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;
88 unsigned long masks[0];
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 struct listeners __rcu *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);
136 static DEFINE_RWLOCK(nl_table_lock);
137 static atomic_t nl_table_users = ATOMIC_INIT(0);
139 static ATOMIC_NOTIFIER_HEAD(netlink_chain);
141 static inline u32 netlink_group_mask(u32 group)
143 return group ? 1 << (group - 1) : 0;
146 static inline struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
148 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
151 static void netlink_destroy_callback(struct netlink_callback *cb)
157 static void netlink_sock_destruct(struct sock *sk)
159 struct netlink_sock *nlk = nlk_sk(sk);
163 nlk->cb->done(nlk->cb);
164 netlink_destroy_callback(nlk->cb);
167 skb_queue_purge(&sk->sk_receive_queue);
169 if (!sock_flag(sk, SOCK_DEAD)) {
170 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
174 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
175 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
176 WARN_ON(nlk_sk(sk)->groups);
179 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
180 * SMP. Look, when several writers sleep and reader wakes them up, all but one
181 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
182 * this, _but_ remember, it adds useless work on UP machines.
185 void netlink_table_grab(void)
186 __acquires(nl_table_lock)
190 write_lock_irq(&nl_table_lock);
192 if (atomic_read(&nl_table_users)) {
193 DECLARE_WAITQUEUE(wait, current);
195 add_wait_queue_exclusive(&nl_table_wait, &wait);
197 set_current_state(TASK_UNINTERRUPTIBLE);
198 if (atomic_read(&nl_table_users) == 0)
200 write_unlock_irq(&nl_table_lock);
202 write_lock_irq(&nl_table_lock);
205 __set_current_state(TASK_RUNNING);
206 remove_wait_queue(&nl_table_wait, &wait);
210 void netlink_table_ungrab(void)
211 __releases(nl_table_lock)
213 write_unlock_irq(&nl_table_lock);
214 wake_up(&nl_table_wait);
218 netlink_lock_table(void)
220 /* read_lock() synchronizes us to netlink_table_grab */
222 read_lock(&nl_table_lock);
223 atomic_inc(&nl_table_users);
224 read_unlock(&nl_table_lock);
228 netlink_unlock_table(void)
230 if (atomic_dec_and_test(&nl_table_users))
231 wake_up(&nl_table_wait);
234 static struct sock *netlink_lookup(struct net *net, int protocol, u32 pid)
236 struct nl_pid_hash *hash = &nl_table[protocol].hash;
237 struct hlist_head *head;
239 struct hlist_node *node;
241 read_lock(&nl_table_lock);
242 head = nl_pid_hashfn(hash, pid);
243 sk_for_each(sk, node, head) {
244 if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->pid == pid)) {
251 read_unlock(&nl_table_lock);
255 static struct hlist_head *nl_pid_hash_zalloc(size_t size)
257 if (size <= PAGE_SIZE)
258 return kzalloc(size, GFP_ATOMIC);
260 return (struct hlist_head *)
261 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
265 static void nl_pid_hash_free(struct hlist_head *table, size_t size)
267 if (size <= PAGE_SIZE)
270 free_pages((unsigned long)table, get_order(size));
273 static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
275 unsigned int omask, mask, shift;
277 struct hlist_head *otable, *table;
280 omask = mask = hash->mask;
281 osize = size = (mask + 1) * sizeof(*table);
285 if (++shift > hash->max_shift)
291 table = nl_pid_hash_zalloc(size);
295 otable = hash->table;
299 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
301 for (i = 0; i <= omask; i++) {
303 struct hlist_node *node, *tmp;
305 sk_for_each_safe(sk, node, tmp, &otable[i])
306 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
309 nl_pid_hash_free(otable, osize);
310 hash->rehash_time = jiffies + 10 * 60 * HZ;
314 static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
316 int avg = hash->entries >> hash->shift;
318 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
321 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
322 nl_pid_hash_rehash(hash, 0);
329 static const struct proto_ops netlink_ops;
332 netlink_update_listeners(struct sock *sk)
334 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
335 struct hlist_node *node;
339 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
341 sk_for_each_bound(sk, node, &tbl->mc_list) {
342 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
343 mask |= nlk_sk(sk)->groups[i];
345 tbl->listeners->masks[i] = mask;
347 /* this function is only called with the netlink table "grabbed", which
348 * makes sure updates are visible before bind or setsockopt return. */
351 static int netlink_insert(struct sock *sk, struct net *net, u32 pid)
353 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
354 struct hlist_head *head;
355 int err = -EADDRINUSE;
357 struct hlist_node *node;
360 netlink_table_grab();
361 head = nl_pid_hashfn(hash, pid);
363 sk_for_each(osk, node, head) {
364 if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->pid == pid))
376 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
379 if (len && nl_pid_hash_dilute(hash, len))
380 head = nl_pid_hashfn(hash, pid);
