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 #include "af_netlink.h"
68 unsigned long masks[0];
71 #define NETLINK_KERNEL_SOCKET 0x1
72 #define NETLINK_RECV_PKTINFO 0x2
73 #define NETLINK_BROADCAST_SEND_ERROR 0x4
74 #define NETLINK_RECV_NO_ENOBUFS 0x8
76 static inline int netlink_is_kernel(struct sock *sk)
78 return nlk_sk(sk)->flags & NETLINK_KERNEL_SOCKET;
81 struct netlink_table *nl_table;
82 EXPORT_SYMBOL_GPL(nl_table);
84 static DECLARE_WAIT_QUEUE_HEAD(nl_table_wait);
86 static int netlink_dump(struct sock *sk);
88 DEFINE_RWLOCK(nl_table_lock);
89 EXPORT_SYMBOL_GPL(nl_table_lock);
90 static atomic_t nl_table_users = ATOMIC_INIT(0);
92 #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
94 static ATOMIC_NOTIFIER_HEAD(netlink_chain);
96 static inline u32 netlink_group_mask(u32 group)
98 return group ? 1 << (group - 1) : 0;
101 static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid)
103 return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask];
106 static void netlink_destroy_callback(struct netlink_callback *cb)
112 static void netlink_consume_callback(struct netlink_callback *cb)
114 consume_skb(cb->skb);
118 static void netlink_sock_destruct(struct sock *sk)
120 struct netlink_sock *nlk = nlk_sk(sk);
124 nlk->cb->done(nlk->cb);
126 module_put(nlk->cb->module);
127 netlink_destroy_callback(nlk->cb);
130 skb_queue_purge(&sk->sk_receive_queue);
132 if (!sock_flag(sk, SOCK_DEAD)) {
133 printk(KERN_ERR "Freeing alive netlink socket %p\n", sk);
137 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
138 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
139 WARN_ON(nlk_sk(sk)->groups);
142 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on
143 * SMP. Look, when several writers sleep and reader wakes them up, all but one
144 * immediately hit write lock and grab all the cpus. Exclusive sleep solves
145 * this, _but_ remember, it adds useless work on UP machines.
148 void netlink_table_grab(void)
149 __acquires(nl_table_lock)
153 write_lock_irq(&nl_table_lock);
155 if (atomic_read(&nl_table_users)) {
156 DECLARE_WAITQUEUE(wait, current);
158 add_wait_queue_exclusive(&nl_table_wait, &wait);
160 set_current_state(TASK_UNINTERRUPTIBLE);
161 if (atomic_read(&nl_table_users) == 0)
163 write_unlock_irq(&nl_table_lock);
165 write_lock_irq(&nl_table_lock);
168 __set_current_state(TASK_RUNNING);
169 remove_wait_queue(&nl_table_wait, &wait);
173 void netlink_table_ungrab(void)
174 __releases(nl_table_lock)
176 write_unlock_irq(&nl_table_lock);
177 wake_up(&nl_table_wait);
181 netlink_lock_table(void)
183 /* read_lock() synchronizes us to netlink_table_grab */
185 read_lock(&nl_table_lock);
186 atomic_inc(&nl_table_users);
187 read_unlock(&nl_table_lock);
191 netlink_unlock_table(void)
193 if (atomic_dec_and_test(&nl_table_users))
194 wake_up(&nl_table_wait);
197 static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
199 struct nl_portid_hash *hash = &nl_table[protocol].hash;
200 struct hlist_head *head;
203 read_lock(&nl_table_lock);
204 head = nl_portid_hashfn(hash, portid);
205 sk_for_each(sk, head) {
206 if (net_eq(sock_net(sk), net) && (nlk_sk(sk)->portid == portid)) {
213 read_unlock(&nl_table_lock);
217 static struct hlist_head *nl_portid_hash_zalloc(size_t size)
219 if (size <= PAGE_SIZE)
220 return kzalloc(size, GFP_ATOMIC);
222 return (struct hlist_head *)
223 __get_free_pages(GFP_ATOMIC | __GFP_ZERO,
227 static void nl_portid_hash_free(struct hlist_head *table, size_t size)
229 if (size <= PAGE_SIZE)
232 free_pages((unsigned long)table, get_order(size));
235 static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow)
237 unsigned int omask, mask, shift;
239 struct hlist_head *otable, *table;
242 omask = mask = hash->mask;
243 osize = size = (mask + 1) * sizeof(*table);
247 if (++shift > hash->max_shift)
253 table = nl_portid_hash_zalloc(size);
257 otable = hash->table;
261 get_random_bytes(&hash->rnd, sizeof(hash->rnd));
263 for (i = 0; i <= omask; i++) {
265 struct hlist_node *tmp;
267 sk_for_each_safe(sk, tmp, &otable[i])
268 __sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid));
271 nl_portid_hash_free(otable, osize);
272 hash->rehash_time = jiffies + 10 * 60 * HZ;
276 static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len)
278 int avg = hash->entries >> hash->shift;
280 if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1))
283 if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
284 nl_portid_hash_rehash(hash, 0);
291 static const struct proto_ops netlink_ops;
294 netlink_update_listeners(struct sock *sk)
296 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
299 struct listeners *listeners;
301 listeners = nl_deref_protected(tbl->listeners);
305 for (i = 0; i < NLGRPLONGS(tbl->groups); i++) {
307 sk_for_each_bound(sk, &tbl->mc_list) {
308 if (i < NLGRPLONGS(nlk_sk(sk)->ngroups))
309 mask |= nlk_sk(sk)->groups[i];
311 listeners->masks[i] = mask;
313 /* this function is only called with the netlink table "grabbed", which
314 * makes sure updates are visible before bind or setsockopt return. */
317 static int netlink_insert(struct sock *sk, struct net *net, u32 portid)
319 struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash;
320 struct hlist_head *head;
321 int err = -EADDRINUSE;
325 netlink_table_grab();
326 head = nl_portid_hashfn(hash, portid);
328 sk_for_each(osk, head) {
329 if (net_eq(sock_net(osk), net) && (nlk_sk(osk)->portid == portid))
337 if (nlk_sk(sk)->portid)
341 if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
344 if (len && nl_portid_hash_dilute(hash, len))
