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>
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/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/divert.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
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 long sock_ioctl(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
,
129 .unlocked_ioctl
= sock_ioctl
,
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 DEFINE_SPINLOCK(net_family_lock
);
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
))
229 return audit_sockaddr(ulen
, kaddr
);
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 __read_mostly
;
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
.flags
= 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 __read_mostly
;
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 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
387 this.hash
= SOCK_INODE(sock
)->i_ino
;
389 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
390 if (!file
->f_dentry
) {
396 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
397 d_add(file
->f_dentry
, SOCK_INODE(sock
));
398 file
->f_vfsmnt
= mntget(sock_mnt
);
399 file
->f_mapping
= file
->f_dentry
->d_inode
->i_mapping
;
402 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
403 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
404 file
->f_flags
= O_RDWR
;
406 file
->private_data
= sock
;
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 if (file
->f_op
== &socket_file_ops
)
440 return file
->private_data
; /* set in sock_map_fd */
442 inode
= file
->f_dentry
->d_inode
;
443 if (!S_ISSOCK(inode
->i_mode
)) {
449 sock
= SOCKET_I(inode
);
450 if (sock
->file
!= file
) {
451 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
458 * sock_alloc - allocate a socket
460 * Allocate a new inode and socket object. The two are bound together
461 * and initialised. The socket is then returned. If we are out of inodes
465 static struct socket
*sock_alloc(void)
467 struct inode
* inode
;
468 struct socket
* sock
;
470 inode
= new_inode(sock_mnt
->mnt_sb
);
474 sock
= SOCKET_I(inode
);
476 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
477 inode
->i_uid
= current
->fsuid
;
478 inode
->i_gid
= current
->fsgid
;
480 get_cpu_var(sockets_in_use
)++;
481 put_cpu_var(sockets_in_use
);
486 * In theory you can't get an open on this inode, but /proc provides
487 * a back door. Remember to keep it shut otherwise you'll let the
488 * creepy crawlies in.
491 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
496 struct file_operations bad_sock_fops
= {
497 .owner
= THIS_MODULE
,
498 .open
= sock_no_open
,
502 * sock_release - close a socket
503 * @sock: socket to close
505 * The socket is released from the protocol stack if it has a release
506 * callback, and the inode is then released if the socket is bound to
507 * an inode not a file.
510 void sock_release(struct socket
*sock
)
513 struct module
*owner
= sock
->ops
->owner
;
515 sock
->ops
->release(sock
);
520 if (sock
->fasync_list
)
521 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
523 get_cpu_var(sockets_in_use
)--;
524 put_cpu_var(sockets_in_use
);
526 iput(SOCK_INODE(sock
));
532 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
533 struct msghdr
*msg
, size_t size
)
535 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
543 err
= security_socket_sendmsg(sock
, msg
, size
);
547 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
550 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
553 struct sock_iocb siocb
;
556 init_sync_kiocb(&iocb
, NULL
);
557 iocb
.private = &siocb
;
558 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
559 if (-EIOCBQUEUED
== ret
)
560 ret
= wait_on_sync_kiocb(&iocb
);
564 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
565 struct kvec
*vec
, size_t num
, size_t size
)
567 mm_segment_t oldfs
= get_fs();
572 * the following is safe, since for compiler definitions of kvec and
573 * iovec are identical, yielding the same in-core layout and alignment
575 msg
->msg_iov
= (struct iovec
*)vec
,
576 msg
->msg_iovlen
= num
;
577 result
= sock_sendmsg(sock
, msg
, size
);
582 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