382 nlk_sk(sk)->pid = pid;
383 sk_add_node(sk, head);
387 netlink_table_ungrab();
391 static void netlink_remove(struct sock *sk)
393 netlink_table_grab();
394 if (sk_del_node_init(sk))
395 nl_table[sk->sk_protocol].hash.entries--;
396 if (nlk_sk(sk)->subscriptions)
397 __sk_del_bind_node(sk);
398 netlink_table_ungrab();
401 static struct proto netlink_proto = {
403 .owner = THIS_MODULE,
404 .obj_size = sizeof(struct netlink_sock),
407 static int __netlink_create(struct net *net, struct socket *sock,
408 struct mutex *cb_mutex, int protocol)
411 struct netlink_sock *nlk;
413 sock->ops = &netlink_ops;
415 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
419 sock_init_data(sock, sk);
423 nlk->cb_mutex = cb_mutex;
425 nlk->cb_mutex = &nlk->cb_def_mutex;
426 mutex_init(nlk->cb_mutex);
428 init_waitqueue_head(&nlk->wait);
430 sk->sk_destruct = netlink_sock_destruct;
431 sk->sk_protocol = protocol;
435 static int netlink_create(struct net *net, struct socket *sock, int protocol,
438 struct module *module = NULL;
439 struct mutex *cb_mutex;
440 struct netlink_sock *nlk;
443 sock->state = SS_UNCONNECTED;
445 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
446 return -ESOCKTNOSUPPORT;
448 if (protocol < 0 || protocol >= MAX_LINKS)
449 return -EPROTONOSUPPORT;
451 netlink_lock_table();
452 #ifdef CONFIG_MODULES
453 if (!nl_table[protocol].registered) {
454 netlink_unlock_table();
455 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
456 netlink_lock_table();
459 if (nl_table[protocol].registered &&
460 try_module_get(nl_table[protocol].module))
461 module = nl_table[protocol].module;
463 err = -EPROTONOSUPPORT;
464 cb_mutex = nl_table[protocol].cb_mutex;
465 netlink_unlock_table();
470 err = __netlink_create(net, sock, cb_mutex, protocol);
475 sock_prot_inuse_add(net, &netlink_proto, 1);
478 nlk = nlk_sk(sock->sk);
479 nlk->module = module;
488 static int netlink_release(struct socket *sock)
490 struct sock *sk = sock->sk;
491 struct netlink_sock *nlk;
501 * OK. Socket is unlinked, any packets that arrive now
506 wake_up_interruptible_all(&nlk->wait);
508 skb_queue_purge(&sk->sk_write_queue);
511 struct netlink_notify n = {
513 .protocol = sk->sk_protocol,
516 atomic_notifier_call_chain(&netlink_chain,
517 NETLINK_URELEASE, &n);
520 module_put(nlk->module);
522 netlink_table_grab();
523 if (netlink_is_kernel(sk)) {
524 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
525 if (--nl_table[sk->sk_protocol].registered == 0) {
526 kfree(nl_table[sk->sk_protocol].listeners);
527 nl_table[sk->sk_protocol].module = NULL;
528 nl_table[sk->sk_protocol].registered = 0;
530 } else if (nlk->subscriptions) {
531 netlink_update_listeners(sk);
533 netlink_table_ungrab();
539 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
545 static int netlink_autobind(struct socket *sock)
547 struct sock *sk = sock->sk;
548 struct net *net = sock_net(sk);
549 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
550 struct hlist_head *head;
552 struct hlist_node *node;
553 s32 pid = task_tgid_vnr(current);
555 static s32 rover = -4097;
559 netlink_table_grab();
560 head = nl_pid_hashfn(hash, pid);
561 sk_for_each(osk, node, head) {
562 if (!net_eq(sock_net(osk), net))
564 if (nlk_sk(osk)->pid == pid) {
565 /* Bind collision, search negative pid values. */
569 netlink_table_ungrab();
573 netlink_table_ungrab();
575 err = netlink_insert(sk, net, pid);
576 if (err == -EADDRINUSE)
579 /* If 2 threads race to autobind, that is fine. */
586 static inline int netlink_capable(const struct socket *sock, unsigned int flag)
588 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
589 capable(CAP_NET_ADMIN);
593 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
595 struct netlink_sock *nlk = nlk_sk(sk);
597 if (nlk->subscriptions && !subscriptions)
598 __sk_del_bind_node(sk);
599 else if (!nlk->subscriptions && subscriptions)
600 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
601 nlk->subscriptions = subscriptions;
604 static int netlink_realloc_groups(struct sock *sk)
606 struct netlink_sock *nlk = nlk_sk(sk);
608 unsigned long *new_groups;
611 netlink_table_grab();
613 groups = nl_table[sk->sk_protocol].groups;
614 if (!nl_table[sk->sk_protocol].registered) {
619 if (nlk->ngroups >= groups)
622 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
623 if (new_groups == NULL) {
627 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
628 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
630 nlk->groups = new_groups;
631 nlk->ngroups = groups;
633 netlink_table_ungrab();
637 static int netlink_bind(struct socket *sock, struct sockaddr *addr,
640 struct sock *sk = sock->sk;
641 struct net *net = sock_net(sk);
642 struct netlink_sock *nlk = nlk_sk(sk);
643 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
646 if (nladdr->nl_family != AF_NETLINK)
649 /* Only superuser is allowed to listen multicasts */
650 if (nladdr->nl_groups) {
651 if (!netlink_capable(sock, NL_NONROOT_RECV))
653 err = netlink_realloc_groups(sk);
659 if (nladdr->nl_pid != nlk->pid)
662 err = nladdr->nl_pid ?
