2 * NETLINK Kernel-user communication protocol.
4 * Authors: Alan Cox <alan@lxorguk.ukuu.org.uk>
5 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
12 * Tue Jun 26 14:36:48 MEST 2001 Herbert "herp" Rosmanith
13 * added netlink_proto_exit
14 * Tue Jan 22 18:32:44 BRST 2002 Arnaldo C. de Melo <acme@conectiva.com.br>
15 * use nlk_sk, as sk->protinfo is on a diet 8)
16 * Fri Jul 22 19:51:12 MEST 2005 Harald Welte <laforge@gnumonks.org>
17 * - inc module use count of module that owns
18 * the kernel socket in case userspace opens
19 * socket of same protocol
20 * - remove all module support, since netlink is
21 * mandatory if CONFIG_NET=y these days
24 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/kernel.h>
28 #include <linux/init.h>
29 #include <linux/signal.h>
30 #include <linux/sched.h>
31 #include <linux/errno.h>
32 #include <linux/string.h>
33 #include <linux/stat.h>
34 #include <linux/socket.h>
36 #include <linux/fcntl.h>
37 #include <linux/termios.h>
38 #include <linux/sockios.h>
39 #include <linux/net.h>
41 #include <linux/slab.h>
42 #include <asm/uaccess.h>
43 #include <linux/skbuff.h>
44 #include <linux/netdevice.h>
45 #include <linux/rtnetlink.h>
46 #include <linux/proc_fs.h>
47 #include <linux/seq_file.h>
48 #include <linux/notifier.h>
49 #include <linux/security.h>
50 #include <linux/jhash.h>
51 #include <linux/jiffies.h>
52 #include <linux/random.h>
53 #include <linux/bitops.h>
55 #include <linux/types.h>
56 #include <linux/audit.h>
57 #include <linux/mutex.h>
59 #include <net/net_namespace.h>
62 #include <net/netlink.h>
64 #define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
65 #define NLGRPLONGS(x) (NLGRPSZ(x)/sizeof(unsigned long))
68 /* struct sock has to be the first member of netlink_sock */
76 unsigned long *groups;
78 wait_queue_head_t wait;
79 struct netlink_callback *cb;
80 struct mutex *cb_mutex;
81 struct mutex cb_def_mutex;
82 void (*netlink_rcv)(struct sk_buff *skb);
83 struct module *module;
86 struct listeners_rcu_head {
87 struct rcu_head rcu_head;
91 #define NETLINK_KERNEL_SOCKET 0x1
92 #define NETLINK_RECV_PKTINFO 0x2
93 #define NETLINK_BROADCAST_SEND_ERROR 0x4
94 #define NETLINK_RECV_NO_ENOBUFS 0x8
96 static inline struct netlink_sock *nlk_sk(struct sock *sk)
98 return container_of(sk, struct netlink_sock, sk);
101 static inline int netlink_is_kernel(struct sock *sk)
103 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
107 struct hlist_head *table;
108 unsigned long rehash_time;
113 unsigned int entries;
114 unsigned int max_shift;
119 struct netlink_table {
120 struct nl_pid_hash hash;
121 struct hlist_head mc_list;
122 unsigned long *listeners;
123 unsigned int nl_nonroot;
125 struct mutex *cb_mutex;
126 struct module *module;
130 static struct netlink_table *nl_table;
132 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
134 static int netlink_dump(struct sock *sk);
135 static void netlink_destroy_callback(struct netlink_callback *cb);
137 static DEFINE_RWLOCK(nl_table_lock);
138 static atomic_t nl_table_users = ATOMIC_INIT(0);
140 static ATOMIC_NOTIFIER_HEAD(netlink_chain);
142 static u32 netlink_group_mask(u32 group)
144 return group ? 1 << (group - 1) : 0;
147 static struct hlist_head *nl_pid_hashfn(struct nl_pid_hash *hash, u32 pid)
149 return &hash->table[jhash_1word(pid, hash->rnd) & hash->mask];
152 static void netlink_sock_destruct(struct sock *sk)
154 struct netlink_sock *nlk = nlk_sk(sk);
158 nlk->cb->done(nlk->cb);
159 netlink_destroy_callback(nlk->cb);
162 skb_queue_purge(&sk->sk_receive_queue);
164 if (!sock_flag(sk, SOCK_DEAD)) {
165 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
169 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
170 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
171 WARN_ON(nlk_sk(sk)->groups);
174 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
175 * SMP. Look, when several writers sleep and reader wakes them up, all but one
176 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
177 * this, _but_ remember, it adds useless work on UP machines.
