NLM_F_ACK
|| err
)
2422 netlink_ack(skb
, nlh
, err
, &extack
);
2425 msglen
= NLMSG_ALIGN(nlh
->nlmsg_len
);
2426 if (msglen
> skb
->len
)
2428 skb_pull(skb
, msglen
);
2433 EXPORT_SYMBOL(netlink_rcv_skb
);
2436 * nlmsg_notify - send a notification netlink message
2437 * @sk: netlink socket to use
2438 * @skb: notification message
2439 * @portid: destination netlink portid for reports or 0
2440 * @group: destination multicast group or 0
2441 * @report: 1 to report back, 0 to disable
2442 * @flags: allocation flags
2444 int nlmsg_notify(struct sock
*sk
, struct sk_buff
*skb
, u32 portid
,
2445 unsigned int group
, int report
, gfp_t flags
)
2450 int exclude_portid
= 0;
2453 refcount_inc(&skb
->users
);
2454 exclude_portid
= portid
;
2457 /* errors reported via destination sk->sk_err, but propagate
2458 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
2459 err
= nlmsg_multicast(sk
, skb
, exclude_portid
, group
, flags
);
2465 err2
= nlmsg_unicast(sk
, skb
, portid
);
2466 if (!err
|| err
== -ESRCH
)
2472 EXPORT_SYMBOL(nlmsg_notify
);
2474 #ifdef CONFIG_PROC_FS
2475 struct nl_seq_iter
{
2476 struct seq_net_private p
;
2477 struct rhashtable_iter hti
;
2481 static int netlink_walk_start(struct nl_seq_iter
*iter
)
2485 err
= rhashtable_walk_init(&nl_table
[iter
->link
].hash
, &iter
->hti
,
2488 iter
->link
= MAX_LINKS
;
2492 err
= rhashtable_walk_start(&iter
->hti
);
2493 return err
== -EAGAIN
? 0 : err
;
2496 static void netlink_walk_stop(struct nl_seq_iter
*iter
)
2498 rhashtable_walk_stop(&iter
->hti
);
2499 rhashtable_walk_exit(&iter
->hti
);
2502 static void *__netlink_seq_next(struct seq_file
*seq
)
2504 struct nl_seq_iter
*iter
= seq
->private;
2505 struct netlink_sock
*nlk
;
2511 nlk
= rhashtable_walk_next(&iter
->hti
);
2514 if (PTR_ERR(nlk
) == -EAGAIN
)
2523 netlink_walk_stop(iter
);
2524 if (++iter
->link
>= MAX_LINKS
)
2527 err
= netlink_walk_start(iter
);
2529 return ERR_PTR(err
);
2531 } while (sock_net(&nlk
->sk
) != seq_file_net(seq
));
2536 static void *netlink_seq_start(struct seq_file
*seq
, loff_t
*posp
)
2538 struct nl_seq_iter
*iter
= seq
->private;
2539 void *obj
= SEQ_START_TOKEN
;
2545 err
= netlink_walk_start(iter
);
2547 return ERR_PTR(err
);
2549 for (pos
= *posp
; pos
&& obj
&& !IS_ERR(obj
); pos
--)
2550 obj
= __netlink_seq_next(seq
);
2555 static void *netlink_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2558 return __netlink_seq_next(seq
);
2561 static void netlink_seq_stop(struct seq_file
*seq
, void *v
)
2563 struct nl_seq_iter
*iter
= seq
->private;
2565 if (iter
->link
>= MAX_LINKS
)
2568 netlink_walk_stop(iter
);
2572 static int netlink_seq_show(struct seq_file
*seq
, void *v
)
2574 if (v
== SEQ_START_TOKEN
) {
2576 "sk Eth Pid Groups "
2577 "Rmem Wmem Dump Locks Drops Inode\n");
2580 struct netlink_sock
*nlk
= nlk_sk(s
);
2582 seq_printf(seq
, "%pK %-3d %-6u %08x %-8d %-8d %d %-8d %-8d %-8lu\n",
2586 nlk
->groups
? (u32
)nlk
->groups
[0] : 0,
2587 sk_rmem_alloc_get(s
),
2588 sk_wmem_alloc_get(s
),
2590 refcount_read(&s
->sk_refcnt
),
2591 atomic_read(&s
->sk_drops
),
2599 static const struct seq_operations netlink_seq_ops
= {
2600 .start
= netlink_seq_start
,
2601 .next
= netlink_seq_next
,
2602 .stop
= netlink_seq_stop
,
2603 .show
= netlink_seq_show
,
2607 static int netlink_seq_open(struct inode
*inode
, struct file
*file
)
2609 return seq_open_net(inode
, file
, &netlink_seq_ops
,
2610 sizeof(struct nl_seq_iter
));
2613 static const struct file_operations netlink_seq_fops
= {
2614 .owner
= THIS_MODULE
,
2615 .open
= netlink_seq_open
,
2617 .llseek
= seq_lseek
,
2618 .release