663 netlink_insert(sk, net, nladdr->nl_pid) :
664 netlink_autobind(sock);
669 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
672 netlink_table_grab();
673 netlink_update_subscriptions(sk, nlk->subscriptions +
674 hweight32(nladdr->nl_groups) -
675 hweight32(nlk->groups[0]));
676 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
677 netlink_update_listeners(sk);
678 netlink_table_ungrab();
683 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
687 struct sock *sk = sock->sk;
688 struct netlink_sock *nlk = nlk_sk(sk);
689 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
691 if (alen < sizeof(addr->sa_family))
694 if (addr->sa_family == AF_UNSPEC) {
695 sk->sk_state = NETLINK_UNCONNECTED;
700 if (addr->sa_family != AF_NETLINK)
703 /* Only superuser is allowed to send multicasts */
704 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
708 err = netlink_autobind(sock);
711 sk->sk_state = NETLINK_CONNECTED;
712 nlk->dst_pid = nladdr->nl_pid;
713 nlk->dst_group = ffs(nladdr->nl_groups);
719 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
720 int *addr_len, int peer)
722 struct sock *sk = sock->sk;
723 struct netlink_sock *nlk = nlk_sk(sk);
724 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
726 nladdr->nl_family = AF_NETLINK;
728 *addr_len = sizeof(*nladdr);
731 nladdr->nl_pid = nlk->dst_pid;
732 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
734 nladdr->nl_pid = nlk->pid;
735 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
740 static void netlink_overrun(struct sock *sk)
742 struct netlink_sock *nlk = nlk_sk(sk);
744 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
745 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
746 sk->sk_err = ENOBUFS;
747 sk->sk_error_report(sk);
750 atomic_inc(&sk->sk_drops);
753 static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
756 struct netlink_sock *nlk;
758 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, pid);
760 return ERR_PTR(-ECONNREFUSED);
762 /* Don't bother queuing skb if kernel socket has no input function */
764 if (sock->sk_state == NETLINK_CONNECTED &&
765 nlk->dst_pid != nlk_sk(ssk)->pid) {
767 return ERR_PTR(-ECONNREFUSED);
772 struct sock *netlink_getsockbyfilp(struct file *filp)
774 struct inode *inode = filp->f_path.dentry->d_inode;
777 if (!S_ISSOCK(inode->i_mode))
778 return ERR_PTR(-ENOTSOCK);
780 sock = SOCKET_I(inode)->sk;
781 if (sock->sk_family != AF_NETLINK)
782 return ERR_PTR(-EINVAL);
789 * Attach a skb to a netlink socket.
790 * The caller must hold a reference to the destination socket. On error, the
791 * reference is dropped. The skb is not send to the destination, just all
792 * all error checks are performed and memory in the queue is reserved.
794 * < 0: error. skb freed, reference to sock dropped.
796 * 1: repeat lookup - reference dropped while waiting for socket memory.
798 int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
799 long *timeo, struct sock *ssk)
801 struct netlink_sock *nlk;
805 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
806 test_bit(0, &nlk->state)) {
807 DECLARE_WAITQUEUE(wait, current);
809 if (!ssk || netlink_is_kernel(ssk))
816 __set_current_state(TASK_INTERRUPTIBLE);
817 add_wait_queue(&nlk->wait, &wait);
819 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
820 test_bit(0, &nlk->state)) &&
821 !sock_flag(sk, SOCK_DEAD))
822 *timeo = schedule_timeout(*timeo);
824 __set_current_state(TASK_RUNNING);
825 remove_wait_queue(&nlk->wait, &wait);
828 if (signal_pending(current)) {
830 return sock_intr_errno(*timeo);
834 skb_set_owner_r(skb, sk);
838 static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
842 skb_queue_tail(&sk->sk_receive_queue, skb);
843 sk->sk_data_ready(sk, len);
847 int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
849 int len = __netlink_sendskb(sk, skb);
855 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
861 static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
867 delta = skb->end - skb->tail;
868 if (delta * 2 < skb->truesize)
871 if (skb_shared(skb)) {
872 struct sk_buff *nskb = skb_clone(skb, allocation);
879 if (!pskb_expand_head(skb, 0, -delta, allocation))
880 skb->truesize -= delta;
885 static void netlink_rcv_wake(struct sock *sk)
887 struct netlink_sock *nlk = nlk_sk(sk);
889 if (skb_queue_empty(&sk->sk_receive_queue))
890 clear_bit(0, &nlk->state);
891 if (!test_bit(0, &nlk->state))
892 wake_up_interruptible(&nlk->wait);
895 static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb)