180 void netlink_table_grab(void)
181 __acquires(nl_table_lock)
185 write_lock_irq(&nl_table_lock);
187 if (atomic_read(&nl_table_users)) {
188 DECLARE_WAITQUEUE(wait, current);
190 add_wait_queue_exclusive(&nl_table_wait, &wait);
192 set_current_state(TASK_UNINTERRUPTIBLE);
193 if (atomic_read(&nl_table_users) == 0)
195 write_unlock_irq(&nl_table_lock);
197 write_lock_irq(&nl_table_lock);
200 __set_current_state(TASK_RUNNING);
201 remove_wait_queue(&nl_table_wait, &wait);
205 void netlink_table_ungrab(void)
206 __releases(nl_table_lock)
208 write_unlock_irq(&nl_table_lock);
209 wake_up(&nl_table_wait);
213 netlink_lock_table(void)
215 /* read_lock() synchronizes us to netlink_table_grab */
217 read_lock(&nl_table_lock);
218 atomic_inc(&nl_table_users);
219 read_unlock(&nl_table_lock);
223 netlink_unlock_table(void)
225 if (atomic_dec_and_test(&nl_table_users))
226 wake_up(&nl_table_wait);
229 static inline struct sock *netlink_lookup(struct net *net, int protocol,
232 struct nl_pid_hash *hash = &nl_table[protocol].hash;
233 struct hlist_head *head;
235 struct hlist_node *node;
237 read_lock(&nl_table_lock);
238 head = nl_pid_hashfn(hash, pid);
239 sk_for_each(sk, node, head) {
240 if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->pid == pid)) {
247 read_unlock(&nl_table_lock);
251 static inline struct hlist_head *nl_pid_hash_zalloc(size_t size)
253 if (size <= PAGE_SIZE)
254 return kzalloc(size, GFP_ATOMIC);
256 return (struct hlist_head *)
257 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
261 static inline void nl_pid_hash_free(struct hlist_head *table, size_t size)
263 if (size <= PAGE_SIZE)
266 free_pages((unsigned long)table, get_order(size));
269 static int nl_pid_hash_rehash(struct nl_pid_hash *hash, int grow)
271 unsigned int omask, mask, shift;
273 struct hlist_head *otable, *table;
276 omask = mask = hash->mask;
277 osize = size = (mask + 1) * sizeof(*table);
281 if (++shift > hash->max_shift)
287 table = nl_pid_hash_zalloc(size);
291 otable = hash->table;
295 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
297 for (i = 0; i <= omask; i++) {
299 struct hlist_node *node, *tmp;
301 sk_for_each_safe(sk, node, tmp, &otable[i])
302 __sk_add_node(sk, nl_pid_hashfn(hash, nlk_sk(sk)->pid));
305 nl_pid_hash_free(otable, osize);
306 hash->rehash_time = jiffies + 10 * 60 * HZ;
310 static inline int nl_pid_hash_dilute(struct nl_pid_hash *hash, int len)
312 int avg = hash->entries >> hash->shift;
314 if (unlikely(avg > 1) && nl_pid_hash_rehash(hash, 1))
317 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
318 nl_pid_hash_rehash(hash, 0);
325 static const struct proto_ops netlink_ops;
328 netlink_update_listeners(struct sock *sk)
330 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
331 struct hlist_node *node;
335 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
337 sk_for_each_bound(sk, node, &tbl->mc_list) {
338 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
339 mask |= nlk_sk(sk)->groups[i];
341 tbl->listeners[i] = mask;
343 /* this function is only called with the netlink table "grabbed", which
344 * makes sure updates are visible before bind or setsockopt return. */
347 static int netlink_insert(struct sock *sk, struct net *net, u32 pid)
349 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
350 struct hlist_head *head;
351 int err = -EADDRINUSE;
353 struct hlist_node *node;
356 netlink_table_grab();
357 head = nl_pid_hashfn(hash, pid);
359 sk_for_each(osk, node, head) {
360 if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->pid == pid))
372 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
375 if (len && nl_pid_hash_dilute(hash, len))
376 head = nl_pid_hashfn(hash, pid);
378 nlk_sk(sk)->pid = pid;
379 sk_add_node(sk, head);
383 netlink_table_ungrab();
387 static void netlink_remove(struct sock *sk)
389 netlink_table_grab();
390 if (sk_del_node_init(sk))
391 nl_table[sk->sk_protocol].hash.entries--;
392 if (nlk_sk(sk)->subscriptions)
393 __sk_del_bind_node(sk);
394 netlink_table_ungrab();
397 static struct proto netlink_proto = {
399 .owner = THIS_MODULE,
400 .obj_size = sizeof(struct netlink_sock),
403 static int __netlink_create(struct net *net, struct socket *sock,
404 struct mutex *cb_mutex, int protocol)
407 struct netlink_sock *nlk;
409 sock->ops = &netlink_ops;
411 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
415 sock_init_data(sock, sk);
419 nlk->cb_mutex = cb_mutex;
421 nlk->cb_mutex = &nlk->cb_def_mutex;
422 mutex_init(nlk->cb_mutex);
424 init_waitqueue_head(&nlk->wait);
426 sk->sk_destruct = netlink_sock_destruct;
427 sk->sk_protocol = protocol;
431 static int netlink_create(struct net *net, struct socket *sock, int protocol,
434 struct module *module = NULL;
435 struct mutex *cb_mutex;
436 struct netlink_sock *nlk;
439 sock->state = SS_UNCONNECTED;
441 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
442 return -ESOCKTNOSUPPORT;
444 if (protocol < 0 || protocol >= MAX_LINKS)
445 return -EPROTONOSUPPORT;
447 netlink_lock_table();
448 #ifdef CONFIG_MODULES
449 if (!nl_table[protocol].registered) {
450 netlink_unlock_table();
451 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
452 netlink_lock_table();
455 if (nl_table[protocol].registered &&
456 try_module_get(nl_table[protocol].module))
457 module = nl_table[protocol].module;
459 err = -EPROTONOSUPPORT;
460 cb_mutex = nl_table[protocol].cb_mutex;
461 netlink_unlock_table();
466 err = __netlink_create(net, sock, cb_mutex, protocol);
471 sock_prot_inuse_add(net, &netlink_proto, 1);
474 nlk = nlk_sk(sock->sk);
475 nlk->module = module;
484 static int netlink_release(struct socket *sock)
486 struct sock *sk = sock->sk;
487 struct netlink_sock *nlk;
497 * OK. Socket is unlinked, any packets that arrive now