583 struct msghdr
*msg
, size_t size
, int flags
)
586 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
594 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
598 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
601 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
602 size_t size
, int flags
)
605 struct sock_iocb siocb
;
608 init_sync_kiocb(&iocb
, NULL
);
609 iocb
.private = &siocb
;
610 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
611 if (-EIOCBQUEUED
== ret
)
612 ret
= wait_on_sync_kiocb(&iocb
);
616 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
617 struct kvec
*vec
, size_t num
,
618 size_t size
, int flags
)
620 mm_segment_t oldfs
= get_fs();
625 * the following is safe, since for compiler definitions of kvec and
626 * iovec are identical, yielding the same in-core layout and alignment
628 msg
->msg_iov
= (struct iovec
*)vec
,
629 msg
->msg_iovlen
= num
;
630 result
= sock_recvmsg(sock
, msg
, size
, flags
);
635 static void sock_aio_dtor(struct kiocb
*iocb
)
637 kfree(iocb
->private);
640 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
641 int offset
, size_t size
, loff_t
*ppos
, int more
)
646 sock
= file
->private_data
;
648 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
652 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
655 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
656 char __user
*ubuf
, size_t size
, struct sock_iocb
*siocb
)
658 if (!is_sync_kiocb(iocb
)) {
659 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
662 iocb
->ki_dtor
= sock_aio_dtor
;
666 siocb
->async_iov
.iov_base
= ubuf
;
667 siocb
->async_iov
.iov_len
= size
;
669 iocb
->private = siocb
;
673 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
674 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
676 struct socket
*sock
= file
->private_data
;
680 for (i
= 0 ; i
< nr_segs
; i
++)
681 size
+= iov
[i
].iov_len
;
683 msg
->msg_name
= NULL
;
684 msg
->msg_namelen
= 0;
685 msg
->msg_control
= NULL
;
686 msg
->msg_controllen
= 0;
687 msg
->msg_iov
= (struct iovec
*) iov
;
688 msg
->msg_iovlen
= nr_segs
;
689 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
691 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
694 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
695 unsigned long nr_segs
, loff_t
*ppos
)
698 struct sock_iocb siocb
;
702 init_sync_kiocb(&iocb
, NULL
);
703 iocb
.private = &siocb
;
705 ret
= do_sock_read(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
706 if (-EIOCBQUEUED
== ret
)
707 ret
= wait_on_sync_kiocb(&iocb
);
711 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
712 size_t count
, loff_t pos
)
714 struct sock_iocb siocb
, *x
;
718 if (count
== 0) /* Match SYS5 behaviour */
721 x
= alloc_sock_iocb(iocb
, ubuf
, count
, &siocb
);
724 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
,
728 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
729 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
731 struct socket
*sock
= file
->private_data
;
735 for (i
= 0 ; i
< nr_segs
; i
++)
736 size
+= iov
[i
].iov_len
;
738 msg
->msg_name
= NULL
;
739 msg
->msg_namelen
= 0;
740 msg
->msg_control
= NULL
;
741 msg
->msg_controllen
= 0;
742 msg
->msg_iov
= (struct iovec
*) iov
;
743 msg
->msg_iovlen
= nr_segs
;
744 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
745 if (sock
->type
== SOCK_SEQPACKET
)
746 msg
->msg_flags
|= MSG_EOR
;
748 return __sock_sendmsg(iocb
, sock
, msg
, size
);
751 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
752 unsigned long nr_segs
, loff_t
*ppos
)
756 struct sock_iocb siocb
;
759 init_sync_kiocb(&iocb
, NULL
);
760 iocb
.private = &siocb
;
762 ret
= do_sock_write(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
763 if (-EIOCBQUEUED
== ret
)
764 ret
= wait_on_sync_kiocb(&iocb
);
768 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
769 size_t count
, loff_t pos
)
771 struct sock_iocb siocb
, *x
;
775 if (count
== 0) /* Match SYS5 behaviour */
778 x
= alloc_sock_iocb(iocb
, (void __user
*)ubuf
, count
, &siocb
);
782 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
,
788 * Atomic setting of ioctl hooks to avoid race
789 * with module unload.