= seq_release_net
,
2623 int netlink_register_notifier(struct notifier_block
*nb
)
2625 return blocking_notifier_chain_register(&netlink_chain
, nb
);
2627 EXPORT_SYMBOL(netlink_register_notifier
);
2629 int netlink_unregister_notifier(struct notifier_block
*nb
)
2631 return blocking_notifier_chain_unregister(&netlink_chain
, nb
);
2633 EXPORT_SYMBOL(netlink_unregister_notifier
);
2635 static const struct proto_ops netlink_ops
= {
2636 .family
= PF_NETLINK
,
2637 .owner
= THIS_MODULE
,
2638 .release
= netlink_release
,
2639 .bind
= netlink_bind
,
2640 .connect
= netlink_connect
,
2641 .socketpair
= sock_no_socketpair
,
2642 .accept
= sock_no_accept
,
2643 .getname
= netlink_getname
,
2644 .poll
= datagram_poll
,
2645 .ioctl
= netlink_ioctl
,
2646 .listen
= sock_no_listen
,
2647 .shutdown
= sock_no_shutdown
,
2648 .setsockopt
= netlink_setsockopt
,
2649 .getsockopt
= netlink_getsockopt
,
2650 .sendmsg
= netlink_sendmsg
,
2651 .recvmsg
= netlink_recvmsg
,
2652 .mmap
= sock_no_mmap
,
2653 .sendpage
= sock_no_sendpage
,
2656 static const struct net_proto_family netlink_family_ops
= {
2657 .family
= PF_NETLINK
,
2658 .create
= netlink_create
,
2659 .owner
= THIS_MODULE
, /* for consistency 8) */
2662 static int __net_init
netlink_net_init(struct net
*net
)
2664 #ifdef CONFIG_PROC_FS
2665 if (!proc_create("netlink", 0, net
->proc_net
, &netlink_seq_fops
))
2671 static void __net_exit
netlink_net_exit(struct net
*net
)
2673 #ifdef CONFIG_PROC_FS
2674 remove_proc_entry("netlink", net
->proc_net
);
2678 static void __init
netlink_add_usersock_entry(void)
2680 struct listeners
*listeners
;
2683 listeners
= kzalloc(sizeof(*listeners
) + NLGRPSZ(groups
), GFP_KERNEL
);
2685 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
2687 netlink_table_grab();
2689 nl_table
[NETLINK_USERSOCK
].groups
= groups
;
2690 rcu_assign_pointer(nl_table
[NETLINK_USERSOCK
].listeners
, listeners
);
2691 nl_table
[NETLINK_USERSOCK
].module
= THIS_MODULE
;
2692 nl_table
[NETLINK_USERSOCK
].registered
= 1;
2693 nl_table
[NETLINK_USERSOCK
].flags
= NL_CFG_F_NONROOT_SEND
;
2695 netlink_table_ungrab();
2698 static struct pernet_operations __net_initdata netlink_net_ops
= {
2699 .init
= netlink_net_init
,
2700 .exit
= netlink_net_exit
,
2703 static inline u32
netlink_hash(const void *data
, u32 len
, u32 seed
)
2705 const struct netlink_sock
*nlk
= data
;
2706 struct netlink_compare_arg arg
;
2708 netlink_compare_arg_init(&arg
, sock_net(&nlk
->sk
), nlk
->portid
);
2709 return jhash2((u32
*)&arg
, netlink_compare_arg_len
/ sizeof(u32
), seed
);
2712 static const struct rhashtable_params netlink_rhashtable_params
= {
2713 .head_offset
= offsetof(struct netlink_sock
, node
),
2714 .key_len
= netlink_compare_arg_len
,
2715 .obj_hashfn
= netlink_hash
,
2716 .obj_cmpfn
= netlink_compare
,
2717 .automatic_shrinking
= true,
2720 static int __init
netlink_proto_init(void)
2723 int err
= proto_register(&netlink_proto
, 0);
2728 BUILD_BUG_ON(sizeof(struct netlink_skb_parms
) > FIELD_SIZEOF(struct sk_buff
, cb
));
2730 nl_table
= kcalloc(MAX_LINKS
, sizeof(*nl_table
), GFP_KERNEL
);
2734 for (i
= 0; i
< MAX_LINKS
; i
++) {
2735 if (rhashtable_init(&nl_table
[i
].hash
,
2736 &netlink_rhashtable_params
) < 0) {
2738 rhashtable_destroy(&nl_table
[i
].hash
);
2744 INIT_LIST_HEAD(&netlink_tap_all
);
2746 netlink_add_usersock_entry();
2748 sock_register(&netlink_family_ops
);
2749 register_pernet_subsys(&netlink_net_ops
);
2750 /* The netlink device handler may be needed early. */
2755 panic("netlink_init: Cannot allocate nl_table\n");
2758 core_initcall(netlink_proto_init
);