2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro, <bir7@leland.Stanford.Edu>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
61 #include <linux/config.h>
63 #include <linux/smp_lock.h>
64 #include <linux/socket.h>
65 #include <linux/file.h>
66 #include <linux/net.h>
67 #include <linux/interrupt.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/init.h>
74 #include <linux/poll.h>
75 #include <linux/cache.h>
76 #include <linux/module.h>
77 #include <linux/highmem.h>
78 #include <linux/divert.h>
79 #include <linux/mount.h>
80 #include <linux/security.h>
81 #include <linux/syscalls.h>
82 #include <linux/compat.h>
83 #include <linux/kmod.h>
85 #ifdef CONFIG_NET_RADIO
86 #include <linux/wireless.h> /* Note : will define WIRELESS_EXT */
87 #endif /* CONFIG_NET_RADIO */
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
99 size_t size, loff_t pos);
100 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
101 size_t size, loff_t pos);
102 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
104 static int sock_close(struct inode *inode, struct file *file);
105 static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107 static int sock_ioctl(struct inode *inode, struct file *file,
108 unsigned int cmd, unsigned long arg);
109 static int sock_fasync(int fd, struct file *filp, int on);
110 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
111 unsigned long count, loff_t *ppos);
112 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
113 unsigned long count, loff_t *ppos);
114 static ssize_t sock_sendpage(struct file *file, struct page *page,
115 int offset, size_t size, loff_t *ppos, int more);
119 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
120 * in the operation structures but are done directly via the socketcall() multiplexor.
123 static struct file_operations socket_file_ops = {
124 .owner = THIS_MODULE,
126 .aio_read = sock_aio_read,
127 .aio_write = sock_aio_write,
131 .open = sock_no_open, /* special open code to disallow open via /proc */
132 .release = sock_close,
133 .fasync = sock_fasync,
135 .writev = sock_writev,
136 .sendpage = sock_sendpage
140 * The protocol list. Each protocol is registered in here.
143 static struct net_proto_family *net_families[NPROTO];
145 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
146 static atomic_t net_family_lockct = ATOMIC_INIT(0);
147 static spinlock_t net_family_lock = SPIN_LOCK_UNLOCKED;
149 /* The strategy is: modifications net_family vector are short, do not
150 sleep and veeery rare, but read access should be free of any exclusive
154 static void net_family_write_lock(void)
156 spin_lock(&net_family_lock);
157 while (atomic_read(&net_family_lockct) != 0) {
158 spin_unlock(&net_family_lock);
162 spin_lock(&net_family_lock);
166 static __inline__ void net_family_write_unlock(void)
168 spin_unlock(&net_family_lock);
171 static __inline__ void net_family_read_lock(void)
173 atomic_inc(&net_family_lockct);
174 spin_unlock_wait(&net_family_lock);
177 static __inline__ void net_family_read_unlock(void)
179 atomic_dec(&net_family_lockct);
183 #define net_family_write_lock() do { } while(0)
184 #define net_family_write_unlock() do { } while(0)
185 #define net_family_read_lock() do { } while(0)
186 #define net_family_read_unlock() do { } while(0)
191 * Statistics counters of the socket lists
194 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
197 * Support routines. Move socket addresses back and forth across the kernel/user
198 * divide and look after the messy bits.
201 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
202 16 for IP, 16 for IPX,
205 must be at least one bigger than
206 the AF_UNIX size (see net/unix/af_unix.c
211 * move_addr_to_kernel - copy a socket address into kernel space
212 * @uaddr: Address in user space
213 * @kaddr: Address in kernel space
214 * @ulen: Length in user space
216 * The address is copied into kernel space. If the provided address is
217 * too long an error code of -EINVAL is returned. If the copy gives
218 * invalid addresses -EFAULT is returned. On a success 0 is returned.
221 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
223 if(ulen<0||ulen>MAX_SOCK_ADDR)
227 if(copy_from_user(kaddr,uaddr,ulen))
233 * move_addr_to_user - copy an address to user space
234 * @kaddr: kernel space address
235 * @klen: length of address in kernel
236 * @uaddr: user space address
237 * @ulen: pointer to user length field
239 * The value pointed to by ulen on entry is the buffer length available.
240 * This is overwritten with the buffer space used. -EINVAL is returned
241 * if an overlong buffer is specified or a negative buffer size. -EFAULT
242 * is returned if either the buffer or the length field are not
244 * After copying the data up to the limit the user specifies, the true
245 * length of the data is written over the length limit the user
246 * specified. Zero is returned for a success.