502 wake_up_interruptible_all(&nlk->wait);
504 skb_queue_purge(&sk->sk_write_queue);
507 struct netlink_notify n = {
509 .protocol = sk->sk_protocol,
512 atomic_notifier_call_chain(&netlink_chain,
513 NETLINK_URELEASE, &n);
516 module_put(nlk->module);
518 netlink_table_grab();
519 if (netlink_is_kernel(sk)) {
520 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
521 if (--nl_table[sk->sk_protocol].registered == 0) {
522 kfree(nl_table[sk->sk_protocol].listeners);
523 nl_table[sk->sk_protocol].module = NULL;
524 nl_table[sk->sk_protocol].registered = 0;
526 } else if (nlk->subscriptions)
527 netlink_update_listeners(sk);
528 netlink_table_ungrab();
534 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
540 static int netlink_autobind(struct socket *sock)
542 struct sock *sk = sock->sk;
543 struct net *net = sock_net(sk);
544 struct nl_pid_hash *hash = &nl_table[sk->sk_protocol].hash;
545 struct hlist_head *head;
547 struct hlist_node *node;
548 s32 pid = current->tgid;
550 static s32 rover = -4097;
554 netlink_table_grab();
555 head = nl_pid_hashfn(hash, pid);
556 sk_for_each(osk, node, head) {
557 if (!net_eq(sock_net(osk), net))
559 if (nlk_sk(osk)->pid == pid) {
560 /* Bind collision, search negative pid values. */
564 netlink_table_ungrab();
568 netlink_table_ungrab();
570 err = netlink_insert(sk, net, pid);
571 if (err == -EADDRINUSE)
574 /* If 2 threads race to autobind, that is fine. */
581 static inline int netlink_capable(struct socket *sock, unsigned int flag)
583 return (nl_table[sock->sk->sk_protocol].nl_nonroot & flag) ||
584 capable(CAP_NET_ADMIN);
588 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
590 struct netlink_sock *nlk = nlk_sk(sk);
592 if (nlk->subscriptions && !subscriptions)
593 __sk_del_bind_node(sk);
594 else if (!nlk->subscriptions && subscriptions)
595 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
596 nlk->subscriptions = subscriptions;
599 static int netlink_realloc_groups(struct sock *sk)
601 struct netlink_sock *nlk = nlk_sk(sk);
603 unsigned long *new_groups;
606 netlink_table_grab();
608 groups = nl_table[sk->sk_protocol].groups;
609 if (!nl_table[sk->sk_protocol].registered) {
614 if (nlk->ngroups >= groups)
617 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
618 if (new_groups == NULL) {
622 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
623 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
625 nlk->groups = new_groups;
626 nlk->ngroups = groups;
628 netlink_table_ungrab();
632 static int netlink_bind(struct socket *sock, struct sockaddr *addr,
635 struct sock *sk = sock->sk;
636 struct net *net = sock_net(sk);
637 struct netlink_sock *nlk = nlk_sk(sk);
638 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
641 if (nladdr->nl_family != AF_NETLINK)
644 /* Only superuser is allowed to listen multicasts */
645 if (nladdr->nl_groups) {
646 if (!netlink_capable(sock, NL_NONROOT_RECV))
648 err = netlink_realloc_groups(sk);
654 if (nladdr->nl_pid != nlk->pid)
657 err = nladdr->nl_pid ?
658 netlink_insert(sk, net, nladdr->nl_pid) :
659 netlink_autobind(sock);
664 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
667 netlink_table_grab();
668 netlink_update_subscriptions(sk, nlk->subscriptions +
669 hweight32(nladdr->nl_groups) -
670 hweight32(nlk->groups[0]));
671 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
672 netlink_update_listeners(sk);
673 netlink_table_ungrab();
678 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
682 struct sock *sk = sock->sk;
683 struct netlink_sock *nlk = nlk_sk(sk);
684 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
686 if (addr->sa_family == AF_UNSPEC) {
687 sk->sk_state = NETLINK_UNCONNECTED;
692 if (addr->sa_family != AF_NETLINK)
695 /* Only superuser is allowed to send multicasts */
696 if (nladdr->nl_groups && !netlink_capable(sock, NL_NONROOT_SEND))
700 err = netlink_autobind(sock);
703 sk->sk_state = NETLINK_CONNECTED;
704 nlk->dst_pid = nladdr->nl_pid;
705 nlk->dst_group = ffs(nladdr->nl_groups);
711 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
712 int *addr_len, int peer)
714 struct sock *sk = sock->sk;
715 struct netlink_sock *nlk = nlk_sk(sk);
716 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
718 nladdr->nl_family = AF_NETLINK;
720 *addr_len = sizeof(*nladdr);
723 nladdr->nl_pid = nlk->dst_pid;
724 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
726 nladdr->nl_pid = nlk->pid;
727 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
732 static void netlink_overrun(struct sock *sk)
734 struct netlink_sock *nlk = nlk_sk(sk);
736 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
737 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
738 sk->sk_err = ENOBUFS;
739 sk->sk_error_report(sk);
742 atomic_inc(&sk->sk_drops);
745 static struct sock *netlink_getsockbypid(struct sock *ssk, u32 pid)
748 struct netlink_sock *nlk;
750 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, pid);
752 return ERR_PTR(-ECONNREFUSED);
754 /* Don't bother queuing skb if kernel socket has no input function */
756 if (sock->sk_state == NETLINK_CONNECTED &&
757 nlk->dst_pid != nlk_sk(ssk)->pid) {
759 return ERR_PTR(-ECONNREFUSED);
764 struct sock *netlink_getsockbyfilp(struct file *filp)
766 struct inode *inode = filp->f_path.dentry->d_inode;
769 if (!S_ISSOCK(inode->i_mode))
770 return ERR_PTR(-ENOTSOCK);
772 sock = SOCKET_I(inode)->sk;
773 if (sock->sk_family != AF_NETLINK)
774 return ERR_PTR(-EINVAL);
781 * Attach a skb to a netlink socket.
782 * The caller must hold a reference to the destination socket. On error, the
783 * reference is dropped. The skb is not send to the destination, just all
784 * all error checks are performed and memory in the queue is reserved.
786 * < 0: error. skb freed, reference to sock dropped.