898 struct netlink_sock *nlk = nlk_sk(sk);
901 if (nlk->netlink_rcv != NULL) {
903 skb_set_owner_r(skb, sk);
904 nlk->netlink_rcv(skb);
911 int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
912 u32 pid, int nonblock)
918 skb = netlink_trim(skb, gfp_any());
920 timeo = sock_sndtimeo(ssk, nonblock);
922 sk = netlink_getsockbypid(ssk, pid);
927 if (netlink_is_kernel(sk))
928 return netlink_unicast_kernel(sk, skb);
930 if (sk_filter(sk, skb)) {
937 err = netlink_attachskb(sk, skb, &timeo, ssk);
943 return netlink_sendskb(sk, skb);
945 EXPORT_SYMBOL(netlink_unicast);
947 int netlink_has_listeners(struct sock *sk, unsigned int group)
950 struct listeners *listeners;
952 BUG_ON(!netlink_is_kernel(sk));
955 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
957 if (group - 1 < nl_table[sk->sk_protocol].groups)
958 res = test_bit(group - 1, listeners->masks);
964 EXPORT_SYMBOL_GPL(netlink_has_listeners);
966 static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
968 struct netlink_sock *nlk = nlk_sk(sk);
970 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
971 !test_bit(0, &nlk->state)) {
972 skb_set_owner_r(skb, sk);
973 __netlink_sendskb(sk, skb);
974 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
979 struct netlink_broadcast_data {
980 struct sock *exclude_sk;
985 int delivery_failure;
989 struct sk_buff *skb, *skb2;
990 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
994 static int do_one_broadcast(struct sock *sk,
995 struct netlink_broadcast_data *p)
997 struct netlink_sock *nlk = nlk_sk(sk);
1000 if (p->exclude_sk == sk)
1003 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
1004 !test_bit(p->group - 1, nlk->groups))
1007 if (!net_eq(sock_net(sk), p->net))
1011 netlink_overrun(sk);
1016 if (p->skb2 == NULL) {
1017 if (skb_shared(p->skb)) {
1018 p->skb2 = skb_clone(p->skb, p->allocation);
1020 p->skb2 = skb_get(p->skb);
1022 * skb ownership may have been set when
1023 * delivered to a previous socket.
1025 skb_orphan(p->skb2);
1028 if (p->skb2 == NULL) {
1029 netlink_overrun(sk);
1030 /* Clone failed. Notify ALL listeners. */
1032 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1033 p->delivery_failure = 1;
1034 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1037 } else if (sk_filter(sk, p->skb2)) {
1040 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1041 netlink_overrun(sk);
1042 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1043 p->delivery_failure = 1;
1045 p->congested |= val;
1055 int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 pid,
1056 u32 group, gfp_t allocation,
1057 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1060 struct net *net = sock_net(ssk);
1061 struct netlink_broadcast_data info;
1062 struct hlist_node *node;
1065 skb = netlink_trim(skb, allocation);
1067 info.exclude_sk = ssk;
1072 info.delivery_failure = 0;
1075 info.allocation = allocation;
1078 info.tx_filter = filter;
1079 info.tx_data = filter_data;
1081 /* While we sleep in clone, do not allow to change socket list */
1083 netlink_lock_table();
1085 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1086 do_one_broadcast(sk, &info);
1090 netlink_unlock_table();
1092 if (info.delivery_failure) {
1093 kfree_skb(info.skb2);
1096 consume_skb(info.skb2);
1098 if (info.delivered) {
1099 if (info.congested && (allocation & __GFP_WAIT))
1105 EXPORT_SYMBOL(netlink_broadcast_filtered);
1107 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
1108 u32 group, gfp_t allocation)
1110 return netlink_broadcast_filtered(ssk, skb, pid, group, allocation,
1113 EXPORT_SYMBOL(netlink_broadcast);
1115 struct netlink_set_err_data {
1116 struct sock *exclude_sk;
1122 static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
1124 struct netlink_sock *nlk = nlk_sk(sk);
1127 if (sk == p->exclude_sk)
1130 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
1133 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
1134 !test_bit(p->group - 1, nlk->groups))
1137 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
1142 sk->sk_err = p->code;
1143 sk->sk_error_report(sk);
1149 * netlink_set_err - report error to broadcast listeners
1150 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1151 * @pid: the PID of a process that we want to skip (if any)
1152 * @groups: the broadcast group that will notice the error
1153 * @code: error code, must be negative (as usual in kernelspace)
1155 * This function returns the number of broadcast listeners that have set the
1156 * NETLINK_RECV_NO_ENOBUFS socket option.