345 head = nl_portid_hashfn(hash, portid);
347 nlk_sk(sk)->portid = portid;
348 sk_add_node(sk, head);
352 netlink_table_ungrab();
356 static void netlink_remove(struct sock *sk)
358 netlink_table_grab();
359 if (sk_del_node_init(sk))
360 nl_table[sk->sk_protocol].hash.entries--;
361 if (nlk_sk(sk)->subscriptions)
362 __sk_del_bind_node(sk);
363 netlink_table_ungrab();
366 static struct proto netlink_proto = {
368 .owner = THIS_MODULE,
369 .obj_size = sizeof(struct netlink_sock),
372 static int __netlink_create(struct net *net, struct socket *sock,
373 struct mutex *cb_mutex, int protocol)
376 struct netlink_sock *nlk;
378 sock->ops = &netlink_ops;
380 sk = sk_alloc(net, PF_NETLINK, GFP_KERNEL, &netlink_proto);
384 sock_init_data(sock, sk);
388 nlk->cb_mutex = cb_mutex;
390 nlk->cb_mutex = &nlk->cb_def_mutex;
391 mutex_init(nlk->cb_mutex);
393 init_waitqueue_head(&nlk->wait);
395 sk->sk_destruct = netlink_sock_destruct;
396 sk->sk_protocol = protocol;
400 static int netlink_create(struct net *net, struct socket *sock, int protocol,
403 struct module *module = NULL;
404 struct mutex *cb_mutex;
405 struct netlink_sock *nlk;
406 void (*bind)(int group);
409 sock->state = SS_UNCONNECTED;
411 if (sock->type != SOCK_RAW && sock->type != SOCK_DGRAM)
412 return -ESOCKTNOSUPPORT;
414 if (protocol < 0 || protocol >= MAX_LINKS)
415 return -EPROTONOSUPPORT;
417 netlink_lock_table();
418 #ifdef CONFIG_MODULES
419 if (!nl_table[protocol].registered) {
420 netlink_unlock_table();
421 request_module("net-pf-%d-proto-%d", PF_NETLINK, protocol);
422 netlink_lock_table();
425 if (nl_table[protocol].registered &&
426 try_module_get(nl_table[protocol].module))
427 module = nl_table[protocol].module;
429 err = -EPROTONOSUPPORT;
430 cb_mutex = nl_table[protocol].cb_mutex;
431 bind = nl_table[protocol].bind;
432 netlink_unlock_table();
437 err = __netlink_create(net, sock, cb_mutex, protocol);
442 sock_prot_inuse_add(net, &netlink_proto, 1);
445 nlk = nlk_sk(sock->sk);
446 nlk->module = module;
447 nlk->netlink_bind = bind;
456 static int netlink_release(struct socket *sock)
458 struct sock *sk = sock->sk;
459 struct netlink_sock *nlk;
469 * OK. Socket is unlinked, any packets that arrive now
474 wake_up_interruptible_all(&nlk->wait);
476 skb_queue_purge(&sk->sk_write_queue);
479 struct netlink_notify n = {
481 .protocol = sk->sk_protocol,
482 .portid = nlk->portid,
484 atomic_notifier_call_chain(&netlink_chain,
485 NETLINK_URELEASE, &n);
488 module_put(nlk->module);
490 netlink_table_grab();
491 if (netlink_is_kernel(sk)) {
492 BUG_ON(nl_table[sk->sk_protocol].registered == 0);
493 if (--nl_table[sk->sk_protocol].registered == 0) {
494 struct listeners *old;
496 old = nl_deref_protected(nl_table[sk->sk_protocol].listeners);
497 RCU_INIT_POINTER(nl_table[sk->sk_protocol].listeners, NULL);
499 nl_table[sk->sk_protocol].module = NULL;
500 nl_table[sk->sk_protocol].bind = NULL;
501 nl_table[sk->sk_protocol].flags = 0;
502 nl_table[sk->sk_protocol].registered = 0;
504 } else if (nlk->subscriptions) {
505 netlink_update_listeners(sk);
507 netlink_table_ungrab();
513 sock_prot_inuse_add(sock_net(sk), &netlink_proto, -1);
519 static int netlink_autobind(struct socket *sock)
521 struct sock *sk = sock->sk;
522 struct net *net = sock_net(sk);
523 struct nl_portid_hash *hash = &nl_table[sk->sk_protocol].hash;
524 struct hlist_head *head;
526 s32 portid = task_tgid_vnr(current);
528 static s32 rover = -4097;
532 netlink_table_grab();
533 head = nl_portid_hashfn(hash, portid);
534 sk_for_each(osk, head) {
535 if (!net_eq(sock_net(osk), net))
537 if (nlk_sk(osk)->portid == portid) {
538 /* Bind collision, search negative portid values. */
542 netlink_table_ungrab();
546 netlink_table_ungrab();
548 err = netlink_insert(sk, net, portid);
549 if (err == -EADDRINUSE)
552 /* If 2 threads race to autobind, that is fine. */
559 static inline int netlink_capable(const struct socket *sock, unsigned int flag)
561 return (nl_table[sock->sk->sk_protocol].flags & flag) ||
562 ns_capable(sock_net(sock->sk)->user_ns, CAP_NET_ADMIN);
566 netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
568 struct netlink_sock *nlk = nlk_sk(sk);
570 if (nlk->subscriptions && !subscriptions)
571 __sk_del_bind_node(sk);
572 else if (!nlk->subscriptions && subscriptions)
573 sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
574 nlk->subscriptions = subscriptions;
577 static int netlink_realloc_groups(struct sock *sk)
579 struct netlink_sock *nlk = nlk_sk(sk);
581 unsigned long *new_groups;
584 netlink_table_grab();
586 groups = nl_table[sk->sk_protocol].groups;
587 if (!nl_table[sk->sk_protocol].registered) {
592 if (nlk->ngroups >= groups)
595 new_groups = krealloc(nlk->groups, NLGRPSZ(groups), GFP_ATOMIC);
596 if (new_groups == NULL) {
600 memset((char *)new_groups + NLGRPSZ(nlk->ngroups), 0,
601 NLGRPSZ(groups) - NLGRPSZ(nlk->ngroups));
603 nlk->groups = new_groups;
604 nlk->ngroups = groups;
606 netlink_table_ungrab();
610 static int netlink_bind(struct socket *sock, struct sockaddr *addr,
613 struct sock *sk = sock->sk;
614 struct net *net = sock_net(sk);
615 struct netlink_sock *nlk = nlk_sk(sk);
616 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
619 if (addr_len < sizeof(struct sockaddr_nl))
622 if (nladdr->nl_family != AF_NETLINK)
625 /* Only superuser is allowed to listen multicasts */
626 if (nladdr->nl_groups) {
627 if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV))
629 err = netlink_realloc_groups(sk);
635 if (nladdr->nl_pid != nlk->portid)
638 err = nladdr->nl_pid ?