249 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
254 if((err=get_user(len, ulen)))
258 if(len<0 || len> MAX_SOCK_ADDR)
262 if(copy_to_user(uaddr,kaddr,len))
266 * "fromlen shall refer to the value before truncation.."
269 return __put_user(klen, ulen);
272 #define SOCKFS_MAGIC 0x534F434B
274 static kmem_cache_t * sock_inode_cachep;
276 static struct inode *sock_alloc_inode(struct super_block *sb)
278 struct socket_alloc *ei;
279 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
282 init_waitqueue_head(&ei->socket.wait);
284 ei->socket.fasync_list = NULL;
285 ei->socket.state = SS_UNCONNECTED;
286 ei->socket.flags = 0;
287 ei->socket.ops = NULL;
288 ei->socket.sk = NULL;
289 ei->socket.file = NULL;
290 ei->socket.passcred = 0;
292 return &ei->vfs_inode;
295 static void sock_destroy_inode(struct inode *inode)
297 kmem_cache_free(sock_inode_cachep,
298 container_of(inode, struct socket_alloc, vfs_inode));
301 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
303 struct socket_alloc *ei = (struct socket_alloc *) foo;
305 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
306 SLAB_CTOR_CONSTRUCTOR)
307 inode_init_once(&ei->vfs_inode);
310 static int init_inodecache(void)
312 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
313 sizeof(struct socket_alloc),
314 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
316 if (sock_inode_cachep == NULL)
321 static struct super_operations sockfs_ops = {
322 .alloc_inode = sock_alloc_inode,
323 .destroy_inode =sock_destroy_inode,
324 .statfs = simple_statfs,
327 static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
328 int flags, const char *dev_name, void *data)
330 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
333 static struct vfsmount *sock_mnt;
335 static struct file_system_type sock_fs_type = {
337 .get_sb = sockfs_get_sb,
338 .kill_sb = kill_anon_super,
340 static int sockfs_delete_dentry(struct dentry *dentry)
344 static struct dentry_operations sockfs_dentry_operations = {
345 .d_delete = sockfs_delete_dentry,
349 * Obtains the first available file descriptor and sets it up for use.
351 * This function creates file structure and maps it to fd space
352 * of current process. On success it returns file descriptor
353 * and file struct implicitly stored in sock->file.
354 * Note that another thread may close file descriptor before we return
355 * from this function. We use the fact that now we do not refer
356 * to socket after mapping. If one day we will need it, this
357 * function will increment ref. count on file by 1.
359 * In any case returned fd MAY BE not valid!
360 * This race condition is unavoidable
361 * with shared fd spaces, we cannot solve it inside kernel,
362 * but we take care of internal coherence yet.
365 int sock_map_fd(struct socket *sock)
372 * Find a file descriptor suitable for return to the user.
375 fd = get_unused_fd();
377 struct file *file = get_empty_filp();
385 sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
387 this.len = strlen(name);
388 this.hash = SOCK_INODE(sock)->i_ino;
390 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
391 if (!file->f_dentry) {
397 file->f_dentry->d_op = &sockfs_dentry_operations;
398 d_add(file->f_dentry, SOCK_INODE(sock));
399 file->f_vfsmnt = mntget(sock_mnt);
400 file->f_mapping = file->f_dentry->d_inode->i_mapping;
403 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
405 file->f_flags = O_RDWR;
407 fd_install(fd, file);
415 * sockfd_lookup - Go from a file number to its socket slot
417 * @err: pointer to an error code return
419 * The file handle passed in is locked and the socket it is bound
420 * too is returned. If an error occurs the err pointer is overwritten
421 * with a negative errno code and NULL is returned. The function checks
422 * for both invalid handles and passing a handle which is not a socket.
424 * On a success the socket object pointer is returned.
427 struct socket *sockfd_lookup(int fd, int *err)
433 if (!(file = fget(fd)))
439 inode = file->f_dentry->d_inode;
440 if (!inode->i_sock || !(sock = SOCKET_I(inode)))
447 if (sock->file != file) {
448 printk(KERN_ERR "socki_lookup: socket file changed!\n");
455 * sock_alloc - allocate a socket
457 * Allocate a new inode and socket object. The two are bound together
458 * and initialised. The socket is then returned. If we are out of inodes
462 struct socket *sock_alloc(void)
464 struct inode * inode;
465 struct socket * sock;
467 inode = new_inode(sock_mnt->mnt_sb);
471 sock = SOCKET_I(inode);
473 inode->i_mode = S_IFSOCK|S_IRWXUGO;
475 inode->i_uid = current->fsuid;
476 inode->i_gid = current->fsgid;
478 get_cpu_var(sockets_in_use)++;
479 put_cpu_var(sockets_in_use);
484 * In theory you can't get an open on this inode, but /proc provides
485 * a back door. Remember to keep it shut otherwise you'll let the
486 * creepy crawlies in.