788 * 1: repeat lookup - reference dropped while waiting for socket memory.
790 int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
791 long *timeo, struct sock *ssk)
793 struct netlink_sock *nlk;
797 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
798 test_bit(0, &nlk->state)) {
799 DECLARE_WAITQUEUE(wait, current);
801 if (!ssk || netlink_is_kernel(ssk))
808 __set_current_state(TASK_INTERRUPTIBLE);
809 add_wait_queue(&nlk->wait, &wait);
811 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
812 test_bit(0, &nlk->state)) &&
813 !sock_flag(sk, SOCK_DEAD))
814 *timeo = schedule_timeout(*timeo);
816 __set_current_state(TASK_RUNNING);
817 remove_wait_queue(&nlk->wait, &wait);
820 if (signal_pending(current)) {
822 return sock_intr_errno(*timeo);
826 skb_set_owner_r(skb, sk);
830 int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
834 skb_queue_tail(&sk->sk_receive_queue, skb);
835 sk->sk_data_ready(sk, len);
840 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
846 static inline struct sk_buff *netlink_trim(struct sk_buff *skb,
853 delta = skb->end - skb->tail;
854 if (delta * 2 < skb->truesize)
857 if (skb_shared(skb)) {
858 struct sk_buff *nskb = skb_clone(skb, allocation);
865 if (!pskb_expand_head(skb, 0, -delta, allocation))
866 skb->truesize -= delta;
871 static inline void netlink_rcv_wake(struct sock *sk)
873 struct netlink_sock *nlk = nlk_sk(sk);
875 if (skb_queue_empty(&sk->sk_receive_queue))
876 clear_bit(0, &nlk->state);
877 if (!test_bit(0, &nlk->state))
878 wake_up_interruptible(&nlk->wait);
881 static inline int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb)
884 struct netlink_sock *nlk = nlk_sk(sk);
887 if (nlk->netlink_rcv != NULL) {
889 skb_set_owner_r(skb, sk);
890 nlk->netlink_rcv(skb);
897 int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
898 u32 pid, int nonblock)
904 skb = netlink_trim(skb, gfp_any());
906 timeo = sock_sndtimeo(ssk, nonblock);
908 sk = netlink_getsockbypid(ssk, pid);
913 if (netlink_is_kernel(sk))
914 return netlink_unicast_kernel(sk, skb);
916 if (sk_filter(sk, skb)) {
923 err = netlink_attachskb(sk, skb, &timeo, ssk);
929 return netlink_sendskb(sk, skb);
931 EXPORT_SYMBOL(netlink_unicast);
933 int netlink_has_listeners(struct sock *sk, unsigned int group)
936 unsigned long *listeners;
938 BUG_ON(!netlink_is_kernel(sk));
941 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
943 if (group - 1 < nl_table[sk->sk_protocol].groups)
944 res = test_bit(group - 1, listeners);
950 EXPORT_SYMBOL_GPL(netlink_has_listeners);
952 static inline int netlink_broadcast_deliver(struct sock *sk,
955 struct netlink_sock *nlk = nlk_sk(sk);
957 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
958 !test_bit(0, &nlk->state)) {
959 skb_set_owner_r(skb, sk);
960 skb_queue_tail(&sk->sk_receive_queue, skb);
961 sk->sk_data_ready(sk, skb->len);
962 return atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf;
967 struct netlink_broadcast_data {
968 struct sock *exclude_sk;
973 int delivery_failure;
977 struct sk_buff *skb, *skb2;
980 static inline int do_one_broadcast(struct sock *sk,
981 struct netlink_broadcast_data *p)
983 struct netlink_sock *nlk = nlk_sk(sk);
986 if (p->exclude_sk == sk)
989 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
990 !test_bit(p->group - 1, nlk->groups))
993 if (!net_eq(sock_net(sk), p->net))
1002 if (p->skb2 == NULL) {
1003 if (skb_shared(p->skb)) {
1004 p->skb2 = skb_clone(p->skb, p->allocation);
1006 p->skb2 = skb_get(p->skb);
1008 * skb ownership may have been set when
1009 * delivered to a previous socket.
1011 skb_orphan(p->skb2);
1014 if (p->skb2 == NULL) {
1015 netlink_overrun(sk);
1016 /* Clone failed. Notify ALL listeners. */
1018 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1019 p->delivery_failure = 1;
1020 } else if (sk_filter(sk, p->skb2)) {
1023 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1024 netlink_overrun(sk);
1025 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1026 p->delivery_failure = 1;
1028 p->congested |= val;
1038 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 pid,
1039 u32 group, gfp_t allocation)
1041 struct net *net = sock_net(ssk);
1042 struct netlink_broadcast_data info;
1043 struct hlist_node *node;
1046 skb = netlink_trim(skb, allocation);
1048 info.exclude_sk = ssk;
1053 info.delivery_failure = 0;
1056 info.allocation = allocation;
1060 /* While we sleep in clone, do not allow to change socket list */
1062 netlink_lock_table();
1064 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1065 do_one_broadcast(sk, &info);
1069 netlink_unlock_table();
1071 kfree_skb(info.skb2);
1073 if (info.delivery_failure)
1076 if (info.delivered) {
1077 if (info.congested && (allocation & __GFP_WAIT))
1083 EXPORT_SYMBOL(netlink_broadcast);
1085 struct netlink_set_err_data {
1086 struct sock *exclude_sk;
1092 static inline int do_one_set_err(struct sock *sk,
1093 struct netlink_set_err_data *p)
1095 struct netlink_sock *nlk = nlk_sk(sk);
1098 if (sk == p->exclude_sk)
1101 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
1104 if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
1105 !test_bit(p->group - 1, nlk->groups))
1108 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
1113 sk->sk_err = p->code;
1114 sk->sk_error_report(sk);
1120 * netlink_set_err - report error to broadcast listeners
1121 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1122 * @pid: the PID of a process that we want to skip (if any)
1123 * @groups: the broadcast group that will notice the error
1124 * @code: error code, must be negative (as usual in kernelspace)
1126 * This function returns the number of broadcast listeners that have set the
1127 * NETLINK_RECV_NO_ENOBUFS socket option.