639 netlink_insert(sk, net, nladdr->nl_pid) :
640 netlink_autobind(sock);
645 if (!nladdr->nl_groups && (nlk->groups == NULL || !(u32)nlk->groups[0]))
648 netlink_table_grab();
649 netlink_update_subscriptions(sk, nlk->subscriptions +
650 hweight32(nladdr->nl_groups) -
651 hweight32(nlk->groups[0]));
652 nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups;
653 netlink_update_listeners(sk);
654 netlink_table_ungrab();
656 if (nlk->netlink_bind && nlk->groups[0]) {
659 for (i=0; i<nlk->ngroups; i++) {
660 if (test_bit(i, nlk->groups))
661 nlk->netlink_bind(i);
668 static int netlink_connect(struct socket *sock, struct sockaddr *addr,
672 struct sock *sk = sock->sk;
673 struct netlink_sock *nlk = nlk_sk(sk);
674 struct sockaddr_nl *nladdr = (struct sockaddr_nl *)addr;
676 if (alen < sizeof(addr->sa_family))
679 if (addr->sa_family == AF_UNSPEC) {
680 sk->sk_state = NETLINK_UNCONNECTED;
685 if (addr->sa_family != AF_NETLINK)
688 /* Only superuser is allowed to send multicasts */
689 if (nladdr->nl_groups && !netlink_capable(sock, NL_CFG_F_NONROOT_SEND))
693 err = netlink_autobind(sock);
696 sk->sk_state = NETLINK_CONNECTED;
697 nlk->dst_portid = nladdr->nl_pid;
698 nlk->dst_group = ffs(nladdr->nl_groups);
704 static int netlink_getname(struct socket *sock, struct sockaddr *addr,
705 int *addr_len, int peer)
707 struct sock *sk = sock->sk;
708 struct netlink_sock *nlk = nlk_sk(sk);
709 DECLARE_SOCKADDR(struct sockaddr_nl *, nladdr, addr);
711 nladdr->nl_family = AF_NETLINK;
713 *addr_len = sizeof(*nladdr);
716 nladdr->nl_pid = nlk->dst_portid;
717 nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
719 nladdr->nl_pid = nlk->portid;
720 nladdr->nl_groups = nlk->groups ? nlk->groups[0] : 0;
725 static void netlink_overrun(struct sock *sk)
727 struct netlink_sock *nlk = nlk_sk(sk);
729 if (!(nlk->flags & NETLINK_RECV_NO_ENOBUFS)) {
730 if (!test_and_set_bit(0, &nlk_sk(sk)->state)) {
731 sk->sk_err = ENOBUFS;
732 sk->sk_error_report(sk);
735 atomic_inc(&sk->sk_drops);
738 static struct sock *netlink_getsockbyportid(struct sock *ssk, u32 portid)
741 struct netlink_sock *nlk;
743 sock = netlink_lookup(sock_net(ssk), ssk->sk_protocol, portid);
745 return ERR_PTR(-ECONNREFUSED);
747 /* Don't bother queuing skb if kernel socket has no input function */
749 if (sock->sk_state == NETLINK_CONNECTED &&
750 nlk->dst_portid != nlk_sk(ssk)->portid) {
752 return ERR_PTR(-ECONNREFUSED);
757 struct sock *netlink_getsockbyfilp(struct file *filp)
759 struct inode *inode = file_inode(filp);
762 if (!S_ISSOCK(inode->i_mode))
763 return ERR_PTR(-ENOTSOCK);
765 sock = SOCKET_I(inode)->sk;
766 if (sock->sk_family != AF_NETLINK)
767 return ERR_PTR(-EINVAL);
774 * Attach a skb to a netlink socket.
775 * The caller must hold a reference to the destination socket. On error, the
776 * reference is dropped. The skb is not send to the destination, just all
777 * all error checks are performed and memory in the queue is reserved.
779 * < 0: error. skb freed, reference to sock dropped.
781 * 1: repeat lookup - reference dropped while waiting for socket memory.