1158 int netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
1160 struct netlink_set_err_data info;
1161 struct hlist_node *node;
1165 info.exclude_sk = ssk;
1168 /* sk->sk_err wants a positive error value */
1171 read_lock(&nl_table_lock);
1173 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1174 ret += do_one_set_err(sk, &info);
1176 read_unlock(&nl_table_lock);
1179 EXPORT_SYMBOL(netlink_set_err);
1181 /* must be called with netlink table grabbed */
1182 static void netlink_update_socket_mc(struct netlink_sock *nlk,
1186 int old, new = !!is_new, subscriptions;
1188 old = test_bit(group - 1, nlk->groups);
1189 subscriptions = nlk->subscriptions - old + new;
1191 __set_bit(group - 1, nlk->groups);
1193 __clear_bit(group - 1, nlk->groups);
1194 netlink_update_subscriptions(&nlk->sk, subscriptions);
1195 netlink_update_listeners(&nlk->sk);
1198 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1199 char __user *optval, unsigned int optlen)
1201 struct sock *sk = sock->sk;
1202 struct netlink_sock *nlk = nlk_sk(sk);
1203 unsigned int val = 0;
1206 if (level != SOL_NETLINK)
1207 return -ENOPROTOOPT;
1209 if (optlen >= sizeof(int) &&
1210 get_user(val, (unsigned int __user *)optval))
1214 case NETLINK_PKTINFO:
1216 nlk->flags |= NETLINK_RECV_PKTINFO;
1218 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1221 case NETLINK_ADD_MEMBERSHIP:
1222 case NETLINK_DROP_MEMBERSHIP: {
1223 if (!netlink_capable(sock, NL_NONROOT_RECV))
1225 err = netlink_realloc_groups(sk);
1228 if (!val || val - 1 >= nlk->ngroups)
1230 netlink_table_grab();
1231 netlink_update_socket_mc(nlk, val,
1232 optname == NETLINK_ADD_MEMBERSHIP);
1233 netlink_table_ungrab();
1237 case NETLINK_BROADCAST_ERROR:
1239 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
1241 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
1244 case NETLINK_NO_ENOBUFS:
1246 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
1247 clear_bit(0, &nlk->state);
1248 wake_up_interruptible(&nlk->wait);
1250 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
1260 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1261 char __user *optval, int __user *optlen)
1263 struct sock *sk = sock->sk;
1264 struct netlink_sock *nlk = nlk_sk(sk);
1267 if (level != SOL_NETLINK)
1268 return -ENOPROTOOPT;
1270 if (get_user(len, optlen))
1276 case NETLINK_PKTINFO:
1277 if (len < sizeof(int))
1280 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
1281 if (put_user(len, optlen) ||
1282 put_user(val, optval))
1286 case NETLINK_BROADCAST_ERROR:
1287 if (len < sizeof(int))
1290 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
1291 if (put_user(len, optlen) ||
1292 put_user(val, optval))
1296 case NETLINK_NO_ENOBUFS:
1297 if (len < sizeof(int))
1300 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
1301 if (put_user(len, optlen) ||
1302 put_user(val, optval))
1312 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1314 struct nl_pktinfo info;
1316 info.group = NETLINK_CB(skb).dst_group;
1317 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1320 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1321 struct msghdr *msg, size_t len)
1323 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1324 struct sock *sk = sock->sk;
1325 struct netlink_sock *nlk = nlk_sk(sk);
1326 struct sockaddr_nl *addr = msg->msg_name;
1329 struct sk_buff *skb;
1331 struct scm_cookie scm;
1333 if (msg->msg_flags&MSG_OOB)
1336 if (NULL == siocb->scm)
1339 err = scm_send(sock, msg, siocb->scm);
1343 if (msg->msg_namelen) {
1345 if (addr->nl_family != AF_NETLINK)
1347 dst_pid = addr->nl_pid;
1348 dst_group = ffs(addr->nl_groups);
1350 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
1353 dst_pid = nlk->dst_pid;
1354 dst_group = nlk->dst_group;
1358 err = netlink_autobind(sock);
1364 if (len > sk->sk_sndbuf - 32)
1367 skb = alloc_skb(len, GFP_KERNEL);
1371 NETLINK_CB(skb).pid = nlk->pid;
1372 NETLINK_CB(skb).dst_group = dst_group;
1373 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
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, *data_skb;
1416 skb = skb_recv_datagram(sk, flags, noblock, &err);
1422 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1423 if (unlikely(skb_shinfo(skb)->frag_list)) {
1425 * If this skb has a frag_list, then here that means that we
1426 * will have to use the frag_list skb's data for compat tasks
1427 * and the regular skb's data for normal (non-compat) tasks.
1429 * If we need to send the compat skb, assign it to the
1430 * 'data_skb' variable so that it will be used below for data
1431 * copying. We keep 'skb' for everything else, including
1432 * freeing both later.
1434 if (flags & MSG_CMSG_COMPAT)
1435 data_skb = skb_shinfo(skb)->frag_list;
1439 msg->msg_namelen = 0;
1441 copied = data_skb->len;
1443 msg->msg_flags |= MSG_TRUNC;
1447 skb_reset_transport_header(data_skb);
1448 err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied);
1450 if (msg->msg_name) {
1451 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
1452 addr->nl_family = AF_NETLINK;
1454 addr->nl_pid = NETLINK_CB(skb).pid;
1455 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1456 msg->msg_namelen = sizeof(*addr);
1459 if (nlk->flags & NETLINK_RECV_PKTINFO)
1460 netlink_cmsg_recv_pktinfo(msg, skb);
1462 if (NULL == siocb->scm) {
1463 memset(&scm, 0, sizeof(scm));
1466 siocb->scm->creds = *NETLINK_CREDS(skb);
1467 if (flags & MSG_TRUNC)
1468 copied = data_skb->len;
1470 skb_free_datagram(sk, skb);
1472 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
1473 ret = netlink_dump(sk);
1476 sk->sk_error_report(sk);
1480 scm_recv(sock, msg, siocb->scm, flags);
1482 netlink_rcv_wake(sk);
1483 return err ? : copied;
1486 static void netlink_data_ready(struct sock *sk, int len)
1492 * We export these functions to other modules. They provide a
1493 * complete set of kernel non-blocking support for message
1498 netlink_kernel_create(struct net *net, int unit, unsigned int groups,
1499 void (*input)(struct sk_buff *skb),
1500 struct mutex *cb_mutex, struct module *module)
1502 struct socket *sock;
1504 struct netlink_sock *nlk;
1505 struct listeners *listeners = NULL;
1509 if (unit < 0 || unit >= MAX_LINKS)
1512 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1516 * We have to just have a reference on the net from sk, but don't
1517 * get_net it. Besides, we cannot get and then put the net here.