792 static DECLARE_MUTEX(br_ioctl_mutex
);
793 static int (*br_ioctl_hook
)(unsigned int cmd
, void __user
*arg
) = NULL
;
795 void brioctl_set(int (*hook
)(unsigned int, void __user
*))
797 down(&br_ioctl_mutex
);
798 br_ioctl_hook
= hook
;
801 EXPORT_SYMBOL(brioctl_set
);
803 static DECLARE_MUTEX(vlan_ioctl_mutex
);
804 static int (*vlan_ioctl_hook
)(void __user
*arg
);
806 void vlan_ioctl_set(int (*hook
)(void __user
*))
808 down(&vlan_ioctl_mutex
);
809 vlan_ioctl_hook
= hook
;
810 up(&vlan_ioctl_mutex
);
812 EXPORT_SYMBOL(vlan_ioctl_set
);
814 static DECLARE_MUTEX(dlci_ioctl_mutex
);
815 static int (*dlci_ioctl_hook
)(unsigned int, void __user
*);
817 void dlci_ioctl_set(int (*hook
)(unsigned int, void __user
*))
819 down(&dlci_ioctl_mutex
);
820 dlci_ioctl_hook
= hook
;
821 up(&dlci_ioctl_mutex
);
823 EXPORT_SYMBOL(dlci_ioctl_set
);
826 * With an ioctl, arg may well be a user mode pointer, but we don't know
827 * what to do with it - that's up to the protocol still.
830 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
833 void __user
*argp
= (void __user
*)arg
;
836 sock
= file
->private_data
;
837 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
838 err
= dev_ioctl(cmd
, argp
);
840 #ifdef CONFIG_WIRELESS_EXT
841 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
842 err
= dev_ioctl(cmd
, argp
);
844 #endif /* CONFIG_WIRELESS_EXT */
849 if (get_user(pid
, (int __user
*)argp
))
851 err
= f_setown(sock
->file
, pid
, 1);
855 err
= put_user(sock
->file
->f_owner
.pid
, (int __user
*)argp
);
863 request_module("bridge");
865 down(&br_ioctl_mutex
);
867 err
= br_ioctl_hook(cmd
, argp
);
873 if (!vlan_ioctl_hook
)
874 request_module("8021q");
876 down(&vlan_ioctl_mutex
);
878 err
= vlan_ioctl_hook(argp
);
879 up(&vlan_ioctl_mutex
);
883 /* Convert this to call through a hook */
884 err
= divert_ioctl(cmd
, argp
);
889 if (!dlci_ioctl_hook
)
890 request_module("dlci");
892 if (dlci_ioctl_hook
) {
893 down(&dlci_ioctl_mutex
);
894 err
= dlci_ioctl_hook(cmd
, argp
);
895 up(&dlci_ioctl_mutex
);
899 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
902 * If this ioctl is unknown try to hand it down
905 if (err
== -ENOIOCTLCMD
)
906 err
= dev_ioctl(cmd
, argp
);
912 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
915 struct socket
*sock
= NULL
;
917 err
= security_socket_create(family
, type
, protocol
, 1);
927 security_socket_post_create(sock
, family
, type
, protocol
, 1);
934 /* No kernel lock held - perfect */
935 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
940 * We can't return errors to poll, so it's either yes or no.
942 sock
= file
->private_data
;
943 return sock
->ops
->poll(file
, sock
, wait
);
946 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
948 struct socket
*sock
= file
->private_data
;
950 return sock
->ops
->mmap(file
, sock
, vma
);
953 static int sock_close(struct inode
*inode
, struct file
*filp
)
956 * It was possible the inode is NULL we were
957 * closing an unfinished socket.
962 printk(KERN_DEBUG
"sock_close: NULL inode\n");
965 sock_fasync(-1, filp
, 0);
966 sock_release(SOCKET_I(inode
));
971 * Update the socket async list
973 * Fasync_list locking strategy.
975 * 1. fasync_list is modified only under process context socket lock
976 * i.e. under semaphore.
977 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
978 * or under socket lock.