489 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
494 struct file_operations bad_sock_fops = {
495 .owner = THIS_MODULE,
496 .open = sock_no_open,
500 * sock_release - close a socket
501 * @sock: socket to close
503 * The socket is released from the protocol stack if it has a release
504 * callback, and the inode is then released if the socket is bound to
505 * an inode not a file.
508 void sock_release(struct socket *sock)
511 struct module *owner = sock->ops->owner;
513 sock->ops->release(sock);
518 if (sock->fasync_list)
519 printk(KERN_ERR "sock_release: fasync list not empty!\n");
521 get_cpu_var(sockets_in_use)--;
522 put_cpu_var(sockets_in_use);
524 iput(SOCK_INODE(sock));
530 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
531 struct msghdr *msg, size_t size)
533 struct sock_iocb *si = kiocb_to_siocb(iocb);
541 err = security_socket_sendmsg(sock, msg, size);
545 return sock->ops->sendmsg(iocb, sock, msg, size);
548 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
553 init_sync_kiocb(&iocb, NULL);
554 ret = __sock_sendmsg(&iocb, sock, msg, size);
555 if (-EIOCBQUEUED == ret)
556 ret = wait_on_sync_kiocb(&iocb);
561 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
562 struct msghdr *msg, size_t size, int flags)
565 struct sock_iocb *si = kiocb_to_siocb(iocb);
573 err = security_socket_recvmsg(sock, msg, size, flags);
577 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
580 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
581 size_t size, int flags)
586 init_sync_kiocb(&iocb, NULL);
587 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
588 if (-EIOCBQUEUED == ret)
589 ret = wait_on_sync_kiocb(&iocb);
594 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
595 * area ubuf...ubuf+size-1 is writable before asking the protocol.
598 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
599 size_t size, loff_t pos)
601 struct sock_iocb *x = kiocb_to_siocb(iocb);
607 if (size==0) /* Match SYS5 behaviour */
610 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
612 x->async_msg.msg_name = NULL;
613 x->async_msg.msg_namelen = 0;
614 x->async_msg.msg_iov = &x->async_iov;
615 x->async_msg.msg_iovlen = 1;
616 x->async_msg.msg_control = NULL;
617 x->async_msg.msg_controllen = 0;
618 x->async_iov.iov_base = ubuf;
619 x->async_iov.iov_len = size;
620 flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
622 return __sock_recvmsg(iocb, sock, &x->async_msg, size, flags);
627 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
628 * is readable by the user process.
631 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
632 size_t size, loff_t pos)
634 struct sock_iocb *x = kiocb_to_siocb(iocb);
639 if(size==0) /* Match SYS5 behaviour */
642 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
644 x->async_msg.msg_name = NULL;
645 x->async_msg.msg_namelen = 0;
646 x->async_msg.msg_iov = &x->async_iov;
647 x->async_msg.msg_iovlen = 1;
648 x->async_msg.msg_control = NULL;
649 x->async_msg.msg_controllen = 0;
650 x->async_msg.msg_flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
651 if (sock->type == SOCK_SEQPACKET)
652 x->async_msg.msg_flags |= MSG_EOR;
653 x->async_iov.iov_base = (void __user *)ubuf;
654 x->async_iov.iov_len = size;
656 return __sock_sendmsg(iocb, sock, &x->async_msg, size);
659 ssize_t sock_sendpage(struct file *file, struct page *page,
660 int offset, size_t size, loff_t *ppos, int more)
665 if (ppos != &file->f_pos)
668 sock = SOCKET_I(file->f_dentry->d_inode);
670 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
674 return sock->ops->sendpage(sock, page, offset, size, flags);
677 int sock_readv_writev(int type, struct inode * inode, struct file * file,
678 const struct iovec * iov, long count, size_t size)
683 sock = SOCKET_I(inode);
687 msg.msg_control = NULL;
688 msg.msg_controllen = 0;
689 msg.msg_iov = (struct iovec *) iov;
690 msg.msg_iovlen = count;
691 msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
693 /* read() does a VERIFY_WRITE */
694 if (type == VERIFY_WRITE)
695 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
697 if (sock->type == SOCK_SEQPACKET)
698 msg.msg_flags |= MSG_EOR;
700 return sock_sendmsg(sock, &msg, size);
703 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
704 unsigned long count, loff_t *ppos)
708 for (i = 0 ; i < count ; i++)
709 tot_len += vector[i].iov_len;
710 return sock_readv_writev(VERIFY_WRITE, file->f_dentry->d_inode,
711 file, vector, count, tot_len);
714 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
715 unsigned long count, loff_t *ppos)
719 for (i = 0 ; i < count ; i++)
720 tot_len += vector[i].iov_len;
721 return sock_readv_writev(VERIFY_READ, file->f_dentry->d_inode,
722 file, vector, count, tot_len);
727 * Atomic setting of ioctl hooks to avoid race
728 * with module unload.