783 int netlink_attachskb(struct sock *sk, struct sk_buff *skb,
784 long *timeo, struct sock *ssk)
786 struct netlink_sock *nlk;
790 if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
791 test_bit(0, &nlk->state)) {
792 DECLARE_WAITQUEUE(wait, current);
794 if (!ssk || netlink_is_kernel(ssk))
801 __set_current_state(TASK_INTERRUPTIBLE);
802 add_wait_queue(&nlk->wait, &wait);
804 if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf ||
805 test_bit(0, &nlk->state)) &&
806 !sock_flag(sk, SOCK_DEAD))
807 *timeo = schedule_timeout(*timeo);
809 __set_current_state(TASK_RUNNING);
810 remove_wait_queue(&nlk->wait, &wait);
813 if (signal_pending(current)) {
815 return sock_intr_errno(*timeo);
819 skb_set_owner_r(skb, sk);
823 static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb)
827 skb_queue_tail(&sk->sk_receive_queue, skb);
828 sk->sk_data_ready(sk, len);
832 int netlink_sendskb(struct sock *sk, struct sk_buff *skb)
834 int len = __netlink_sendskb(sk, skb);
840 void netlink_detachskb(struct sock *sk, struct sk_buff *skb)
846 static struct sk_buff *netlink_trim(struct sk_buff *skb, gfp_t allocation)
852 delta = skb->end - skb->tail;
853 if (delta * 2 < skb->truesize)
856 if (skb_shared(skb)) {
857 struct sk_buff *nskb = skb_clone(skb, allocation);
864 if (!pskb_expand_head(skb, 0, -delta, allocation))
865 skb->truesize -= delta;
870 static void netlink_rcv_wake(struct sock *sk)
872 struct netlink_sock *nlk = nlk_sk(sk);
874 if (skb_queue_empty(&sk->sk_receive_queue))
875 clear_bit(0, &nlk->state);
876 if (!test_bit(0, &nlk->state))
877 wake_up_interruptible(&nlk->wait);
880 static 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 NETLINK_CB(skb).ssk = ssk;
891 nlk->netlink_rcv(skb);
900 int netlink_unicast(struct sock *ssk, struct sk_buff *skb,
901 u32 portid, int nonblock)
907 skb = netlink_trim(skb, gfp_any());
909 timeo = sock_sndtimeo(ssk, nonblock);
911 sk = netlink_getsockbyportid(ssk, portid);
916 if (netlink_is_kernel(sk))
917 return netlink_unicast_kernel(sk, skb, ssk);
919 if (sk_filter(sk, skb)) {
926 err = netlink_attachskb(sk, skb, &timeo, ssk);
932 return netlink_sendskb(sk, skb);
934 EXPORT_SYMBOL(netlink_unicast);
936 int netlink_has_listeners(struct sock *sk, unsigned int group)
939 struct listeners *listeners;
941 BUG_ON(!netlink_is_kernel(sk));
944 listeners = rcu_dereference(nl_table[sk->sk_protocol].listeners);
946 if (listeners && group - 1 < nl_table[sk->sk_protocol].groups)
947 res = test_bit(group - 1, listeners->masks);
953 EXPORT_SYMBOL_GPL(netlink_has_listeners);
955 static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb)
957 struct netlink_sock *nlk = nlk_sk(sk);
959 if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf &&
960 !test_bit(0, &nlk->state)) {
961 skb_set_owner_r(skb, sk);
962 __netlink_sendskb(sk, skb);
963 return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1);
968 struct netlink_broadcast_data {
969 struct sock *exclude_sk;
974 int delivery_failure;
978 struct sk_buff *skb, *skb2;
979 int (*tx_filter)(struct sock *dsk, struct sk_buff *skb, void *data);
983 static int do_one_broadcast(struct sock *sk,
984 struct netlink_broadcast_data *p)
986 struct netlink_sock *nlk = nlk_sk(sk);
989 if (p->exclude_sk == sk)
992 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
993 !test_bit(p->group - 1, nlk->groups))
996 if (!net_eq(sock_net(sk), p->net))
1000 netlink_overrun(sk);
1005 if (p->skb2 == NULL) {
1006 if (skb_shared(p->skb)) {
1007 p->skb2 = skb_clone(p->skb, p->allocation);
1009 p->skb2 = skb_get(p->skb);
1011 * skb ownership may have been set when
1012 * delivered to a previous socket.