979 * 3. fasync_list can be used from softirq context, so that
980 * modification under socket lock have to be enhanced with
981 * write_lock_bh(&sk->sk_callback_lock).
985 static int sock_fasync(int fd
, struct file
*filp
, int on
)
987 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
993 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
998 sock
= filp
->private_data
;
1000 if ((sk
=sock
->sk
) == NULL
) {
1007 prev
=&(sock
->fasync_list
);
1009 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
1010 if (fa
->fa_file
==filp
)
1017 write_lock_bh(&sk
->sk_callback_lock
);
1019 write_unlock_bh(&sk
->sk_callback_lock
);
1026 fna
->magic
=FASYNC_MAGIC
;
1027 fna
->fa_next
=sock
->fasync_list
;
1028 write_lock_bh(&sk
->sk_callback_lock
);
1029 sock
->fasync_list
=fna
;
1030 write_unlock_bh(&sk
->sk_callback_lock
);
1036 write_lock_bh(&sk
->sk_callback_lock
);
1038 write_unlock_bh(&sk
->sk_callback_lock
);
1044 release_sock(sock
->sk
);
1048 /* This function may be called only under socket lock or callback_lock */
1050 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1052 if (!sock
|| !sock
->fasync_list
)
1058 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1062 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1067 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1070 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1075 static int __sock_create(int family
, int type
, int protocol
, struct socket
**res
, int kern
)
1078 struct socket
*sock
;
1081 * Check protocol is in range
1083 if (family
< 0 || family
>= NPROTO
)
1084 return -EAFNOSUPPORT
;
1085 if (type
< 0 || type
>= SOCK_MAX
)
1090 This uglymoron is moved from INET layer to here to avoid
1091 deadlock in module load.
1093 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1097 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
1102 err
= security_socket_create(family
, type
, protocol
, kern
);
1106 #if defined(CONFIG_KMOD)
1107 /* Attempt to load a protocol module if the find failed.
1109 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1110 * requested real, full-featured networking support upon configuration.
1111 * Otherwise module support will break!
1113 if (net_families
[family
]==NULL
)
1115 request_module("net-pf-%d",family
);
1119 net_family_read_lock();
1120 if (net_families
[family
] == NULL
) {
1121 err
= -EAFNOSUPPORT
;
1126 * Allocate the socket and allow the family to set things up. if
1127 * the protocol is 0, the family is instructed to select an appropriate
1131 if (!(sock
= sock_alloc())) {
1132 printk(KERN_WARNING
"socket: no more sockets\n");
1133 err
= -ENFILE
; /* Not exactly a match, but its the
1134 closest posix thing */
1141 * We will call the ->create function, that possibly is in a loadable
1142 * module, so we have to bump that loadable module refcnt first.
1144 err
= -EAFNOSUPPORT
;
1145 if (!try_module_get(net_families
[family
]->owner
))
1148 if ((err
= net_families
[family
]->create(sock
, protocol
)) < 0) {
1150 goto out_module_put
;
1154 * Now to bump the refcnt of the [loadable] module that owns this
1155 * socket at sock_release time we decrement its refcnt.
1157 if (!try_module_get(sock
->ops
->owner
)) {
1159 goto out_module_put
;
1162 * Now that we're done with the ->create function, the [loadable]
1163 * module can have its refcnt decremented
1165 module_put(net_families
[family
]->owner
);
1167 security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1170 net_family_read_unlock();
1173 module_put(net_families
[family
]->owner
);
1179 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1181 return __sock_create(family
, type
, protocol
, res
, 0);
1184 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1186 return __sock_create(family
, type
, protocol
, res
, 1);
1189 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1192 struct socket
*sock
;
1194 retval
= sock_create(family
, type
, protocol
, &sock
);
1198 retval
= sock_map_fd(sock
);
1203 /* It may be already another descriptor 8) Not kernel problem. */
1212 * Create a pair of connected sockets.
1215 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1217 struct socket
*sock1
, *sock2
;
1221 * Obtain the first socket and check if the underlying protocol
1222 * supports the socketpair call.