731 static DECLARE_MUTEX(br_ioctl_mutex);
732 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
734 void brioctl_set(int (*hook)(unsigned int, void __user *))
736 down(&br_ioctl_mutex);
737 br_ioctl_hook = hook;
740 EXPORT_SYMBOL(brioctl_set);
742 static DECLARE_MUTEX(vlan_ioctl_mutex);
743 static int (*vlan_ioctl_hook)(void __user *arg);
745 void vlan_ioctl_set(int (*hook)(void __user *))
747 down(&vlan_ioctl_mutex);
748 vlan_ioctl_hook = hook;
749 up(&vlan_ioctl_mutex);
751 EXPORT_SYMBOL(vlan_ioctl_set);
753 static DECLARE_MUTEX(dlci_ioctl_mutex);
754 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
756 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
758 down(&dlci_ioctl_mutex);
759 dlci_ioctl_hook = hook;
760 up(&dlci_ioctl_mutex);
762 EXPORT_SYMBOL(dlci_ioctl_set);
765 * With an ioctl, arg may well be a user mode pointer, but we don't know
766 * what to do with it - that's up to the protocol still.
769 static int sock_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
773 void __user *argp = (void __user *)arg;
777 sock = SOCKET_I(inode);
778 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
779 err = dev_ioctl(cmd, argp);
782 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
783 err = dev_ioctl(cmd, argp);
785 #endif /* WIRELESS_EXT */
790 if (get_user(pid, (int __user *)argp))
792 err = f_setown(sock->file, pid, 1);
796 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
804 request_module("bridge");
806 down(&br_ioctl_mutex);
808 err = br_ioctl_hook(cmd, argp);
814 if (!vlan_ioctl_hook)
815 request_module("8021q");
817 down(&vlan_ioctl_mutex);
819 err = vlan_ioctl_hook(argp);
820 up(&vlan_ioctl_mutex);
824 /* Convert this to call through a hook */
825 err = divert_ioctl(cmd, argp);
830 if (!dlci_ioctl_hook)
831 request_module("dlci");
833 if (dlci_ioctl_hook) {
834 down(&dlci_ioctl_mutex);
835 err = dlci_ioctl_hook(cmd, argp);
836 up(&dlci_ioctl_mutex);
840 err = sock->ops->ioctl(sock, cmd, arg);
848 int sock_create_lite(int family, int type, int protocol, struct socket **res)
851 struct socket *sock = NULL;
853 err = security_socket_create(family, type, protocol, 1);
863 security_socket_post_create(sock, family, type, protocol, 1);
870 /* No kernel lock held - perfect */
871 static unsigned int sock_poll(struct file *file, poll_table * wait)
876 * We can't return errors to poll, so it's either yes or no.
878 sock = SOCKET_I(file->f_dentry->d_inode);
879 return sock->ops->poll(file, sock, wait);
882 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
884 struct socket *sock = SOCKET_I(file->f_dentry->d_inode);
886 return sock->ops->mmap(file, sock, vma);
889 int sock_close(struct inode *inode, struct file *filp)
892 * It was possible the inode is NULL we were
893 * closing an unfinished socket.
898 printk(KERN_DEBUG "sock_close: NULL inode\n");
901 sock_fasync(-1, filp, 0);
902 sock_release(SOCKET_I(inode));
907 * Update the socket async list
909 * Fasync_list locking strategy.
911 * 1. fasync_list is modified only under process context socket lock
912 * i.e. under semaphore.
913 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
914 * or under socket lock.
915 * 3. fasync_list can be used from softirq context, so that
916 * modification under socket lock have to be enhanced with
917 * write_lock_bh(&sk->sk_callback_lock).