1014 skb_orphan(p->skb2);
1017 if (p->skb2 == NULL) {
1018 netlink_overrun(sk);
1019 /* Clone failed. Notify ALL listeners. */
1021 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1022 p->delivery_failure = 1;
1023 } else if (p->tx_filter && p->tx_filter(sk, p->skb2, p->tx_data)) {
1026 } else if (sk_filter(sk, p->skb2)) {
1029 } else if ((val = netlink_broadcast_deliver(sk, p->skb2)) < 0) {
1030 netlink_overrun(sk);
1031 if (nlk->flags & NETLINK_BROADCAST_SEND_ERROR)
1032 p->delivery_failure = 1;
1034 p->congested |= val;
1044 int netlink_broadcast_filtered(struct sock *ssk, struct sk_buff *skb, u32 portid,
1045 u32 group, gfp_t allocation,
1046 int (*filter)(struct sock *dsk, struct sk_buff *skb, void *data),
1049 struct net *net = sock_net(ssk);
1050 struct netlink_broadcast_data info;
1053 skb = netlink_trim(skb, allocation);
1055 info.exclude_sk = ssk;
1057 info.portid = portid;
1060 info.delivery_failure = 0;
1063 info.allocation = allocation;
1066 info.tx_filter = filter;
1067 info.tx_data = filter_data;
1069 /* While we sleep in clone, do not allow to change socket list */
1071 netlink_lock_table();
1073 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1074 do_one_broadcast(sk, &info);
1078 netlink_unlock_table();
1080 if (info.delivery_failure) {
1081 kfree_skb(info.skb2);
1084 consume_skb(info.skb2);
1086 if (info.delivered) {
1087 if (info.congested && (allocation & __GFP_WAIT))
1093 EXPORT_SYMBOL(netlink_broadcast_filtered);
1095 int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, u32 portid,
1096 u32 group, gfp_t allocation)
1098 return netlink_broadcast_filtered(ssk, skb, portid, group, allocation,
1101 EXPORT_SYMBOL(netlink_broadcast);
1103 struct netlink_set_err_data {
1104 struct sock *exclude_sk;
1110 static int do_one_set_err(struct sock *sk, struct netlink_set_err_data *p)
1112 struct netlink_sock *nlk = nlk_sk(sk);
1115 if (sk == p->exclude_sk)
1118 if (!net_eq(sock_net(sk), sock_net(p->exclude_sk)))
1121 if (nlk->portid == p->portid || p->group - 1 >= nlk->ngroups ||
1122 !test_bit(p->group - 1, nlk->groups))
1125 if (p->code == ENOBUFS && nlk->flags & NETLINK_RECV_NO_ENOBUFS) {
1130 sk->sk_err = p->code;
1131 sk->sk_error_report(sk);
1137 * netlink_set_err - report error to broadcast listeners
1138 * @ssk: the kernel netlink socket, as returned by netlink_kernel_create()
1139 * @portid: the PORTID of a process that we want to skip (if any)
1140 * @groups: the broadcast group that will notice the error
1141 * @code: error code, must be negative (as usual in kernelspace)
1143 * This function returns the number of broadcast listeners that have set the
1144 * NETLINK_RECV_NO_ENOBUFS socket option.
1146 int netlink_set_err(struct sock *ssk, u32 portid, u32 group, int code)
1148 struct netlink_set_err_data info;
1152 info.exclude_sk = ssk;
1153 info.portid = portid;
1155 /* sk->sk_err wants a positive error value */
1158 read_lock(&nl_table_lock);
1160 sk_for_each_bound(sk, &nl_table[ssk->sk_protocol].mc_list)
1161 ret += do_one_set_err(sk, &info);
1163 read_unlock(&nl_table_lock);
1166 EXPORT_SYMBOL(netlink_set_err);
1168 /* must be called with netlink table grabbed */
1169 static void netlink_update_socket_mc(struct netlink_sock *nlk,
1173 int old, new = !!is_new, subscriptions;
1175 old = test_bit(group - 1, nlk->groups);
1176 subscriptions = nlk->subscriptions - old + new;
1178 __set_bit(group - 1, nlk->groups);
1180 __clear_bit(group - 1, nlk->groups);
1181 netlink_update_subscriptions(&nlk->sk, subscriptions);
1182 netlink_update_listeners(&nlk->sk);
1185 static int netlink_setsockopt(struct socket *sock, int level, int optname,
1186 char __user *optval, unsigned int optlen)
1188 struct sock *sk = sock->sk;
1189 struct netlink_sock *nlk = nlk_sk(sk);
1190 unsigned int val = 0;
1193 if (level != SOL_NETLINK)
1194 return -ENOPROTOOPT;
1196 if (optlen >= sizeof(int) &&
1197 get_user(val, (unsigned int __user *)optval))
1201 case NETLINK_PKTINFO:
1203 nlk->flags |= NETLINK_RECV_PKTINFO;
1205 nlk->flags &= ~NETLINK_RECV_PKTINFO;
1208 case NETLINK_ADD_MEMBERSHIP:
1209 case NETLINK_DROP_MEMBERSHIP: {
1210 if (!netlink_capable(sock, NL_CFG_F_NONROOT_RECV))
1212 err = netlink_realloc_groups(sk);
1215 if (!val || val - 1 >= nlk->ngroups)
1217 netlink_table_grab();
1218 netlink_update_socket_mc(nlk, val,
1219 optname == NETLINK_ADD_MEMBERSHIP);
1220 netlink_table_ungrab();
1222 if (nlk->netlink_bind)
1223 nlk->netlink_bind(val);
1228 case NETLINK_BROADCAST_ERROR:
1230 nlk->flags |= NETLINK_BROADCAST_SEND_ERROR;
1232 nlk->flags &= ~NETLINK_BROADCAST_SEND_ERROR;
1235 case NETLINK_NO_ENOBUFS:
1237 nlk->flags |= NETLINK_RECV_NO_ENOBUFS;
1238 clear_bit(0, &nlk->state);
1239 wake_up_interruptible(&nlk->wait);
1241 nlk->flags &= ~NETLINK_RECV_NO_ENOBUFS;
1251 static int netlink_getsockopt(struct socket *sock, int level, int optname,
1252 char __user *optval, int __user *optlen)
1254 struct sock *sk = sock->sk;
1255 struct netlink_sock *nlk = nlk_sk(sk);