1518 * So we create one inside init_net and the move it to net.
1521 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
1522 goto out_sock_release_nosk;
1525 sk_change_net(sk, net);
1530 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
1532 goto out_sock_release;
1534 sk->sk_data_ready = netlink_data_ready;
1536 nlk_sk(sk)->netlink_rcv = input;
1538 if (netlink_insert(sk, net, 0))
1539 goto out_sock_release;
1542 nlk->flags |= NETLINK_KERNEL_SOCKET;
1544 netlink_table_grab();
1545 if (!nl_table[unit].registered) {
1546 nl_table[unit].groups = groups;
1547 rcu_assign_pointer(nl_table[unit].listeners, listeners);
1548 nl_table[unit].cb_mutex = cb_mutex;
1549 nl_table[unit].module = module;
1550 nl_table[unit].registered = 1;
1553 nl_table[unit].registered++;
1555 netlink_table_ungrab();
1560 netlink_kernel_release(sk);
1563 out_sock_release_nosk:
1567 EXPORT_SYMBOL(netlink_kernel_create);
1571 netlink_kernel_release(struct sock *sk)
1573 sk_release_kernel(sk);
1575 EXPORT_SYMBOL(netlink_kernel_release);
1577 int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
1579 struct listeners *new, *old;
1580 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
1585 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
1586 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
1589 old = rcu_dereference_protected(tbl->listeners, 1);
1590 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
1591 rcu_assign_pointer(tbl->listeners, new);
1593 kfree_rcu(old, rcu);
1595 tbl->groups = groups;
1601 * netlink_change_ngroups - change number of multicast groups
1603 * This changes the number of multicast groups that are available
1604 * on a certain netlink family. Note that it is not possible to
1605 * change the number of groups to below 32. Also note that it does
1606 * not implicitly call netlink_clear_multicast_users() when the
1607 * number of groups is reduced.
1609 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1610 * @groups: The new number of groups.
1612 int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1616 netlink_table_grab();
1617 err = __netlink_change_ngroups(sk, groups);
1618 netlink_table_ungrab();
1623 void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1626 struct hlist_node *node;
1627 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1629 sk_for_each_bound(sk, node, &tbl->mc_list)
1630 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1634 * netlink_clear_multicast_users - kick off multicast listeners
1636 * This function removes all listeners from the given group.
1637 * @ksk: The kernel netlink socket, as returned by
1638 * netlink_kernel_create().
1639 * @group: The multicast group to clear.
1641 void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1643 netlink_table_grab();
1644 __netlink_clear_multicast_users(ksk, group);
1645 netlink_table_ungrab();
1648 void netlink_set_nonroot(int protocol, unsigned int flags)
1650 if ((unsigned int)protocol < MAX_LINKS)
1651 nl_table[protocol].nl_nonroot = flags;
1653 EXPORT_SYMBOL(netlink_set_nonroot);
1656 __nlmsg_put(struct sk_buff *skb, u32 pid, u32 seq, int type, int len, int flags)
1658 struct nlmsghdr *nlh;
1659 int size = NLMSG_LENGTH(len);
1661 nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size));
1662 nlh->nlmsg_type = type;
1663 nlh->nlmsg_len = size;
1664 nlh->nlmsg_flags = flags;
1665 nlh->nlmsg_pid = pid;
1666 nlh->nlmsg_seq = seq;
1667 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
1668 memset(NLMSG_DATA(nlh) + len, 0, NLMSG_ALIGN(size) - size);
1671 EXPORT_SYMBOL(__nlmsg_put);
1674 * It looks a bit ugly.
1675 * It would be better to create kernel thread.