921 static int sock_fasync(int fd, struct file *filp, int on)
923 struct fasync_struct *fa, *fna=NULL, **prev;
929 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
934 sock = SOCKET_I(filp->f_dentry->d_inode);
936 if ((sk=sock->sk) == NULL) {
944 prev=&(sock->fasync_list);
946 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
947 if (fa->fa_file==filp)
954 write_lock_bh(&sk->sk_callback_lock);
956 write_unlock_bh(&sk->sk_callback_lock);
963 fna->magic=FASYNC_MAGIC;
964 fna->fa_next=sock->fasync_list;
965 write_lock_bh(&sk->sk_callback_lock);
966 sock->fasync_list=fna;
967 write_unlock_bh(&sk->sk_callback_lock);
973 write_lock_bh(&sk->sk_callback_lock);
975 write_unlock_bh(&sk->sk_callback_lock);
981 release_sock(sock->sk);
985 /* This function may be called only under socket lock or callback_lock */
987 int sock_wake_async(struct socket *sock, int how, int band)
989 if (!sock || !sock->fasync_list)
995 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
999 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1004 __kill_fasync(sock->fasync_list, SIGIO, band);
1007 __kill_fasync(sock->fasync_list, SIGURG, band);
1012 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1016 struct socket *sock;
1019 * Check protocol is in range
1021 if (family < 0 || family >= NPROTO)
1022 return -EAFNOSUPPORT;
1023 if (type < 0 || type >= SOCK_MAX)
1028 This uglymoron is moved from INET layer to here to avoid
1029 deadlock in module load.
1031 if (family == PF_INET && type == SOCK_PACKET) {
1035 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1040 err = security_socket_create(family, type, protocol, kern);
1044 #if defined(CONFIG_KMOD)
1045 /* Attempt to load a protocol module if the find failed.
1047 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1048 * requested real, full-featured networking support upon configuration.
1049 * Otherwise module support will break!
1051 if (net_families[family]==NULL)
1053 request_module("net-pf-%d",family);
1057 net_family_read_lock();
1058 if (net_families[family] == NULL) {
1064 * Allocate the socket and allow the family to set things up. if
1065 * the protocol is 0, the family is instructed to select an appropriate
1069 if (!(sock = sock_alloc()))
1071 printk(KERN_WARNING "socket: no more sockets\n");
1072 i = -ENFILE; /* Not exactly a match, but its the
1073 closest posix thing */
1080 * We will call the ->create function, that possibly is in a loadable
1081 * module, so we have to bump that loadable module refcnt first.
1084 if (!try_module_get(net_families[family]->owner))
1087 if ((i = net_families[family]->create(sock, protocol)) < 0)
1088 goto out_module_put;
1090 * Now to bump the refcnt of the [loadable] module that owns this
1091 * socket at sock_release time we decrement its refcnt.
1093 if (!try_module_get(sock->ops->owner)) {
1095 goto out_module_put;
1098 * Now that we're done with the ->create function, the [loadable]
1099 * module can have its refcnt decremented
1101 module_put(net_families[family]->owner);
1103 security_socket_post_create(sock, family, type, protocol, kern);
1106 net_family_read_unlock();
1109 module_put(net_families[family]->owner);
1115 int sock_create(int family, int type, int protocol, struct socket **res)
1117 return __sock_create(family, type, protocol, res, 0);
1120 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1122 return __sock_create(family, type, protocol, res, 1);
1125 asmlinkage long sys_socket(int family, int type, int protocol)
1128 struct socket *sock;
1130 retval = sock_create(family, type, protocol, &sock);
1134 retval = sock_map_fd(sock);
1139 /* It may be already another descriptor 8) Not kernel problem. */
1148 * Create a pair of connected sockets.
1151 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1153 struct socket *sock1, *sock2;
1157 * Obtain the first socket and check if the underlying protocol
1158 * supports the socketpair call.
1161 err = sock_create(family, type, protocol, &sock1);
1165 err = sock_create(family, type, protocol, &sock2);
1169 err = sock1->ops->socketpair(sock1, sock2);
1171 goto out_release_both;
1175 err = sock_map_fd(sock1);
1177 goto out_release_both;
1180 err = sock_map_fd(sock2);
1185 /* fd1 and fd2 may be already another descriptors.
1186 * Not kernel problem.
1189 err = put_user(fd1, &usockvec[0]);
1191 err = put_user(fd2, &usockvec[1]);
1200 sock_release(sock2);
1205 sock_release(sock2);
1207 sock_release(sock1);
1214 * Bind a name to a socket. Nothing much to do here since it's
1215 * the protocol's responsibility to handle the local address.
1217 * We move the socket address to kernel space before we call
1218 * the protocol layer (having also checked the address is ok).
