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
;
404 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
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
)
551 struct sock_iocb siocb
;
554 init_sync_kiocb(&iocb
, NULL
);
555 iocb
.private = &siocb
;
556 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
557 if (-EIOCBQUEUED
== ret
)
558 ret
= wait_on_sync_kiocb(&iocb
);
562 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
563 struct kvec
*vec
, size_t num
, size_t size
)
565 mm_segment_t oldfs
= get_fs();
570 * the following is safe, since for compiler definitions of kvec and
571 * iovec are identical, yielding the same in-core layout and alignment
573 msg
->msg_iov
= (struct iovec
*)vec
,
574 msg
->msg_iovlen
= num
;
575 result
= sock_sendmsg(sock
, msg
, size
);
580 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
581 struct msghdr
*msg
, size_t size
, int flags
)
584 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
592 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
596 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
599 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
600 size_t size
, int flags
)
603 struct sock_iocb siocb
;
606 init_sync_kiocb(&iocb
, NULL
);
607 iocb
.private = &siocb
;
608 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
609 if (-EIOCBQUEUED
== ret
)
610 ret
= wait_on_sync_kiocb(&iocb
);
614 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
615 struct kvec
*vec
, size_t num
,
616 size_t size
, int flags
)
618 mm_segment_t oldfs
= get_fs();
623 * the following is safe, since for compiler definitions of kvec and
624 * iovec are identical, yielding the same in-core layout and alignment
626 msg
->msg_iov
= (struct iovec
*)vec
,
627 msg
->msg_iovlen
= num
;
628 result
= sock_recvmsg(sock
, msg
, size
, flags
);
633 static void sock_aio_dtor(struct kiocb
*iocb
)
635 kfree(iocb
->private);
639 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
640 * area ubuf...ubuf+size-1 is writable before asking the protocol.
643 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
644 size_t size
, loff_t pos
)
646 struct sock_iocb
*x
, siocb
;
652 if (size
==0) /* Match SYS5 behaviour */
655 if (is_sync_kiocb(iocb
))
658 x
= kmalloc(sizeof(struct sock_iocb
), GFP_KERNEL
);
661 iocb
->ki_dtor
= sock_aio_dtor
;
665 sock
= SOCKET_I(iocb
->ki_filp
->f_dentry
->d_inode
);
667 x
->async_msg
.msg_name
= NULL
;
668 x
->async_msg
.msg_namelen
= 0;
669 x
->async_msg
.msg_iov
= &x
->async_iov
;
670 x
->async_msg
.msg_iovlen
= 1;
671 x
->async_msg
.msg_control
= NULL
;
672 x
->async_msg
.msg_controllen
= 0;
673 x
->async_iov
.iov_base
= ubuf
;
674 x
->async_iov
.iov_len
= size
;
675 flags
= !(iocb
->ki_filp
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
677 return __sock_recvmsg(iocb
, sock
, &x
->async_msg
, size
, flags
);
682 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
683 * is readable by the user process.
686 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
687 size_t size
, loff_t pos
)
689 struct sock_iocb
*x
, siocb
;
694 if(size
==0) /* Match SYS5 behaviour */
697 if (is_sync_kiocb(iocb
))
700 x
= kmalloc(sizeof(struct sock_iocb
), GFP_KERNEL
);
703 iocb
->ki_dtor
= sock_aio_dtor
;
707 sock
= SOCKET_I(iocb
->ki_filp
->f_dentry
->d_inode
);
709 x
->async_msg
.msg_name
= NULL
;
710 x
->async_msg
.msg_namelen
= 0;
711 x
->async_msg
.msg_iov
= &x
->async_iov
;
712 x
->async_msg
.msg_iovlen
= 1;
713 x
->async_msg
.msg_control
= NULL
;
714 x
->async_msg
.msg_controllen
= 0;
715 x
->async_msg
.msg_flags
= !(iocb
->ki_filp
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
716 if (sock
->type
== SOCK_SEQPACKET
)
717 x
->async_msg
.msg_flags
|= MSG_EOR
;
718 x
->async_iov
.iov_base
= (void __user
*)ubuf
;
719 x
->async_iov
.iov_len
= size
;
721 return __sock_sendmsg(iocb
, sock
, &x
->async_msg
, size
);
724 ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
725 int offset
, size_t size
, loff_t
*ppos
, int more
)
730 sock
= SOCKET_I(file
->f_dentry
->d_inode
);
732 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
736 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
739 int sock_readv_writev(int type
, struct inode
* inode
, struct file
* file
,
740 const struct iovec
* iov
, long count
, size_t size
)
745 sock
= SOCKET_I(inode
);
749 msg
.msg_control
= NULL
;
750 msg
.msg_controllen
= 0;
751 msg
.msg_iov
= (struct iovec
*) iov
;
752 msg
.msg_iovlen
= count
;
753 msg
.msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
755 /* read() does a VERIFY_WRITE */
756 if (type
== VERIFY_WRITE
)
757 return sock_recvmsg(sock
, &msg
, size
, msg
.msg_flags
);
759 if (sock
->type
== SOCK_SEQPACKET
)
760 msg
.msg_flags
|= MSG_EOR
;
762 return sock_sendmsg(sock
, &msg
, size
);
765 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
766 unsigned long count
, loff_t
*ppos
)
770 for (i
= 0 ; i
< count
; i
++)
771 tot_len
+= vector
[i
].iov_len
;
772 return sock_readv_writev(VERIFY_WRITE
, file
->f_dentry
->d_inode
,
773 file
, vector
, count
, tot_len
);
776 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
777 unsigned long count
, loff_t
*ppos
)
781 for (i
= 0 ; i
< count
; i
++)
782 tot_len
+= vector
[i
].iov_len
;
783 return sock_readv_writev(VERIFY_READ
, file
->f_dentry
->d_inode
,
784 file
, vector
, count
, tot_len
);
789 * Atomic setting of ioctl hooks to avoid race
790 * with module unload.