1258 if (level != SOL_NETLINK)
1259 return -ENOPROTOOPT;
1261 if (get_user(len, optlen))
1267 case NETLINK_PKTINFO:
1268 if (len < sizeof(int))
1271 val = nlk->flags & NETLINK_RECV_PKTINFO ? 1 : 0;
1272 if (put_user(len, optlen) ||
1273 put_user(val, optval))
1277 case NETLINK_BROADCAST_ERROR:
1278 if (len < sizeof(int))
1281 val = nlk->flags & NETLINK_BROADCAST_SEND_ERROR ? 1 : 0;
1282 if (put_user(len, optlen) ||
1283 put_user(val, optval))
1287 case NETLINK_NO_ENOBUFS:
1288 if (len < sizeof(int))
1291 val = nlk->flags & NETLINK_RECV_NO_ENOBUFS ? 1 : 0;
1292 if (put_user(len, optlen) ||
1293 put_user(val, optval))
1303 static void netlink_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
1305 struct nl_pktinfo info;
1307 info.group = NETLINK_CB(skb).dst_group;
1308 put_cmsg(msg, SOL_NETLINK, NETLINK_PKTINFO, sizeof(info), &info);
1311 static int netlink_sendmsg(struct kiocb *kiocb, struct socket *sock,
1312 struct msghdr *msg, size_t len)
1314 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1315 struct sock *sk = sock->sk;
1316 struct netlink_sock *nlk = nlk_sk(sk);
1317 struct sockaddr_nl *addr = msg->msg_name;
1320 struct sk_buff *skb;
1322 struct scm_cookie scm;
1324 if (msg->msg_flags&MSG_OOB)
1327 if (NULL == siocb->scm)
1330 err = scm_send(sock, msg, siocb->scm, true);
1334 if (msg->msg_namelen) {
1336 if (addr->nl_family != AF_NETLINK)
1338 dst_portid = addr->nl_pid;
1339 dst_group = ffs(addr->nl_groups);
1341 if ((dst_group || dst_portid) &&
1342 !netlink_capable(sock, NL_CFG_F_NONROOT_SEND))
1345 dst_portid = nlk->dst_portid;
1346 dst_group = nlk->dst_group;
1350 err = netlink_autobind(sock);
1356 if (len > sk->sk_sndbuf - 32)
1359 skb = alloc_skb(len, GFP_KERNEL);
1363 NETLINK_CB(skb).portid = nlk->portid;
1364 NETLINK_CB(skb).dst_group = dst_group;
1365 NETLINK_CB(skb).creds = siocb->scm->creds;
1368 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
1373 err = security_netlink_send(sk, skb);
1380 atomic_inc(&skb->users);
1381 netlink_broadcast(sk, skb, dst_portid, dst_group, GFP_KERNEL);
1383 err = netlink_unicast(sk, skb, dst_portid, msg->msg_flags&MSG_DONTWAIT);
1386 scm_destroy(siocb->scm);
1390 static int netlink_recvmsg(struct kiocb *kiocb, struct socket *sock,
1391 struct msghdr *msg, size_t len,
1394 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1395 struct scm_cookie scm;
1396 struct sock *sk = sock->sk;
1397 struct netlink_sock *nlk = nlk_sk(sk);
1398 int noblock = flags&MSG_DONTWAIT;
1400 struct sk_buff *skb, *data_skb;
1408 skb = skb_recv_datagram(sk, flags, noblock, &err);
1414 #ifdef CONFIG_COMPAT_NETLINK_MESSAGES
1415 if (unlikely(skb_shinfo(skb)->frag_list)) {
1417 * If this skb has a frag_list, then here that means that we
1418 * will have to use the frag_list skb's data for compat tasks
1419 * and the regular skb's data for normal (non-compat) tasks.
1421 * If we need to send the compat skb, assign it to the
1422 * 'data_skb' variable so that it will be used below for data
1423 * copying. We keep 'skb' for everything else, including
1424 * freeing both later.
1426 if (flags & MSG_CMSG_COMPAT)
1427 data_skb = skb_shinfo(skb)->frag_list;
1431 msg->msg_namelen = 0;
1433 copied = data_skb->len;
1435 msg->msg_flags |= MSG_TRUNC;
1439 skb_reset_transport_header(data_skb);
1440 err = skb_copy_datagram_iovec(data_skb, 0, msg->msg_iov, copied);
1442 if (msg->msg_name) {
1443 struct sockaddr_nl *addr = (struct sockaddr_nl *)msg->msg_name;
1444 addr->nl_family = AF_NETLINK;
1446 addr->nl_pid = NETLINK_CB(skb).portid;
1447 addr->nl_groups = netlink_group_mask(NETLINK_CB(skb).dst_group);
1448 msg->msg_namelen = sizeof(*addr);
1451 if (nlk->flags & NETLINK_RECV_PKTINFO)
1452 netlink_cmsg_recv_pktinfo(msg, skb);
1454 if (NULL == siocb->scm) {
1455 memset(&scm, 0, sizeof(scm));
1458 siocb->scm->creds = *NETLINK_CREDS(skb);
1459 if (flags & MSG_TRUNC)
1460 copied = data_skb->len;
1462 skb_free_datagram(sk, skb);
1464 if (nlk->cb && atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf / 2) {
1465 ret = netlink_dump(sk);
1468 sk->sk_error_report(sk);
1472 scm_recv(sock, msg, siocb->scm, flags);
1474 netlink_rcv_wake(sk);
1475 return err ? : copied;
1478 static void netlink_data_ready(struct sock *sk, int len)
1484 * We export these functions to other modules. They provide a
1485 * complete set of kernel non-blocking support for message
1490 __netlink_kernel_create(struct net *net, int unit, struct module *module,
1491 struct netlink_kernel_cfg *cfg)
1493 struct socket *sock;
1495 struct netlink_sock *nlk;
1496 struct listeners *listeners = NULL;
1497 struct mutex *cb_mutex = cfg ? cfg->cb_mutex : NULL;
1498 unsigned int groups;
1502 if (unit < 0 || unit >= MAX_LINKS)
1505 if (sock_create_lite(PF_NETLINK, SOCK_DGRAM, unit, &sock))
1509 * We have to just have a reference on the net from sk, but don't
1510 * get_net it. Besides, we cannot get and then put the net here.
1511 * So we create one inside init_net and the move it to net.