1221 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1223 struct socket *sock;
1224 char address[MAX_SOCK_ADDR];
1227 if((sock = sockfd_lookup(fd,&err))!=NULL)
1229 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1230 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1235 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1244 * Perform a listen. Basically, we allow the protocol to do anything
1245 * necessary for a listen, and if that works, we mark the socket as
1246 * ready for listening.
1249 int sysctl_somaxconn = SOMAXCONN;
1251 asmlinkage long sys_listen(int fd, int backlog)
1253 struct socket *sock;
1256 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1257 if ((unsigned) backlog > sysctl_somaxconn)
1258 backlog = sysctl_somaxconn;
1260 err = security_socket_listen(sock, backlog);
1266 err=sock->ops->listen(sock, backlog);
1274 * For accept, we attempt to create a new socket, set up the link
1275 * with the client, wake up the client, then return the new
1276 * connected fd. We collect the address of the connector in kernel
1277 * space and move it to user at the very end. This is unclean because
1278 * we open the socket then return an error.
1280 * 1003.1g adds the ability to recvmsg() to query connection pending
1281 * status to recvmsg. We need to add that support in a way thats
1282 * clean when we restucture accept also.
1285 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1287 struct socket *sock, *newsock;
1289 char address[MAX_SOCK_ADDR];
1291 sock = sockfd_lookup(fd, &err);
1296 if (!(newsock = sock_alloc()))
1299 newsock->type = sock->type;
1300 newsock->ops = sock->ops;
1302 err = security_socket_accept(sock, newsock);
1307 * We don't need try_module_get here, as the listening socket (sock)
1308 * has the protocol module (sock->ops->owner) held.
1310 __module_get(newsock->ops->owner);
1312 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1316 if (upeer_sockaddr) {
1317 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1318 err = -ECONNABORTED;
1321 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1326 /* File flags are not inherited via accept() unlike another OSes. */
1328 if ((err = sock_map_fd(newsock)) < 0)
1331 security_socket_post_accept(sock, newsock);
1338 sock_release(newsock);
1344 * Attempt to connect to a socket with the server address. The address
1345 * is in user space so we verify it is OK and move it to kernel space.
1347 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1350 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1351 * other SEQPACKET protocols that take time to connect() as it doesn't
1352 * include the -EINPROGRESS status for such sockets.
1355 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1357 struct socket *sock;
1358 char address[MAX_SOCK_ADDR];
1361 sock = sockfd_lookup(fd, &err);
1364 err = move_addr_to_kernel(uservaddr, addrlen, address);
1368 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1372 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1373 sock->file->f_flags);
1381 * Get the local address ('name') of a socket object. Move the obtained
1382 * name to user space.
1385 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1387 struct socket *sock;
1388 char address[MAX_SOCK_ADDR];
1391 sock = sockfd_lookup(fd, &err);
1395 err = security_socket_getsockname(sock);
1399 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1402 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1411 * Get the remote address ('name') of a socket object. Move the obtained
1412 * name to user space.
1415 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1417 struct socket *sock;
1418 char address[MAX_SOCK_ADDR];
1421 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1423 err = security_socket_getpeername(sock);
1429 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1431 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1438 * Send a datagram to a given address. We move the address into kernel
1439 * space and check the user space data area is readable before invoking
1443 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1444 struct sockaddr __user *addr, int addr_len)
1446 struct socket *sock;
1447 char address[MAX_SOCK_ADDR];
1452 sock = sockfd_lookup(fd, &err);
1460 msg.msg_control=NULL;
1461 msg.msg_controllen=0;
1465 err = move_addr_to_kernel(addr, addr_len, address);
1468 msg.msg_name=address;
1469 msg.msg_namelen=addr_len;
1471 if (sock->file->f_flags & O_NONBLOCK)
1472 flags |= MSG_DONTWAIT;
1473 msg.msg_flags = flags;
1474 err = sock_sendmsg(sock, &msg, len);
1483 * Send a datagram down a socket.
1486 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1488 return sys_sendto(fd, buff, len, flags, NULL, 0);
1492 * Receive a frame from the socket and optionally record the address of the
1493 * sender. We verify the buffers are writable and if needed move the
1494 * sender address from kernel to user space.