793 static DECLARE_MUTEX(br_ioctl_mutex
);
794 static int (*br_ioctl_hook
)(unsigned int cmd
, void __user
*arg
) = NULL
;
796 void brioctl_set(int (*hook
)(unsigned int, void __user
*))
798 down(&br_ioctl_mutex
);
799 br_ioctl_hook
= hook
;
802 EXPORT_SYMBOL(brioctl_set
);
804 static DECLARE_MUTEX(vlan_ioctl_mutex
);
805 static int (*vlan_ioctl_hook
)(void __user
*arg
);
807 void vlan_ioctl_set(int (*hook
)(void __user
*))
809 down(&vlan_ioctl_mutex
);
810 vlan_ioctl_hook
= hook
;
811 up(&vlan_ioctl_mutex
);
813 EXPORT_SYMBOL(vlan_ioctl_set
);
815 static DECLARE_MUTEX(dlci_ioctl_mutex
);
816 static int (*dlci_ioctl_hook
)(unsigned int, void __user
*);
818 void dlci_ioctl_set(int (*hook
)(unsigned int, void __user
*))
820 down(&dlci_ioctl_mutex
);
821 dlci_ioctl_hook
= hook
;
822 up(&dlci_ioctl_mutex
);
824 EXPORT_SYMBOL(dlci_ioctl_set
);
827 * With an ioctl, arg may well be a user mode pointer, but we don't know
828 * what to do with it - that's up to the protocol still.
831 static int sock_ioctl(struct inode
*inode
, struct file
*file
, unsigned int cmd
,
835 void __user
*argp
= (void __user
*)arg
;
839 sock
= SOCKET_I(inode
);
840 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
841 err
= dev_ioctl(cmd
, argp
);
844 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
845 err
= dev_ioctl(cmd
, argp
);
847 #endif /* WIRELESS_EXT */
852 if (get_user(pid
, (int __user
*)argp
))
854 err
= f_setown(sock
->file
, pid
, 1);
858 err
= put_user(sock
->file
->f_owner
.pid
, (int __user
*)argp
);
866 request_module("bridge");
868 down(&br_ioctl_mutex
);
870 err
= br_ioctl_hook(cmd
, argp
);
876 if (!vlan_ioctl_hook
)
877 request_module("8021q");
879 down(&vlan_ioctl_mutex
);
881 err
= vlan_ioctl_hook(argp
);
882 up(&vlan_ioctl_mutex
);
886 /* Convert this to call through a hook */
887 err
= divert_ioctl(cmd
, argp
);
892 if (!dlci_ioctl_hook
)
893 request_module("dlci");
895 if (dlci_ioctl_hook
) {
896 down(&dlci_ioctl_mutex
);
897 err
= dlci_ioctl_hook(cmd
, argp
);
898 up(&dlci_ioctl_mutex
);
902 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
910 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
913 struct socket
*sock
= NULL
;
915 err
= security_socket_create(family
, type
, protocol
, 1);
925 security_socket_post_create(sock
, family
, type
, protocol
, 1);
932 /* No kernel lock held - perfect */
933 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
938 * We can't return errors to poll, so it's either yes or no.