1129 int netlink_set_err(struct sock *ssk, u32 pid, u32 group, int code)
1131 struct netlink_set_err_data info;
1132 struct hlist_node *node;
1136 info.exclude_sk = ssk;
1139 /* sk->sk_err wants a positive error value */
1142 read_lock(&nl_table_lock);
1144 sk_for_each_bound(sk, node, &nl_table[ssk->sk_protocol].mc_list)
1145 ret += do_one_set_err(sk, &info);
1147 read_unlock(&nl_table_lock);
1150 EXPORT_SYMBOL(netlink_set_err);
1152 /* must be called with netlink table grabbed */
1153 static void netlink_update_socket_mc(struct netlink_sock *nlk,
1157 int old, new = !!is_new, subscriptions;
1159 old = test_bit(group - 1, nlk->groups);
1160 subscriptions = nlk->subscriptions - old + new;
1162 __set_bit(group - 1, nlk->groups);
1164 __clear_bit(group - 1, nlk->groups);
1165 netlink_update_subscriptions(&nlk->sk, subscriptions);
1166 netlink_update_listeners(&nlk->sk);
1169 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1170 char __user *optval, unsigned int optlen)
1172 struct sock *sk = sock->sk;
1173 struct netlink_sock *nlk = nlk_sk(sk);
1174 unsigned int val = 0;
1177 if (level != SOL_NETLINK)
1178 return -ENOPROTOOPT;
1180 if (optlen >= sizeof(int) &&
1181 get_user(val, (unsigned int __user *)optval))
1185 case NETLINK_PKTINFO:
1187 nlk->flags |= NETLINK_RECV_PKTINFO;
1189 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1192 case NETLINK_ADD_MEMBERSHIP:
1193 case NETLINK_DROP_MEMBERSHIP: {
1194 if (!netlink_capable(sock, NL_NONROOT_RECV))
1196 err = netlink_realloc_groups(sk);
1199 if (!val || val - 1 >= nlk->ngroups)
1201 netlink_table_grab();
1202 netlink_update_socket_mc(nlk, val,
1203 optname == NETLINK_ADD_MEMBERSHIP);
1204 netlink_table_ungrab();
1208 case NETLINK_BROADCAST_ERROR:
1210 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
1212 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
1215 case NETLINK_NO_ENOBUFS:
1217 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
1218 clear_bit(0, &nlk->state);
1219 wake_up_interruptible(&nlk->wait);
1221 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
1230 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1231 char __user *optval, int __user *optlen)
1233 struct sock *sk = sock->sk;
1234 struct netlink_sock *nlk = nlk_sk(sk);
1237 if (level != SOL_NETLINK)
1238 return -ENOPROTOOPT;
1240 if (get_user(len, optlen))
1246 case NETLINK_PKTINFO:
1247 if (len < sizeof(int))
1250 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
1251 if (put_user(len, optlen) ||
1252 put_user(val, optval))
1256 case NETLINK_BROADCAST_ERROR:
1257 if (len < sizeof(int))
1260 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
1261 if (put_user(len, optlen) ||
1262 put_user(val, optval))
1266 case NETLINK_NO_ENOBUFS:
1267 if (len < sizeof(int))
1270 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
1271 if (put_user(len, optlen) ||
1272 put_user(val, optval))
1282 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1284 struct nl_pktinfo info;
1286 info.group = NETLINK_CB(skb).dst_group;
1287 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1290 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1291 struct msghdr *msg, size_t len)
1293 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1294 struct sock *sk = sock->sk;
1295 struct netlink_sock *nlk = nlk_sk(sk);
1296 struct sockaddr_nl *addr = msg->msg_name;
1299 struct sk_buff *skb;
1301 struct scm_cookie scm;
1303 if (msg->msg_flags&MSG_OOB)
1306 if (NULL == siocb->scm)
1308 err = scm_send(sock, msg, siocb->scm);
1312 if (msg->msg_namelen) {
1313 if (addr->nl_family != AF_NETLINK)
1315 dst_pid = addr->nl_pid;
1316 dst_group = ffs(addr->nl_groups);
1317 if (dst_group && !netlink_capable(sock, NL_NONROOT_SEND))
1320 dst_pid = nlk->dst_pid;
1321 dst_group = nlk->dst_group;
1325 err = netlink_autobind(sock);
1331 if (len > sk->sk_sndbuf - 32)
1334 skb = alloc_skb(len, GFP_KERNEL);
1338 NETLINK_CB(skb).pid = nlk->pid;
1339 NETLINK_CB(skb).dst_group = dst_group;
1340 NETLINK_CB(skb).loginuid = audit_get_loginuid(current);
1341 NETLINK_CB(skb).sessionid = audit_get_sessionid(current);
1342 security_task_getsecid(current, &(NETLINK_CB(skb).sid));
1343 memcpy(NETLINK_CREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
1345 /* What can I do? Netlink is asynchronous, so that
1346 we will have to save current capabilities to
1347 check them, when this message will be delivered
1348 to corresponding kernel module. --ANK (980802)
1352 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1357 err = security_netlink_send(sk, skb);
1364 atomic_inc(&skb->users);
1365 netlink_broadcast(sk, skb, dst_pid, dst_group, GFP_KERNEL);
1367 err = netlink_unicast(sk, skb, dst_pid, msg->msg_flags&MSG_DONTWAIT);
1373 static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1374 struct msghdr *msg, size_t len,
1377 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1378 struct scm_cookie scm;
1379 struct sock *sk = sock->sk;
1380 struct netlink_sock *nlk = nlk_sk(sk);
1381 int noblock = flags&MSG_DONTWAIT;
1383 struct sk_buff *skb, *frag __maybe_unused = NULL;
1391 skb = skb_recv_datagram(sk, flags, noblock, &err);
1395 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1396 if (unlikely(skb_shinfo(skb)->frag_list)) {
1397 bool need_compat = !!(flags & MSG_CMSG_COMPAT);
1400 * If this skb has a frag_list, then here that means that
1401 * we will have to use the frag_list skb for compat tasks
1402 * and the regular skb for non-compat tasks.