1497 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1498 struct sockaddr __user *addr, int __user *addr_len)
1500 struct socket *sock;
1503 char address[MAX_SOCK_ADDR];
1506 sock = sockfd_lookup(fd, &err);
1510 msg.msg_control=NULL;
1511 msg.msg_controllen=0;
1516 msg.msg_name=address;
1517 msg.msg_namelen=MAX_SOCK_ADDR;
1518 if (sock->file->f_flags & O_NONBLOCK)
1519 flags |= MSG_DONTWAIT;
1520 err=sock_recvmsg(sock, &msg, size, flags);
1522 if(err >= 0 && addr != NULL)
1524 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1534 * Receive a datagram from a socket.
1537 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1539 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1543 * Set a socket option. Because we don't know the option lengths we have
1544 * to pass the user mode parameter for the protocols to sort out.
1547 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1550 struct socket *sock;
1555 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1557 err = security_socket_setsockopt(sock,level,optname);
1563 if (level == SOL_SOCKET)
1564 err=sock_setsockopt(sock,level,optname,optval,optlen);
1566 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1573 * Get a socket option. Because we don't know the option lengths we have
1574 * to pass a user mode parameter for the protocols to sort out.
1577 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1580 struct socket *sock;
1582 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1584 err = security_socket_getsockopt(sock, level,
1591 if (level == SOL_SOCKET)
1592 err=sock_getsockopt(sock,level,optname,optval,optlen);
1594 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1602 * Shutdown a socket.
1605 asmlinkage long sys_shutdown(int fd, int how)
1608 struct socket *sock;
1610 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1612 err = security_socket_shutdown(sock, how);
1618 err=sock->ops->shutdown(sock, how);
1624 /* A couple of helpful macros for getting the address of the 32/64 bit
1625 * fields which are the same type (int / unsigned) on our platforms.
1627 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1628 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1629 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1633 * BSD sendmsg interface
1636 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1638 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1639 struct socket *sock;
1640 char address[MAX_SOCK_ADDR];
1641 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1642 unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1643 unsigned char *ctl_buf = ctl;
1644 struct msghdr msg_sys;
1645 int err, ctl_len, iov_size, total_len;
1648 if (MSG_CMSG_COMPAT & flags) {
1649 if (get_compat_msghdr(&msg_sys, msg_compat))
1651 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1654 sock = sockfd_lookup(fd, &err);
1658 /* do not move before msg_sys is valid */
1660 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1663 /* Check whether to allocate the iovec area*/
1665 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1666 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1667 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1672 /* This will also move the address data into kernel space */
1673 if (MSG_CMSG_COMPAT & flags) {
1674 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1676 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1683 if (msg_sys.msg_controllen > INT_MAX)
1685 ctl_len = msg_sys.msg_controllen;
1686 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1687 err = cmsghdr_from_user_compat_to_kern(&msg_sys, ctl, sizeof(ctl));
1690 ctl_buf = msg_sys.msg_control;
1691 } else if (ctl_len) {
1692 if (ctl_len > sizeof(ctl))
1694 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1695 if (ctl_buf == NULL)
1700 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1701 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1702 * checking falls down on this.
1704 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1706 msg_sys.msg_control = ctl_buf;
1708 msg_sys.msg_flags = flags;
1710 if (sock->file->f_flags & O_NONBLOCK)
1711 msg_sys.msg_flags |= MSG_DONTWAIT;
1712 err = sock_sendmsg(sock, &msg_sys, total_len);
1716 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1718 if (iov != iovstack)
1719 sock_kfree_s(sock->sk, iov, iov_size);
1727 * BSD recvmsg interface
1730 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1732 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1733 struct socket *sock;
1734 struct iovec iovstack[UIO_FASTIOV];
1735 struct iovec *iov=iovstack;
1736 struct msghdr msg_sys;
1737 unsigned long cmsg_ptr;
1738 int err, iov_size, total_len, len;
1740 /* kernel mode address */
1741 char addr[MAX_SOCK_ADDR];
1743 /* user mode address pointers */
1744 struct sockaddr __user *uaddr;
1745 int __user *uaddr_len;
1747 if (MSG_CMSG_COMPAT & flags) {
1748 if (get_compat_msghdr(&msg_sys, msg_compat))
1751 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1754 sock = sockfd_lookup(fd, &err);
1759 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1762 /* Check whether to allocate the iovec area*/
1764 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1765 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1766 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1772 * Save the user-mode address (verify_iovec will change the
1773 * kernel msghdr to use the kernel address space)
1776 uaddr = (void __user *) msg_sys.msg_name;
1777 uaddr_len = COMPAT_NAMELEN(msg);
1778 if (MSG_CMSG_COMPAT & flags) {
1779 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1781 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1786 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1787 msg_sys.msg_flags = 0;
1788 if (MSG_CMSG_COMPAT & flags)
1789 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1791 if (sock->file->f_flags & O_NONBLOCK)
1792 flags |= MSG_DONTWAIT;
1793 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1798 if (uaddr != NULL) {
1799 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1803 err = __put_user(msg_sys.msg_flags, COMPAT_FLAGS(msg));
1806 if (MSG_CMSG_COMPAT & flags)
1807 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1808 &msg_compat->msg_controllen);
1810 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1811 &msg->msg_controllen);
1817 if (iov != iovstack)
1818 sock_kfree_s(sock->sk, iov, iov_size);
1825 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1827 /* Argument list sizes for sys_socketcall */
1828 #define AL(x) ((x) * sizeof(unsigned long))
1829 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1830 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1831 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1835 * System call vectors.