1225 err
= sock_create(family
, type
, protocol
, &sock1
);
1229 err
= sock_create(family
, type
, protocol
, &sock2
);
1233 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1235 goto out_release_both
;
1239 err
= sock_map_fd(sock1
);
1241 goto out_release_both
;
1244 err
= sock_map_fd(sock2
);
1249 /* fd1 and fd2 may be already another descriptors.
1250 * Not kernel problem.
1253 err
= put_user(fd1
, &usockvec
[0]);
1255 err
= put_user(fd2
, &usockvec
[1]);
1264 sock_release(sock2
);
1269 sock_release(sock2
);
1271 sock_release(sock1
);
1278 * Bind a name to a socket. Nothing much to do here since it's
1279 * the protocol's responsibility to handle the local address.
1281 * We move the socket address to kernel space before we call
1282 * the protocol layer (having also checked the address is ok).
1285 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1287 struct socket
*sock
;
1288 char address
[MAX_SOCK_ADDR
];
1291 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
1293 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0) {
1294 err
= security_socket_bind(sock
, (struct sockaddr
*)address
, addrlen
);
1299 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
1308 * Perform a listen. Basically, we allow the protocol to do anything
1309 * necessary for a listen, and if that works, we mark the socket as
1310 * ready for listening.
1313 int sysctl_somaxconn
= SOMAXCONN
;
1315 asmlinkage
long sys_listen(int fd
, int backlog
)
1317 struct socket
*sock
;
1320 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
1321 if ((unsigned) backlog
> sysctl_somaxconn
)
1322 backlog
= sysctl_somaxconn
;
1324 err
= security_socket_listen(sock
, backlog
);
1330 err
=sock
->ops
->listen(sock
, backlog
);
1338 * For accept, we attempt to create a new socket, set up the link
1339 * with the client, wake up the client, then return the new
1340 * connected fd. We collect the address of the connector in kernel
1341 * space and move it to user at the very end. This is unclean because
1342 * we open the socket then return an error.
1344 * 1003.1g adds the ability to recvmsg() to query connection pending
1345 * status to recvmsg. We need to add that support in a way thats
1346 * clean when we restucture accept also.
1349 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
, int __user
*upeer_addrlen
)
1351 struct socket
*sock
, *newsock
;
1353 char address
[MAX_SOCK_ADDR
];
1355 sock
= sockfd_lookup(fd
, &err
);
1360 if (!(newsock
= sock_alloc()))
1363 newsock
->type
= sock
->type
;
1364 newsock
->ops
= sock
->ops
;
1367 * We don't need try_module_get here, as the listening socket (sock)
1368 * has the protocol module (sock->ops->owner) held.
1370 __module_get(newsock
->ops
->owner
);
1372 err
= security_socket_accept(sock
, newsock
);
1376 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1380 if (upeer_sockaddr
) {
1381 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1382 err
= -ECONNABORTED
;
1385 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1390 /* File flags are not inherited via accept() unlike another OSes. */
1392 if ((err
= sock_map_fd(newsock
)) < 0)
1395 security_socket_post_accept(sock
, newsock
);
1402 sock_release(newsock
);
1408 * Attempt to connect to a socket with the server address. The address
1409 * is in user space so we verify it is OK and move it to kernel space.
1411 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1414 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1415 * other SEQPACKET protocols that take time to connect() as it doesn't
1416 * include the -EINPROGRESS status for such sockets.
1419 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1421 struct socket
*sock
;
1422 char address
[MAX_SOCK_ADDR
];
1425 sock
= sockfd_lookup(fd
, &err
);
1428 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1432 err
= security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1436 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1437 sock
->file
->f_flags
);
1445 * Get the local address ('name') of a socket object. Move the obtained
1446 * name to user space.