1404 * The skb might (and likely will) be cloned, so we can't
1405 * just reset frag_list and go on with things -- we need to
1406 * keep that. For the compat case that's easy -- simply get
1407 * a reference to the compat skb and free the regular one
1408 * including the frag. For the non-compat case, we need to
1409 * avoid sending the frag to the user -- so assign NULL but
1410 * restore it below before freeing the skb.
1413 struct sk_buff *compskb = skb_shinfo(skb)->frag_list;
1418 frag = skb_shinfo(skb)->frag_list;
1419 skb_shinfo(skb)->frag_list = NULL;
1424 msg->msg_namelen = 0;
1428 msg->msg_flags |= MSG_TRUNC;
1432 skb_reset_transport_header(skb);
1433 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1435 if (msg->msg_name) {
1436 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
1437 addr->nl_family = AF_NETLINK;
1439 addr->nl_pid = NETLINK_CB(skb).pid;
1440 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1441 msg->msg_namelen = sizeof(*addr);
1444 if (nlk->flags & NETLINK_RECV_PKTINFO)
1445 netlink_cmsg_recv_pktinfo(msg, skb);
1447 if (NULL == siocb->scm) {
1448 memset(&scm, 0, sizeof(scm));
1451 siocb->scm->creds = *NETLINK_CREDS(skb);
1452 if (flags & MSG_TRUNC)
1455 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1456 skb_shinfo(skb)->frag_list = frag;
1459 skb_free_datagram(sk, skb);
1461 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2)
1464 scm_recv(sock, msg, siocb->scm, flags);
1466 netlink_rcv_wake(sk);
1467 return err ? : copied;
1470 static void netlink_data_ready(struct sock *sk, int len)
1476 * We export these functions to other modules. They provide a
1477 * complete set of kernel non-blocking support for message
1482 netlink_kernel_create(struct net *net, int unit, unsigned int groups,
1483 void (*input)(struct sk_buff *skb),
1484 struct mutex *cb_mutex, struct module *module)
1486 struct socket *sock;
1488 struct netlink_sock *nlk;
1489 unsigned long *listeners = NULL;
1493 if (unit < 0 || unit >= MAX_LINKS)
1496 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1500 * We have to just have a reference on the net from sk, but don't
1501 * get_net it. Besides, we cannot get and then put the net here.
1502 * So we create one inside init_net and the move it to net.
1505 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
1506 goto out_sock_release_nosk;
1509 sk_change_net(sk, net);
1514 listeners = kzalloc(NLGRPSZ(groups) + sizeof(struct listeners_rcu_head),
1517 goto out_sock_release;
1519 sk->sk_data_ready = netlink_data_ready;
1521 nlk_sk(sk)->netlink_rcv = input;
1523 if (netlink_insert(sk, net, 0))
1524 goto out_sock_release;
1527 nlk->flags |= NETLINK_KERNEL_SOCKET;
1529 netlink_table_grab();
1530 if (!nl_table[unit].registered) {
1531 nl_table[unit].groups = groups;
1532 nl_table[unit].listeners = listeners;
1533 nl_table[unit].cb_mutex = cb_mutex;
1534 nl_table[unit].module = module;
1535 nl_table[unit].registered = 1;
1538 nl_table[unit].registered++;
1540 netlink_table_ungrab();
1545 netlink_kernel_release(sk);
1548 out_sock_release_nosk:
1552 EXPORT_SYMBOL(netlink_kernel_create);
1556 netlink_kernel_release(struct sock *sk)
1558 sk_release_kernel(sk);
1560 EXPORT_SYMBOL(netlink_kernel_release);
1563 static void netlink_free_old_listeners(struct rcu_head *rcu_head)
1565 struct listeners_rcu_head *lrh;
1567 lrh = container_of(rcu_head, struct listeners_rcu_head, rcu_head);
1571 int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
1573 unsigned long *listeners, *old = NULL;
1574 struct listeners_rcu_head *old_rcu_head;
1575 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
1580 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
1581 listeners = kzalloc(NLGRPSZ(groups) +
1582 sizeof(struct listeners_rcu_head),
1586 old = tbl->listeners;
1587 memcpy(listeners, old, NLGRPSZ(tbl->groups));
1588 rcu_assign_pointer(tbl->listeners, listeners);
1590 * Free the old memory after an RCU grace period so we
1591 * don't leak it. We use call_rcu() here in order to be
1592 * able to call this function from atomic contexts. The
1593 * allocation of this memory will have reserved enough
1594 * space for struct listeners_rcu_head at the end.
1596 old_rcu_head = (void *)(tbl->listeners +
1597 NLGRPLONGS(tbl->groups));
1598 old_rcu_head->ptr = old;
1599 call_rcu(&old_rcu_head->rcu_head, netlink_free_old_listeners);
1601 tbl->groups = groups;
1607 * netlink_change_ngroups - change number of multicast groups
1609 * This changes the number of multicast groups that are available
1610 * on a certain netlink family. Note that it is not possible to
1611 * change the number of groups to below 32. Also note that it does
1612 * not implicitly call netlink_clear_multicast_users() when the
1613 * number of groups is reduced.