940 sock
= SOCKET_I(file
->f_dentry
->d_inode
);
941 return sock
->ops
->poll(file
, sock
, wait
);
944 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
946 struct socket
*sock
= SOCKET_I(file
->f_dentry
->d_inode
);
948 return sock
->ops
->mmap(file
, sock
, vma
);
951 int sock_close(struct inode
*inode
, struct file
*filp
)
954 * It was possible the inode is NULL we were
955 * closing an unfinished socket.
960 printk(KERN_DEBUG
"sock_close: NULL inode\n");
963 sock_fasync(-1, filp
, 0);
964 sock_release(SOCKET_I(inode
));
969 * Update the socket async list
971 * Fasync_list locking strategy.
973 * 1. fasync_list is modified only under process context socket lock
974 * i.e. under semaphore.
975 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
976 * or under socket lock.
977 * 3. fasync_list can be used from softirq context, so that
978 * modification under socket lock have to be enhanced with
979 * write_lock_bh(&sk->sk_callback_lock).
983 static int sock_fasync(int fd
, struct file
*filp
, int on
)
985 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
991 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
996 sock
= SOCKET_I(filp
->f_dentry
->d_inode
);
998 if ((sk
=sock
->sk
) == NULL
) {
1006 prev
=&(sock
->fasync_list
);
1008 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
1009 if (fa
->fa_file
==filp
)
1016 write_lock_bh(&sk
->sk_callback_lock
);
1018 write_unlock_bh(&sk
->sk_callback_lock
);
1025 fna
->magic
=FASYNC_MAGIC
;
1026 fna
->fa_next
=sock
->fasync_list
;
1027 write_lock_bh(&sk
->sk_callback_lock
);
1028 sock
->fasync_list
=fna
;
1029 write_unlock_bh(&sk
->sk_callback_lock
);
1035 write_lock_bh(&sk
->sk_callback_lock
);
1037 write_unlock_bh(&sk
->sk_callback_lock
);
1043 release_sock(sock
->sk
);
1047 /* This function may be called only under socket lock or callback_lock */
1049 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1051 if (!sock
|| !sock
->fasync_list
)
1057 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1061 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1066 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1069 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1074 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
) {
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()))
1133 printk(KERN_WARNING
"socket: no more sockets\n");
1134 i
= -ENFILE
; /* Not exactly a match, but its the
1135 closest posix thing */
1142 * We will call the ->create function, that possibly is in a loadable
1143 * module, so we have to bump that loadable module refcnt first.
1146 if (!try_module_get(net_families
[family
]->owner
))
1149 if ((i
= net_families
[family
]->create(sock
, protocol
)) < 0)
1150 goto out_module_put
;
1152 * Now to bump the refcnt of the [loadable] module that owns this
1153 * socket at sock_release time we decrement its refcnt.
1155 if (!try_module_get(sock
->ops
->owner
)) {
1157 goto out_module_put
;
1160 * Now that we're done with the ->create function, the [loadable]
1161 * module can have its refcnt decremented
1163 module_put(net_families
[family
]->owner
);
1165 security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1168 net_family_read_unlock();
1171 module_put(net_families
[family
]->owner
);
1177 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1179 return __sock_create(family
, type
, protocol
, res
, 0);
1182 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1184 return __sock_create(family
, type
, protocol
, res
, 1);
1187 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1190 struct socket
*sock
;
1192 retval
= sock_create(family
, type
, protocol
, &sock
);
1196 retval
= sock_map_fd(sock
);
1201 /* It may be already another descriptor 8) Not kernel problem. */
1210 * Create a pair of connected sockets.
1213 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1215 struct socket
*sock1
, *sock2
;
1219 * Obtain the first socket and check if the underlying protocol
1220 * supports the socketpair call.
1223 err
= sock_create(family
, type
, protocol
, &sock1
);
1227 err
= sock_create(family
, type
, protocol
, &sock2
);
1231 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1233 goto out_release_both
;
1237 err
= sock_map_fd(sock1
);
1239 goto out_release_both
;
1242 err
= sock_map_fd(sock2
);
1247 /* fd1 and fd2 may be already another descriptors.
1248 * Not kernel problem.
1251 err
= put_user(fd1
, &usockvec
[0]);
1253 err
= put_user(fd2
, &usockvec
[1]);
1262 sock_release(sock2
);
1267 sock_release(sock2
);
1269 sock_release(sock1
);
1276 * Bind a name to a socket. Nothing much to do here since it's
1277 * the protocol's responsibility to handle the local address.
1279 * We move the socket address to kernel space before we call
1280 * the protocol layer (having also checked the address is ok).