1449 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1451 struct socket
*sock
;
1452 char address
[MAX_SOCK_ADDR
];
1455 sock
= sockfd_lookup(fd
, &err
);
1459 err
= security_socket_getsockname(sock
);
1463 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1466 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1475 * Get the remote address ('name') of a socket object. Move the obtained
1476 * name to user space.
1479 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1481 struct socket
*sock
;
1482 char address
[MAX_SOCK_ADDR
];
1485 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1487 err
= security_socket_getpeername(sock
);
1493 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1495 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1502 * Send a datagram to a given address. We move the address into kernel
1503 * space and check the user space data area is readable before invoking
1507 asmlinkage
long sys_sendto(int fd
, void __user
* buff
, size_t len
, unsigned flags
,
1508 struct sockaddr __user
*addr
, int addr_len
)
1510 struct socket
*sock
;
1511 char address
[MAX_SOCK_ADDR
];
1516 sock
= sockfd_lookup(fd
, &err
);
1524 msg
.msg_control
=NULL
;
1525 msg
.msg_controllen
=0;
1529 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1532 msg
.msg_name
=address
;
1533 msg
.msg_namelen
=addr_len
;
1535 if (sock
->file
->f_flags
& O_NONBLOCK
)
1536 flags
|= MSG_DONTWAIT
;
1537 msg
.msg_flags
= flags
;
1538 err
= sock_sendmsg(sock
, &msg
, len
);
1547 * Send a datagram down a socket.
1550 asmlinkage
long sys_send(int fd
, void __user
* buff
, size_t len
, unsigned flags
)
1552 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1556 * Receive a frame from the socket and optionally record the address of the
1557 * sender. We verify the buffers are writable and if needed move the
1558 * sender address from kernel to user space.
1561 asmlinkage
long sys_recvfrom(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
,
1562 struct sockaddr __user
*addr
, int __user
*addr_len
)
1564 struct socket
*sock
;
1567 char address
[MAX_SOCK_ADDR
];
1570 sock
= sockfd_lookup(fd
, &err
);
1574 msg
.msg_control
=NULL
;
1575 msg
.msg_controllen
=0;
1580 msg
.msg_name
=address
;
1581 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1582 if (sock
->file
->f_flags
& O_NONBLOCK
)
1583 flags
|= MSG_DONTWAIT
;
1584 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1586 if(err
>= 0 && addr
!= NULL
)
1588 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1598 * Receive a datagram from a socket.
1601 asmlinkage
long sys_recv(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
)
1603 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1607 * Set a socket option. Because we don't know the option lengths we have
1608 * to pass the user mode parameter for the protocols to sort out.
1611 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char __user
*optval
, int optlen
)
1614 struct socket
*sock
;
1619 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1621 err
= security_socket_setsockopt(sock
,level
,optname
);
1627 if (level
== SOL_SOCKET
)
1628 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1630 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1637 * Get a socket option. Because we don't know the option lengths we have
1638 * to pass a user mode parameter for the protocols to sort out.
1641 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
, int __user
*optlen
)
1644 struct socket
*sock
;
1646 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1648 err
= security_socket_getsockopt(sock
, level
,
1655 if (level
== SOL_SOCKET
)
1656 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1658 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1666 * Shutdown a socket.
1669 asmlinkage
long sys_shutdown(int fd
, int how
)
1672 struct socket
*sock
;
1674 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1676 err
= security_socket_shutdown(sock
, how
);
1682 err
=sock
->ops
->shutdown(sock
, how
);
1688 /* A couple of helpful macros for getting the address of the 32/64 bit
1689 * fields which are the same type (int / unsigned) on our platforms.