1615 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1616 * @groups: The new number of groups.
1618 int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1622 netlink_table_grab();
1623 err = __netlink_change_ngroups(sk, groups);
1624 netlink_table_ungrab();
1629 void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1632 struct hlist_node *node;
1633 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1635 sk_for_each_bound(sk, node, &tbl->mc_list)
1636 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1640 * netlink_clear_multicast_users - kick off multicast listeners
1642 * This function removes all listeners from the given group.
1643 * @ksk: The kernel netlink socket, as returned by
1644 * netlink_kernel_create().
1645 * @group: The multicast group to clear.
1647 void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1649 netlink_table_grab();
1650 __netlink_clear_multicast_users(ksk, group);
1651 netlink_table_ungrab();
1654 void netlink_set_nonroot(int protocol, unsigned int flags)
1656 if ((unsigned int)protocol < MAX_LINKS)
1657 nl_table[protocol].nl_nonroot = flags;
1659 EXPORT_SYMBOL(netlink_set_nonroot);
1661 static void netlink_destroy_callback(struct netlink_callback *cb)
1668 * It looks a bit ugly.
1669 * It would be better to create kernel thread.
1672 static int netlink_dump(struct sock *sk)
1674 struct netlink_sock *nlk = nlk_sk(sk);
1675 struct netlink_callback *cb;
1676 struct sk_buff *skb;
1677 struct nlmsghdr *nlh;
1678 int len, err = -ENOBUFS;
1680 skb = sock_rmalloc(sk, NLMSG_GOODSIZE, 0, GFP_KERNEL);
1684 mutex_lock(nlk->cb_mutex);
1692 len = cb->dump(skb, cb);
1695 mutex_unlock(nlk->cb_mutex);
1697 if (sk_filter(sk, skb))
1700 skb_queue_tail(&sk->sk_receive_queue, skb);
1701 sk->sk_data_ready(sk, skb->len);
1706 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1710 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1712 if (sk_filter(sk, skb))
1715 skb_queue_tail(&sk->sk_receive_queue, skb);
1716 sk->sk_data_ready(sk, skb->len);
1722 mutex_unlock(nlk->cb_mutex);
1724 netlink_destroy_callback(cb);
1728 mutex_unlock(nlk->cb_mutex);
1734 int netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1735 const struct nlmsghdr *nlh,
1736 int (*dump)(struct sk_buff *skb,
1737 struct netlink_callback *),
1738 int (*done)(struct netlink_callback *))
1740 struct netlink_callback *cb;
1742 struct netlink_sock *nlk;
1744 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
1751 atomic_inc(&skb->users);
1754 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).pid);
1756 netlink_destroy_callback(cb);
1757 return -ECONNREFUSED;
1760 /* A dump is in progress... */
1761 mutex_lock(nlk->cb_mutex);
1763 mutex_unlock(nlk->cb_mutex);
1764 netlink_destroy_callback(cb);
1769 mutex_unlock(nlk->cb_mutex);
1774 /* We successfully started a dump, by returning -EINTR we
1775 * signal not to send ACK even if it was requested.
1779 EXPORT_SYMBOL(netlink_dump_start);
1781 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1783 struct sk_buff *skb;
1784 struct nlmsghdr *rep;
1785 struct nlmsgerr *errmsg;
1786 size_t payload = sizeof(*errmsg);
1788 /* error messages get the original request appened */
1790 payload += nlmsg_len(nlh);
1792 skb = nlmsg_new(payload, GFP_KERNEL);
1796 sk = netlink_lookup(sock_net(in_skb->sk),
1797 in_skb->sk->sk_protocol,
1798 NETLINK_CB(in_skb).pid);
1800 sk->sk_err = ENOBUFS;
1801 sk->sk_error_report(sk);
1807 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).pid, nlh->nlmsg_seq,
1808 NLMSG_ERROR, payload, 0);
1809 errmsg = nlmsg_data(rep);
1810 errmsg->error = err;
1811 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
1812 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
1814 EXPORT_SYMBOL(netlink_ack);
1816 int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
1819 struct nlmsghdr *nlh;
1822 while (skb->len >= nlmsg_total_size(0)) {
1825 nlh = nlmsg_hdr(skb);
1828 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
1831 /* Only requests are handled by the kernel */
1832 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
1835 /* Skip control messages */
1836 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
1844 if (nlh->nlmsg_flags & NLM_F_ACK || err)
1845 netlink_ack(skb, nlh, err);
1848 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1849 if (msglen > skb->len)
1851 skb_pull(skb, msglen);
1856 EXPORT_SYMBOL(netlink_rcv_skb);
1859 * nlmsg_notify - send a notification netlink message
1860 * @sk: netlink socket to use
1861 * @skb: notification message
1862 * @pid: destination netlink pid for reports or 0
1863 * @group: destination multicast group or 0
1864 * @report: 1 to report back, 0 to disable
1865 * @flags: allocation flags
1867 int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 pid,
1868 unsigned int group, int report, gfp_t flags)
1873 int exclude_pid = 0;
1876 atomic_inc(&skb->users);
1880 /* errors reported via destination sk->sk_err, but propagate
1881 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
1882 err = nlmsg_multicast(sk, skb, exclude_pid, group, flags);