1514 if (__netlink_create(&init_net, sock, cb_mutex, unit) < 0)
1515 goto out_sock_release_nosk;
1518 sk_change_net(sk, net);
1520 if (!cfg || cfg->groups < 32)
1523 groups = cfg->groups;
1525 listeners = kzalloc(sizeof(*listeners) + NLGRPSZ(groups), GFP_KERNEL);
1527 goto out_sock_release;
1529 sk->sk_data_ready = netlink_data_ready;
1530 if (cfg && cfg->input)
1531 nlk_sk(sk)->netlink_rcv = cfg->input;
1533 if (netlink_insert(sk, net, 0))
1534 goto out_sock_release;
1537 nlk->flags |= NETLINK_KERNEL_SOCKET;
1539 netlink_table_grab();
1540 if (!nl_table[unit].registered) {
1541 nl_table[unit].groups = groups;
1542 rcu_assign_pointer(nl_table[unit].listeners, listeners);
1543 nl_table[unit].cb_mutex = cb_mutex;
1544 nl_table[unit].module = module;
1546 nl_table[unit].bind = cfg->bind;
1547 nl_table[unit].flags = cfg->flags;
1549 nl_table[unit].registered = 1;
1552 nl_table[unit].registered++;
1554 netlink_table_ungrab();
1559 netlink_kernel_release(sk);
1562 out_sock_release_nosk:
1566 EXPORT_SYMBOL(__netlink_kernel_create);
1569 netlink_kernel_release(struct sock *sk)
1571 sk_release_kernel(sk);
1573 EXPORT_SYMBOL(netlink_kernel_release);
1575 int __netlink_change_ngroups(struct sock *sk, unsigned int groups)
1577 struct listeners *new, *old;
1578 struct netlink_table *tbl = &nl_table[sk->sk_protocol];
1583 if (NLGRPSZ(tbl->groups) < NLGRPSZ(groups)) {
1584 new = kzalloc(sizeof(*new) + NLGRPSZ(groups), GFP_ATOMIC);
1587 old = nl_deref_protected(tbl->listeners);
1588 memcpy(new->masks, old->masks, NLGRPSZ(tbl->groups));
1589 rcu_assign_pointer(tbl->listeners, new);
1591 kfree_rcu(old, rcu);
1593 tbl->groups = groups;
1599 * netlink_change_ngroups - change number of multicast groups
1601 * This changes the number of multicast groups that are available
1602 * on a certain netlink family. Note that it is not possible to
1603 * change the number of groups to below 32. Also note that it does
1604 * not implicitly call netlink_clear_multicast_users() when the
1605 * number of groups is reduced.
1607 * @sk: The kernel netlink socket, as returned by netlink_kernel_create().
1608 * @groups: The new number of groups.
1610 int netlink_change_ngroups(struct sock *sk, unsigned int groups)
1614 netlink_table_grab();
1615 err = __netlink_change_ngroups(sk, groups);
1616 netlink_table_ungrab();
1621 void __netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1624 struct netlink_table *tbl = &nl_table[ksk->sk_protocol];
1626 sk_for_each_bound(sk, &tbl->mc_list)
1627 netlink_update_socket_mc(nlk_sk(sk), group, 0);
1631 * netlink_clear_multicast_users - kick off multicast listeners
1633 * This function removes all listeners from the given group.
1634 * @ksk: The kernel netlink socket, as returned by
1635 * netlink_kernel_create().
1636 * @group: The multicast group to clear.
1638 void netlink_clear_multicast_users(struct sock *ksk, unsigned int group)
1640 netlink_table_grab();
1641 __netlink_clear_multicast_users(ksk, group);
1642 netlink_table_ungrab();
1646 __nlmsg_put(struct sk_buff *skb, u32 portid, u32 seq, int type, int len, int flags)
1648 struct nlmsghdr *nlh;
1649 int size = nlmsg_msg_size(len);
1651 nlh = (struct nlmsghdr*)skb_put(skb, NLMSG_ALIGN(size));
1652 nlh->nlmsg_type = type;
1653 nlh->nlmsg_len = size;
1654 nlh->nlmsg_flags = flags;
1655 nlh->nlmsg_pid = portid;
1656 nlh->nlmsg_seq = seq;
1657 if (!__builtin_constant_p(size) || NLMSG_ALIGN(size) - size != 0)
1658 memset(nlmsg_data(nlh) + len, 0, NLMSG_ALIGN(size) - size);
1661 EXPORT_SYMBOL(__nlmsg_put);
1664 * It looks a bit ugly.
1665 * It would be better to create kernel thread.
1668 static int netlink_dump(struct sock *sk)
1670 struct netlink_sock *nlk = nlk_sk(sk);
1671 struct netlink_callback *cb;
1672 struct sk_buff *skb = NULL;
1673 struct nlmsghdr *nlh;
1674 int len, err = -ENOBUFS;
1677 mutex_lock(nlk->cb_mutex);
1685 alloc_size = max_t(int, cb->min_dump_alloc, NLMSG_GOODSIZE);
1687 skb = sock_rmalloc(sk, alloc_size, 0, GFP_KERNEL);
1691 len = cb->dump(skb, cb);
1694 mutex_unlock(nlk->cb_mutex);
1696 if (sk_filter(sk, skb))
1699 __netlink_sendskb(sk, skb);
1703 nlh = nlmsg_put_answer(skb, cb, NLMSG_DONE, sizeof(len), NLM_F_MULTI);
1707 nl_dump_check_consistent(cb, nlh);
1709 memcpy(nlmsg_data(nlh), &len, sizeof(len));
1711 if (sk_filter(sk, skb))
1714 __netlink_sendskb(sk, skb);
1719 mutex_unlock(nlk->cb_mutex);
1721 module_put(cb->module);
1722 netlink_consume_callback(cb);
1726 mutex_unlock(nlk->cb_mutex);
1731 int __netlink_dump_start(struct sock *ssk, struct sk_buff *skb,
1732 const struct nlmsghdr *nlh,
1733 struct netlink_dump_control *control)
1735 struct netlink_callback *cb;
1737 struct netlink_sock *nlk;
1740 cb = kzalloc(sizeof(*cb), GFP_KERNEL);
1744 cb->dump = control->dump;
1745 cb->done = control->done;
1747 cb->data = control->data;
1748 cb->module = control->module;
1749 cb->min_dump_alloc = control->min_dump_alloc;
1750 atomic_inc(&skb->users);
1753 sk = netlink_lookup(sock_net(ssk), ssk->sk_protocol, NETLINK_CB(skb).portid);
1755 netlink_destroy_callback(cb);
1756 return -ECONNREFUSED;
1760 mutex_lock(nlk->cb_mutex);
1761 /* A dump is in progress... */
1763 mutex_unlock(nlk->cb_mutex);
1764 netlink_destroy_callback(cb);
1768 /* add reference of module which cb->dump belongs to */
1769 if (!try_module_get(cb->module)) {
1770 mutex_unlock(nlk->cb_mutex);
1771 netlink_destroy_callback(cb);
1772 ret = -EPROTONOSUPPORT;
1777 mutex_unlock(nlk->cb_mutex);
1779 ret = netlink_dump(sk);
1786 /* We successfully started a dump, by returning -EINTR we
1787 * signal not to send ACK even if it was requested.