1283 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1285 struct socket
*sock
;
1286 char address
[MAX_SOCK_ADDR
];
1289 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
1291 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0) {
1292 err
= security_socket_bind(sock
, (struct sockaddr
*)address
, addrlen
);
1297 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
1306 * Perform a listen. Basically, we allow the protocol to do anything
1307 * necessary for a listen, and if that works, we mark the socket as
1308 * ready for listening.
1311 int sysctl_somaxconn
= SOMAXCONN
;
1313 asmlinkage
long sys_listen(int fd
, int backlog
)
1315 struct socket
*sock
;
1318 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
1319 if ((unsigned) backlog
> sysctl_somaxconn
)
1320 backlog
= sysctl_somaxconn
;
1322 err
= security_socket_listen(sock
, backlog
);
1328 err
=sock
->ops
->listen(sock
, backlog
);
1336 * For accept, we attempt to create a new socket, set up the link
1337 * with the client, wake up the client, then return the new
1338 * connected fd. We collect the address of the connector in kernel
1339 * space and move it to user at the very end. This is unclean because
1340 * we open the socket then return an error.
1342 * 1003.1g adds the ability to recvmsg() to query connection pending
1343 * status to recvmsg. We need to add that support in a way thats
1344 * clean when we restucture accept also.
1347 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
, int __user
*upeer_addrlen
)
1349 struct socket
*sock
, *newsock
;
1351 char address
[MAX_SOCK_ADDR
];
1353 sock
= sockfd_lookup(fd
, &err
);
1358 if (!(newsock
= sock_alloc()))
1361 newsock
->type
= sock
->type
;
1362 newsock
->ops
= sock
->ops
;
1364 err
= security_socket_accept(sock
, newsock
);
1369 * We don't need try_module_get here, as the listening socket (sock)
1370 * has the protocol module (sock->ops->owner) held.
1372 __module_get(newsock
->ops
->owner
);
1374 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1378 if (upeer_sockaddr
) {
1379 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1380 err
= -ECONNABORTED
;
1383 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1388 /* File flags are not inherited via accept() unlike another OSes. */
1390 if ((err
= sock_map_fd(newsock
)) < 0)
1393 security_socket_post_accept(sock
, newsock
);
1400 sock_release(newsock
);
1406 * Attempt to connect to a socket with the server address. The address
1407 * is in user space so we verify it is OK and move it to kernel space.
1409 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1412 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1413 * other SEQPACKET protocols that take time to connect() as it doesn't
1414 * include the -EINPROGRESS status for such sockets.
1417 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1419 struct socket
*sock
;
1420 char address
[MAX_SOCK_ADDR
];
1423 sock
= sockfd_lookup(fd
, &err
);
1426 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1430 err
= security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1434 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1435 sock
->file
->f_flags
);
1443 * Get the local address ('name') of a socket object. Move the obtained
1444 * name to user space.
1447 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1449 struct socket
*sock
;
1450 char address
[MAX_SOCK_ADDR
];
1453 sock
= sockfd_lookup(fd
, &err
);
1457 err
= security_socket_getsockname(sock
);
1461 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1464 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1473 * Get the remote address ('name') of a socket object. Move the obtained
1474 * name to user space.
1477 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1479 struct socket
*sock
;
1480 char address
[MAX_SOCK_ADDR
];
1483 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1485 err
= security_socket_getpeername(sock
);
1491 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1493 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1500 * Send a datagram to a given address. We move the address into kernel
1501 * space and check the user space data area is readable before invoking
1505 asmlinkage
long sys_sendto(int fd
, void __user
* buff
, size_t len
, unsigned flags
,
1506 struct sockaddr __user
*addr
, int addr_len
)
1508 struct socket
*sock
;
1509 char address
[MAX_SOCK_ADDR
];
1514 sock
= sockfd_lookup(fd
, &err
);
1522 msg
.msg_control
=NULL
;
1523 msg
.msg_controllen
=0;
1527 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1530 msg
.msg_name
=address
;
1531 msg
.msg_namelen
=addr_len
;
1533 if (sock
->file
->f_flags
& O_NONBLOCK
)
1534 flags
|= MSG_DONTWAIT
;
1535 msg
.msg_flags
= flags
;
1536 err
= sock_sendmsg(sock
, &msg
, len
);
1545 * Send a datagram down a socket.
1548 asmlinkage
long sys_send(int fd
, void __user
* buff
, size_t len
, unsigned flags
)
1550 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1554 * Receive a frame from the socket and optionally record the address of the
1555 * sender. We verify the buffers are writable and if needed move the
1556 * sender address from kernel to user space.