1691 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1692 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1693 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1697 * BSD sendmsg interface
1700 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1702 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1703 struct socket
*sock
;
1704 char address
[MAX_SOCK_ADDR
];
1705 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1706 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1707 __attribute__ ((aligned (sizeof(__kernel_size_t
))));
1708 /* 20 is size of ipv6_pktinfo */
1709 unsigned char *ctl_buf
= ctl
;
1710 struct msghdr msg_sys
;
1711 int err
, ctl_len
, iov_size
, total_len
;
1714 if (MSG_CMSG_COMPAT
& flags
) {
1715 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1717 } else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1720 sock
= sockfd_lookup(fd
, &err
);
1724 /* do not move before msg_sys is valid */
1726 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1729 /* Check whether to allocate the iovec area*/
1731 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1732 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1733 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1738 /* This will also move the address data into kernel space */
1739 if (MSG_CMSG_COMPAT
& flags
) {
1740 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1742 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1749 if (msg_sys
.msg_controllen
> INT_MAX
)
1751 ctl_len
= msg_sys
.msg_controllen
;
1752 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1753 err
= cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
, sizeof(ctl
));
1756 ctl_buf
= msg_sys
.msg_control
;
1757 ctl_len
= msg_sys
.msg_controllen
;
1758 } else if (ctl_len
) {
1759 if (ctl_len
> sizeof(ctl
))
1761 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1762 if (ctl_buf
== NULL
)
1767 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1768 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1769 * checking falls down on this.
1771 if (copy_from_user(ctl_buf
, (void __user
*) msg_sys
.msg_control
, ctl_len
))
1773 msg_sys
.msg_control
= ctl_buf
;
1775 msg_sys
.msg_flags
= flags
;
1777 if (sock
->file
->f_flags
& O_NONBLOCK
)
1778 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1779 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1783 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1785 if (iov
!= iovstack
)
1786 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1794 * BSD recvmsg interface
1797 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned int flags
)
1799 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1800 struct socket
*sock
;
1801 struct iovec iovstack
[UIO_FASTIOV
];
1802 struct iovec
*iov
=iovstack
;
1803 struct msghdr msg_sys
;
1804 unsigned long cmsg_ptr
;
1805 int err
, iov_size
, total_len
, len
;
1807 /* kernel mode address */
1808 char addr
[MAX_SOCK_ADDR
];
1810 /* user mode address pointers */
1811 struct sockaddr __user
*uaddr
;
1812 int __user
*uaddr_len
;
1814 if (MSG_CMSG_COMPAT
& flags
) {
1815 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1818 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1821 sock
= sockfd_lookup(fd
, &err
);
1826 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1829 /* Check whether to allocate the iovec area*/
1831 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1832 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1833 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1839 * Save the user-mode address (verify_iovec will change the
1840 * kernel msghdr to use the kernel address space)
1843 uaddr
= (void __user
*) msg_sys
.msg_name
;
1844 uaddr_len
= COMPAT_NAMELEN(msg
);
1845 if (MSG_CMSG_COMPAT
& flags
) {
1846 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1848 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1853 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1854 msg_sys
.msg_flags
= 0;
1855 if (MSG_CMSG_COMPAT
& flags
)
1856 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1858 if (sock
->file
->f_flags
& O_NONBLOCK
)
1859 flags
|= MSG_DONTWAIT
;
1860 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1865 if (uaddr
!= NULL
) {
1866 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1870 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1874 if (MSG_CMSG_COMPAT
& flags
)
1875 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1876 &msg_compat
->msg_controllen
);
1878 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1879 &msg
->msg_controllen
);
1885 if (iov
!= iovstack
)
1886 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1893 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1895 /* Argument list sizes for sys_socketcall */
1896 #define AL(x) ((x) * sizeof(unsigned long))
1897 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1898 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1899 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1903 * System call vectors.
1905 * Argument checking cleaned up. Saved 20% in size.
1906 * This function doesn't need to set the kernel lock because
1907 * it is set by the callees.