1837 * Argument checking cleaned up. Saved 20% in size.
1838 * This function doesn't need to set the kernel lock because
1839 * it is set by the callees.
1842 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1845 unsigned long a0,a1;
1848 if(call<1||call>SYS_RECVMSG)
1851 /* copy_from_user should be SMP safe. */
1852 if (copy_from_user(a, args, nargs[call]))
1861 err = sys_socket(a0,a1,a[2]);
1864 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1867 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1870 err = sys_listen(a0,a1);
1873 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1875 case SYS_GETSOCKNAME:
1876 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1878 case SYS_GETPEERNAME:
1879 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1881 case SYS_SOCKETPAIR:
1882 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1885 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1888 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1889 (struct sockaddr __user *)a[4], a[5]);
1892 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1895 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1896 (struct sockaddr __user *)a[4], (int __user *)a[5]);
1899 err = sys_shutdown(a0,a1);
1901 case SYS_SETSOCKOPT:
1902 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
1904 case SYS_GETSOCKOPT:
1905 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
1908 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
1911 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
1920 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
1923 * This function is called by a protocol handler that wants to
1924 * advertise its address family, and have it linked into the
1928 int sock_register(struct net_proto_family *ops)
1932 if (ops->family >= NPROTO) {
1933 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
1936 net_family_write_lock();
1938 if (net_families[ops->family] == NULL) {
1939 net_families[ops->family]=ops;
1942 net_family_write_unlock();
1943 printk(KERN_INFO "NET: Registered protocol family %d\n",
1949 * This function is called by a protocol handler that wants to
1950 * remove its address family, and have it unlinked from the
1954 int sock_unregister(int family)
1956 if (family < 0 || family >= NPROTO)
1959 net_family_write_lock();
1960 net_families[family]=NULL;
1961 net_family_write_unlock();
1962 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
1968 extern void sk_init(void);
1970 void __init sock_init(void)
1975 * Initialize all address (protocol) families.
1978 for (i = 0; i < NPROTO; i++)
1979 net_families[i] = NULL;
1982 * Initialize sock SLAB cache.
1989 * Initialize skbuff SLAB cache
1995 * Initialize the protocols module.
1999 register_filesystem(&sock_fs_type);
2000 sock_mnt = kern_mount(&sock_fs_type);
2001 /* The real protocol initialization is performed when
2002 * do_initcalls is run.
2005 #ifdef CONFIG_NETFILTER
2010 #ifdef CONFIG_PROC_FS
2011 void socket_seq_show(struct seq_file *seq)
2016 for (cpu = 0; cpu < NR_CPUS; cpu++)
2017 counter += per_cpu(sockets_in_use, cpu);
2019 /* It can be negative, by the way. 8) */
2023 seq_printf(seq, "sockets: used %d\n", counter);
2025 #endif /* CONFIG_PROC_FS */
2027 /* ABI emulation layers need these two */
2028 EXPORT_SYMBOL(move_addr_to_kernel);
2029 EXPORT_SYMBOL(move_addr_to_user);
2030 EXPORT_SYMBOL(sock_alloc);
2031 EXPORT_SYMBOL(sock_alloc_inode);
2032 EXPORT_SYMBOL(sock_create);
2033 EXPORT_SYMBOL(sock_create_kern);
2034 EXPORT_SYMBOL(sock_create_lite);
2035 EXPORT_SYMBOL(sock_map_fd);
2036 EXPORT_SYMBOL(sock_recvmsg);
2037 EXPORT_SYMBOL(sock_register);
2038 EXPORT_SYMBOL(sock_release);
2039 EXPORT_SYMBOL(sock_sendmsg);
2040 EXPORT_SYMBOL(sock_unregister);
2041 EXPORT_SYMBOL(sock_wake_async);
2042 EXPORT_SYMBOL(sockfd_lookup);