1678 static int netlink_dump(struct sock *sk)
1680 struct netlink_sock *nlk = nlk_sk(sk);
1681 struct netlink_callback *cb;
1682 struct sk_buff *skb = NULL;
1683 struct nlmsghdr *nlh;
1684 int len, err = -ENOBUFS;
1687 mutex_lock(nlk->cb_mutex);
1695 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
1697 skb = sock_rmalloc(sk, alloc_size, 0, GFP_KERNEL);
1701 len = cb->dump(skb, cb);
1704 mutex_unlock(nlk->cb_mutex);
1706 if (sk_filter(sk, skb))
1709 __netlink_sendskb(sk, skb);
1713 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1717 nl_dump_check_consistent(cb, nlh);
1719 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1721 if (sk_filter(sk, skb))
1724 __netlink_sendskb(sk, skb);
1729 mutex_unlock(nlk->cb_mutex);
1731 netlink_destroy_callback(cb);
1735 mutex_unlock(nlk->cb_mutex);
1740 int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1741 const struct nlmsghdr *nlh,
1742 struct netlink_dump_control *control)
1744 struct netlink_callback *cb;
1746 struct netlink_sock *nlk;
1749 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
1753 cb->dump = control->dump;
1754 cb->done = control->done;
1756 cb->data = control->data;
1757 cb->min_dump_alloc = control->min_dump_alloc;
1758 atomic_inc(&skb->users);
1761 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).pid);
1763 netlink_destroy_callback(cb);
1764 return -ECONNREFUSED;
1767 /* A dump is in progress... */
1768 mutex_lock(nlk->cb_mutex);
1770 mutex_unlock(nlk->cb_mutex);
1771 netlink_destroy_callback(cb);
1776 mutex_unlock(nlk->cb_mutex);
1778 ret = netlink_dump(sk);
1785 /* We successfully started a dump, by returning -EINTR we
1786 * signal not to send ACK even if it was requested.
1790 EXPORT_SYMBOL(netlink_dump_start);
1792 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1794 struct sk_buff *skb;
1795 struct nlmsghdr *rep;
1796 struct nlmsgerr *errmsg;
1797 size_t payload = sizeof(*errmsg);
1799 /* error messages get the original request appened */
1801 payload += nlmsg_len(nlh);
1803 skb = nlmsg_new(payload, GFP_KERNEL);
1807 sk = netlink_lookup(sock_net(in_skb->sk),
1808 in_skb->sk->sk_protocol,
1809 NETLINK_CB(in_skb).pid);
1811 sk->sk_err = ENOBUFS;
1812 sk->sk_error_report(sk);
1818 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
1819 NLMSG_ERROR, payload, 0);
1820 errmsg = nlmsg_data(rep);
1821 errmsg->error = err;
1822 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
1823 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1825 EXPORT_SYMBOL(netlink_ack);
1827 int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
1830 struct nlmsghdr *nlh;
1833 while (skb->len >= nlmsg_total_size(0)) {
1836 nlh = nlmsg_hdr(skb);
1839 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
1842 /* Only requests are handled by the kernel */
1843 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
1846 /* Skip control messages */
1847 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
1855 if (nlh->nlmsg_flags & NLM_F_ACK || err)
1856 netlink_ack(skb, nlh, err);
1859 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1860 if (msglen > skb->len)
1862 skb_pull(skb, msglen);
1867 EXPORT_SYMBOL(netlink_rcv_skb);
1870 * nlmsg_notify - send a notification netlink message
1871 * @sk: netlink socket to use
1872 * @skb: notification message
1873 * @pid: destination netlink pid for reports or 0
1874 * @group: destination multicast group or 0
1875 * @report: 1 to report back, 0 to disable
1876 * @flags: allocation flags
1878 int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid,
1879 unsigned int group, int report, gfp_t flags)
1884 int exclude_pid = 0;
1887 atomic_inc(&skb->users);
1891 /* errors reported via destination sk->sk_err, but propagate
1892 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
1893 err = nlmsg_multicast(sk, skb, exclude_pid, group, flags);
1899 err2 = nlmsg_unicast(sk, skb, pid);
1900 if (!err || err == -ESRCH)
1906 EXPORT_SYMBOL(nlmsg_notify);
1908 #ifdef CONFIG_PROC_FS
1909 struct nl_seq_iter {
1910 struct seq_net_private p;
1915 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1917 struct nl_seq_iter *iter = seq->private;
1920 struct hlist_node *node;
1923 for (i = 0; i < MAX_LINKS; i++) {
1924 struct nl_pid_hash *hash = &nl_table[i].hash;
1926 for (j = 0; j <= hash->mask; j++) {
1927 sk_for_each(s, node, &hash->table[j]) {
1928 if (sock_net(s) != seq_file_net(seq))
1942 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
1943 __acquires(nl_table_lock)
1945 read_lock(&nl_table_lock);
1946 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1949 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1952 struct nl_seq_iter *iter;
1957 if (v == SEQ_START_TOKEN)
1958 return netlink_seq_socket_idx(seq, 0);