1910 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1913 unsigned long a0
,a1
;
1916 if(call
<1||call
>SYS_RECVMSG
)
1919 /* copy_from_user should be SMP safe. */
1920 if (copy_from_user(a
, args
, nargs
[call
]))
1923 err
= audit_socketcall(nargs
[call
]/sizeof(unsigned long), a
);
1933 err
= sys_socket(a0
,a1
,a
[2]);
1936 err
= sys_bind(a0
,(struct sockaddr __user
*)a1
, a
[2]);
1939 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
1942 err
= sys_listen(a0
,a1
);
1945 err
= sys_accept(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1947 case SYS_GETSOCKNAME
:
1948 err
= sys_getsockname(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1950 case SYS_GETPEERNAME
:
1951 err
= sys_getpeername(a0
, (struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1953 case SYS_SOCKETPAIR
:
1954 err
= sys_socketpair(a0
,a1
, a
[2], (int __user
*)a
[3]);
1957 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
1960 err
= sys_sendto(a0
,(void __user
*)a1
, a
[2], a
[3],
1961 (struct sockaddr __user
*)a
[4], a
[5]);
1964 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
1967 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
1968 (struct sockaddr __user
*)a
[4], (int __user
*)a
[5]);
1971 err
= sys_shutdown(a0
,a1
);
1973 case SYS_SETSOCKOPT
:
1974 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
1976 case SYS_GETSOCKOPT
:
1977 err
= sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], (int __user
*)a
[4]);
1980 err
= sys_sendmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1983 err
= sys_recvmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1992 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
1995 * This function is called by a protocol handler that wants to
1996 * advertise its address family, and have it linked into the
2000 int sock_register(struct net_proto_family
*ops
)
2004 if (ops
->family
>= NPROTO
) {
2005 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2008 net_family_write_lock();
2010 if (net_families
[ops
->family
] == NULL
) {
2011 net_families
[ops
->family
]=ops
;
2014 net_family_write_unlock();
2015 printk(KERN_INFO
"NET: Registered protocol family %d\n",
2021 * This function is called by a protocol handler that wants to
2022 * remove its address family, and have it unlinked from the
2026 int sock_unregister(int family
)
2028 if (family
< 0 || family
>= NPROTO
)
2031 net_family_write_lock();
2032 net_families
[family
]=NULL
;
2033 net_family_write_unlock();
2034 printk(KERN_INFO
"NET: Unregistered protocol family %d\n",
2039 static int __init
sock_init(void)
2042 * Initialize sock SLAB cache.
2048 * Initialize skbuff SLAB cache
2053 * Initialize the protocols module.
2057 register_filesystem(&sock_fs_type
);
2058 sock_mnt
= kern_mount(&sock_fs_type
);
2060 /* The real protocol initialization is performed in later initcalls.
2063 #ifdef CONFIG_NETFILTER
2070 core_initcall(sock_init
); /* early initcall */
2072 #ifdef CONFIG_PROC_FS
2073 void socket_seq_show(struct seq_file
*seq
)
2079 counter
+= per_cpu(sockets_in_use
, cpu
);
2081 /* It can be negative, by the way. 8) */
2085 seq_printf(seq
, "sockets: used %d\n", counter
);
2087 #endif /* CONFIG_PROC_FS */
2089 /* ABI emulation layers need these two */
2090 EXPORT_SYMBOL(move_addr_to_kernel
);
2091 EXPORT_SYMBOL(move_addr_to_user
);
2092 EXPORT_SYMBOL(sock_create
);
2093 EXPORT_SYMBOL(sock_create_kern
);
2094 EXPORT_SYMBOL(sock_create_lite
);
2095 EXPORT_SYMBOL(sock_map_fd
);
2096 EXPORT_SYMBOL(sock_recvmsg
);
2097 EXPORT_SYMBOL(sock_register
);
2098 EXPORT_SYMBOL(sock_release
);
2099 EXPORT_SYMBOL(sock_sendmsg
);
2100 EXPORT_SYMBOL(sock_unregister
);
2101 EXPORT_SYMBOL(sock_wake_async
);
2102 EXPORT_SYMBOL(sockfd_lookup
);
2103 EXPORT_SYMBOL(kernel_sendmsg
);
2104 EXPORT_SYMBOL(kernel_recvmsg
);