1888 err2 = nlmsg_unicast(sk, skb, pid);
1889 if (!err || err == -ESRCH)
1895 EXPORT_SYMBOL(nlmsg_notify);
1897 #ifdef CONFIG_PROC_FS
1898 struct nl_seq_iter {
1899 struct seq_net_private p;
1904 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1906 struct nl_seq_iter *iter = seq->private;
1909 struct hlist_node *node;
1912 for (i = 0; i < MAX_LINKS; i++) {
1913 struct nl_pid_hash *hash = &nl_table[i].hash;
1915 for (j = 0; j <= hash->mask; j++) {
1916 sk_for_each(s, node, &hash->table[j]) {
1917 if (sock_net(s) != seq_file_net(seq))
1931 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
1932 __acquires(nl_table_lock)
1934 read_lock(&nl_table_lock);
1935 return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1938 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1941 struct nl_seq_iter *iter;
1946 if (v == SEQ_START_TOKEN)
1947 return netlink_seq_socket_idx(seq, 0);
1949 iter = seq->private;
1953 } while (s && sock_net(s) != seq_file_net(seq));
1958 j = iter->hash_idx + 1;
1961 struct nl_pid_hash *hash = &nl_table[i].hash;
1963 for (; j <= hash->mask; j++) {
1964 s = sk_head(&hash->table[j]);
1965 while (s && sock_net(s) != seq_file_net(seq))
1975 } while (++i < MAX_LINKS);
1980 static void netlink_seq_stop(struct seq_file *seq, void *v)
1981 __releases(nl_table_lock)
1983 read_unlock(&nl_table_lock);
1987 static int netlink_seq_show(struct seq_file *seq, void *v)
1989 if (v == SEQ_START_TOKEN)
1991 "sk Eth Pid Groups "
1992 "Rmem Wmem Dump Locks Drops Inode\n");
1995 struct netlink_sock *nlk = nlk_sk(s);
1997 seq_printf(seq, "%p %-3d %-6d %08x %-8d %-8d %p %-8d %-8d %-8lu\n",
2001 nlk->groups ? (u32)nlk->groups[0] : 0,
2002 sk_rmem_alloc_get(s),
2003 sk_wmem_alloc_get(s),
2005 atomic_read(&s->sk_refcnt),
2006 atomic_read(&s->sk_drops),
2014 static const struct seq_operations netlink_seq_ops = {
2015 .start = netlink_seq_start,
2016 .next = netlink_seq_next,
2017 .stop = netlink_seq_stop,
2018 .show = netlink_seq_show,
2022 static int netlink_seq_open(struct inode *inode, struct file *file)
2024 return seq_open_net(inode, file, &netlink_seq_ops,
2025 sizeof(struct nl_seq_iter));
2028 static const struct file_operations netlink_seq_fops = {
2029 .owner = THIS_MODULE,
2030 .open = netlink_seq_open,
2032 .llseek = seq_lseek,
2033 .release = seq_release_net,
2038 int netlink_register_notifier(struct notifier_block *nb)
2040 return atomic_notifier_chain_register(&netlink_chain, nb);
2042 EXPORT_SYMBOL(netlink_register_notifier);
2044 int netlink_unregister_notifier(struct notifier_block *nb)
2046 return atomic_notifier_chain_unregister(&netlink_chain, nb);
2048 EXPORT_SYMBOL(netlink_unregister_notifier);
2050 static const struct proto_ops netlink_ops = {
2051 .family = PF_NETLINK,
2052 .owner = THIS_MODULE,
2053 .release = netlink_release,
2054 .bind = netlink_bind,
2055 .connect = netlink_connect,
2056 .socketpair = sock_no_socketpair,
2057 .accept = sock_no_accept,
2058 .getname = netlink_getname,
2059 .poll = datagram_poll,
2060 .ioctl = sock_no_ioctl,
2061 .listen = sock_no_listen,
2062 .shutdown = sock_no_shutdown,
2063 .setsockopt = netlink_setsockopt,
2064 .getsockopt = netlink_getsockopt,
2065 .sendmsg = netlink_sendmsg,
2066 .recvmsg = netlink_recvmsg,
2067 .mmap = sock_no_mmap,
2068 .sendpage = sock_no_sendpage,
2071 static const struct net_proto_family netlink_family_ops = {
2072 .family = PF_NETLINK,
2073 .create = netlink_create,
2074 .owner = THIS_MODULE, /* for consistency 8) */
2077 static int __net_init netlink_net_init(struct net *net)
2079 #ifdef CONFIG_PROC_FS
2080 if (!proc_net_fops_create(net, "netlink", 0, &netlink_seq_fops))
2086 static void __net_exit netlink_net_exit(struct net *net)
2088 #ifdef CONFIG_PROC_FS
2089 proc_net_remove(net, "netlink");
2093 static struct pernet_operations __net_initdata netlink_net_ops = {
2094 .init = netlink_net_init,
2095 .exit = netlink_net_exit,
2098 static int __init netlink_proto_init(void)
2100 struct sk_buff *dummy_skb;
2102 unsigned long limit;
2104 int err = proto_register(&netlink_proto, 0);
2109 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > sizeof(dummy_skb->cb));
2111 nl_table = kcalloc(MAX_LINKS, sizeof(*nl_table), GFP_KERNEL);
2115 if (totalram_pages >= (128 * 1024))
2116 limit = totalram_pages >> (21 - PAGE_SHIFT);
2118 limit = totalram_pages >> (23 - PAGE_SHIFT);
2120 order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
2121 limit = (1UL << order) / sizeof(struct hlist_head);
2122 order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
2124 for (i = 0; i < MAX_LINKS; i++) {
2125 struct nl_pid_hash *hash = &nl_table[i].hash;
2127 hash->table = nl_pid_hash_zalloc(1 * sizeof(*hash->table));
2130 nl_pid_hash_free(nl_table[i].hash.table,
2131 1 * sizeof(*hash->table));
2135 hash->max_shift = order;
2138 hash->rehash_time = jiffies;
2141 sock_register(&netlink_family_ops);
2142 register_pernet_subsys(&netlink_net_ops);
2143 /* The netlink device handler may be needed early. */
2148 panic("netlink_init: Cannot allocate nl_table\n");
2151 core_initcall(netlink_proto_init);