1960 iter = seq->private;
1964 } while (s && sock_net(s) != seq_file_net(seq));
1969 j = iter->hash_idx + 1;
1972 struct nl_pid_hash *hash = &nl_table[i].hash;
1974 for (; j <= hash->mask; j++) {
1975 s = sk_head(&hash->table[j]);
1976 while (s && sock_net(s) != seq_file_net(seq))
1986 } while (++i < MAX_LINKS);
1991 static void netlink_seq_stop(struct seq_file *seq, void *v)
1992 __releases(nl_table_lock)
1994 read_unlock(&nl_table_lock);
1998 static int netlink_seq_show(struct seq_file *seq, void *v)
2000 if (v == SEQ_START_TOKEN) {
2002 "sk Eth Pid Groups "
2003 "Rmem Wmem Dump Locks Drops Inode\n");
2006 struct netlink_sock *nlk = nlk_sk(s);
2008 seq_printf(seq, "%pK %-3d %-6d %08x %-8d %-8d %pK %-8d %-8d %-8lu\n",
2012 nlk->groups ? (u32)nlk->groups[0] : 0,
2013 sk_rmem_alloc_get(s),
2014 sk_wmem_alloc_get(s),
2016 atomic_read(&s->sk_refcnt),
2017 atomic_read(&s->sk_drops),
2025 static const struct seq_operations netlink_seq_ops = {
2026 .start = netlink_seq_start,
2027 .next = netlink_seq_next,
2028 .stop = netlink_seq_stop,
2029 .show = netlink_seq_show,
2033 static int netlink_seq_open(struct inode *inode, struct file *file)
2035 return seq_open_net(inode, file, &netlink_seq_ops,
2036 sizeof(struct nl_seq_iter));
2039 static const struct file_operations netlink_seq_fops = {
2040 .owner = THIS_MODULE,
2041 .open = netlink_seq_open,
2043 .llseek = seq_lseek,
2044 .release = seq_release_net,
2049 int netlink_register_notifier(struct notifier_block *nb)
2051 return atomic_notifier_chain_register(&netlink_chain, nb);
2053 EXPORT_SYMBOL(netlink_register_notifier);
2055 int netlink_unregister_notifier(struct notifier_block *nb)
2057 return atomic_notifier_chain_unregister(&netlink_chain, nb);
2059 EXPORT_SYMBOL(netlink_unregister_notifier);
2061 static const struct proto_ops netlink_ops = {
2062 .family = PF_NETLINK,
2063 .owner = THIS_MODULE,
2064 .release = netlink_release,
2065 .bind = netlink_bind,
2066 .connect = netlink_connect,
2067 .socketpair = sock_no_socketpair,
2068 .accept = sock_no_accept,
2069 .getname = netlink_getname,
2070 .poll = datagram_poll,
2071 .ioctl = sock_no_ioctl,
2072 .listen = sock_no_listen,
2073 .shutdown = sock_no_shutdown,
2074 .setsockopt = netlink_setsockopt,
2075 .getsockopt = netlink_getsockopt,
2076 .sendmsg = netlink_sendmsg,
2077 .recvmsg = netlink_recvmsg,
2078 .mmap = sock_no_mmap,
2079 .sendpage = sock_no_sendpage,
2082 static const struct net_proto_family netlink_family_ops = {
2083 .family = PF_NETLINK,
2084 .create = netlink_create,
2085 .owner = THIS_MODULE, /* for consistency 8) */
2088 static int __net_init netlink_net_init(struct net *net)
2090 #ifdef CONFIG_PROC_FS
2091 if (!proc_net_fops_create(net, "netlink", 0, &netlink_seq_fops))
2097 static void __net_exit netlink_net_exit(struct net *net)
2099 #ifdef CONFIG_PROC_FS
2100 proc_net_remove(net, "netlink");
2104 static void __init netlink_add_usersock_entry(void)
2106 struct listeners *listeners;
2109 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
2111 panic("netlink_add_usersock_entry: Cannot allocate listeners\n");
2113 netlink_table_grab();
2115 nl_table[NETLINK_USERSOCK].groups = groups;
2116 rcu_assign_pointer(nl_table[NETLINK_USERSOCK].listeners, listeners);
2117 nl_table[NETLINK_USERSOCK].module = THIS_MODULE;
2118 nl_table[NETLINK_USERSOCK].registered = 1;
2120 netlink_table_ungrab();
2123 static struct pernet_operations __net_initdata netlink_net_ops = {
2124 .init = netlink_net_init,
2125 .exit = netlink_net_exit,
2128 static int __init netlink_proto_init(void)
2130 struct sk_buff *dummy_skb;
2132 unsigned long limit;
2134 int err = proto_register(&netlink_proto, 0);
2139 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
2141 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
2145 if (totalram_pages >= (128 * 1024))
2146 limit = totalram_pages >> (21 - PAGE_SHIFT);
2148 limit = totalram_pages >> (23 - PAGE_SHIFT);
2150 order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
2151 limit = (1UL << order) / sizeof(struct hlist_head);
2152 order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
2154 for (i = 0; i < MAX_LINKS; i++) {
2155 struct nl_pid_hash *hash = &nl_table[i].hash;
2157 hash->table = nl_pid_hash_zalloc(1 * sizeof(*hash->table));
2160 nl_pid_hash_free(nl_table[i].hash.table,
2161 1 * sizeof(*hash->table));
2165 hash->max_shift = order;
2168 hash->rehash_time = jiffies;
2171 netlink_add_usersock_entry();
2173 sock_register(&netlink_family_ops);
2174 register_pernet_subsys(&netlink_net_ops);
2175 /* The netlink device handler may be needed early. */
2180 panic("netlink_init: Cannot allocate nl_table\n");
2183 core_initcall(netlink_proto_init);