1791 EXPORT_SYMBOL(__netlink_dump_start);
1793 void netlink_ack(struct sk_buff *in_skb, struct nlmsghdr *nlh, int err)
1795 struct sk_buff *skb;
1796 struct nlmsghdr *rep;
1797 struct nlmsgerr *errmsg;
1798 size_t payload = sizeof(*errmsg);
1800 /* error messages get the original request appened */
1802 payload += nlmsg_len(nlh);
1804 skb = nlmsg_new(payload, GFP_KERNEL);
1808 sk = netlink_lookup(sock_net(in_skb->sk),
1809 in_skb->sk->sk_protocol,
1810 NETLINK_CB(in_skb).portid);
1812 sk->sk_err = ENOBUFS;
1813 sk->sk_error_report(sk);
1819 rep = __nlmsg_put(skb, NETLINK_CB(in_skb).portid, nlh->nlmsg_seq,
1820 NLMSG_ERROR, payload, 0);
1821 errmsg = nlmsg_data(rep);
1822 errmsg->error = err;
1823 memcpy(&errmsg->msg, nlh, err ? nlh->nlmsg_len : sizeof(*nlh));
1824 netlink_unicast(in_skb->sk, skb, NETLINK_CB(in_skb).portid, MSG_DONTWAIT);
1826 EXPORT_SYMBOL(netlink_ack);
1828 int netlink_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *,
1831 struct nlmsghdr *nlh;
1834 while (skb->len >= nlmsg_total_size(0)) {
1837 nlh = nlmsg_hdr(skb);
1840 if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len)
1843 /* Only requests are handled by the kernel */
1844 if (!(nlh->nlmsg_flags & NLM_F_REQUEST))
1847 /* Skip control messages */
1848 if (nlh->nlmsg_type < NLMSG_MIN_TYPE)
1856 if (nlh->nlmsg_flags & NLM_F_ACK || err)
1857 netlink_ack(skb, nlh, err);
1860 msglen = NLMSG_ALIGN(nlh->nlmsg_len);
1861 if (msglen > skb->len)
1863 skb_pull(skb, msglen);
1868 EXPORT_SYMBOL(netlink_rcv_skb);
1871 * nlmsg_notify - send a notification netlink message
1872 * @sk: netlink socket to use
1873 * @skb: notification message
1874 * @portid: destination netlink portid for reports or 0
1875 * @group: destination multicast group or 0
1876 * @report: 1 to report back, 0 to disable
1877 * @flags: allocation flags
1879 int nlmsg_notify(struct sock *sk, struct sk_buff *skb, u32 portid,
1880 unsigned int group, int report, gfp_t flags)
1885 int exclude_portid = 0;
1888 atomic_inc(&skb->users);
1889 exclude_portid = portid;
1892 /* errors reported via destination sk->sk_err, but propagate
1893 * delivery errors if NETLINK_BROADCAST_ERROR flag is set */
1894 err = nlmsg_multicast(sk, skb, exclude_portid, group, flags);
1900 err2 = nlmsg_unicast(sk, skb, portid);
1901 if (!err || err == -ESRCH)
1907 EXPORT_SYMBOL(nlmsg_notify);
1909 #ifdef CONFIG_PROC_FS
1910 struct nl_seq_iter {
1911 struct seq_net_private p;
1916 static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
1918 struct nl_seq_iter *iter = seq->private;
1923 for (i = 0; i < MAX_LINKS; i++) {
1924 struct nl_portid_hash *hash = &nl_table[i].hash;
1926 for (j = 0; j <= hash->mask; j++) {
1927 sk_for_each(s, &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_portid_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 %-6u %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_create("netlink", 0, net->proc_net, &netlink_seq_fops))
2097 static void __net_exit netlink_net_exit(struct net *net)
2099 #ifdef CONFIG_PROC_FS
2100 remove_proc_entry("netlink", net->proc_net);
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;
2119 nl_table[NETLINK_USERSOCK].flags = NL_CFG_F_NONROOT_SEND;
2121 netlink_table_ungrab();
2124 static struct pernet_operations __net_initdata netlink_net_ops = {
2125 .init = netlink_net_init,
2126 .exit = netlink_net_exit,
2129 static int __init netlink_proto_init(void)
2132 unsigned long limit;
2134 int err = proto_register(&netlink_proto, 0);
2139 BUILD_BUG_ON(sizeof(struct netlink_skb_parms) > FIELD_SIZEOF(struct sk_buff, 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_portid_hash *hash = &nl_table[i].hash;
2157 hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table));
2160 nl_portid_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);