1559 asmlinkage
long sys_recvfrom(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
,
1560 struct sockaddr __user
*addr
, int __user
*addr_len
)
1562 struct socket
*sock
;
1565 char address
[MAX_SOCK_ADDR
];
1568 sock
= sockfd_lookup(fd
, &err
);
1572 msg
.msg_control
=NULL
;
1573 msg
.msg_controllen
=0;
1578 msg
.msg_name
=address
;
1579 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1580 if (sock
->file
->f_flags
& O_NONBLOCK
)
1581 flags
|= MSG_DONTWAIT
;
1582 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1584 if(err
>= 0 && addr
!= NULL
)
1586 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1596 * Receive a datagram from a socket.
1599 asmlinkage
long sys_recv(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
)
1601 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1605 * Set a socket option. Because we don't know the option lengths we have
1606 * to pass the user mode parameter for the protocols to sort out.
1609 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char __user
*optval
, int optlen
)
1612 struct socket
*sock
;
1617 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1619 err
= security_socket_setsockopt(sock
,level
,optname
);
1625 if (level
== SOL_SOCKET
)
1626 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1628 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1635 * Get a socket option. Because we don't know the option lengths we have
1636 * to pass a user mode parameter for the protocols to sort out.
1639 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
, int __user
*optlen
)
1642 struct socket
*sock
;
1644 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1646 err
= security_socket_getsockopt(sock
, level
,
1653 if (level
== SOL_SOCKET
)
1654 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1656 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1664 * Shutdown a socket.
1667 asmlinkage
long sys_shutdown(int fd
, int how
)
1670 struct socket
*sock
;
1672 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1674 err
= security_socket_shutdown(sock
, how
);
1680 err
=sock
->ops
->shutdown(sock
, how
);
1686 /* A couple of helpful macros for getting the address of the 32/64 bit
1687 * fields which are the same type (int / unsigned) on our platforms.
1689 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1690 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1691 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1695 * BSD sendmsg interface
1698 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1700 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1701 struct socket
*sock
;
1702 char address
[MAX_SOCK_ADDR
];
1703 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1704 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]; /* 20 is size of ipv6_pktinfo */
1705 unsigned char *ctl_buf
= ctl
;
1706 struct msghdr msg_sys
;
1707 int err
, ctl_len
, iov_size
, total_len
;
1710 if (MSG_CMSG_COMPAT
& flags
) {
1711 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1713 } else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1716 sock
= sockfd_lookup(fd
, &err
);
1720 /* do not move before msg_sys is valid */
1722 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1725 /* Check whether to allocate the iovec area*/
1727 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1728 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1729 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1734 /* This will also move the address data into kernel space */
1735 if (MSG_CMSG_COMPAT
& flags
) {
1736 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1738 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1745 if (msg_sys
.msg_controllen
> INT_MAX
)
1747 ctl_len
= msg_sys
.msg_controllen
;
1748 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1749 err
= cmsghdr_from_user_compat_to_kern(&msg_sys
, ctl
, sizeof(ctl
));
1752 ctl_buf
= msg_sys
.msg_control
;
1753 } else if (ctl_len
) {
1754 if (ctl_len
> sizeof(ctl
))
1756 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1757 if (ctl_buf
== NULL
)
1762 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1763 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1764 * checking falls down on this.
1766 if (copy_from_user(ctl_buf
, (void __user
*) msg_sys
.msg_control
, ctl_len
))
1768 msg_sys
.msg_control
= ctl_buf
;
1770 msg_sys
.msg_flags
= flags
;
1772 if (sock
->file
->f_flags
& O_NONBLOCK
)
1773 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1774 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1778 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1780 if (iov
!= iovstack
)
1781 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1789 * BSD recvmsg interface
1792 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned int flags
)
1794 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1795 struct socket
*sock
;
1796 struct iovec iovstack
[UIO_FASTIOV
];
1797 struct iovec
*iov
=iovstack
;
1798 struct msghdr msg_sys
;
1799 unsigned long cmsg_ptr
;
1800 int err
, iov_size
, total_len
, len
;
1802 /* kernel mode address */
1803 char addr
[MAX_SOCK_ADDR
];
1805 /* user mode address pointers */
1806 struct sockaddr __user
*uaddr
;
1807 int __user
*uaddr_len
;
1809 if (MSG_CMSG_COMPAT
& flags
) {
1810 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1813 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1816 sock
= sockfd_lookup(fd
, &err
);
1821 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1824 /* Check whether to allocate the iovec area*/
1826 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1827 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1828 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1834 * Save the user-mode address (verify_iovec will change the
1835 * kernel msghdr to use the kernel address space)
1838 uaddr
= (void __user
*) msg_sys
.msg_name
;
1839 uaddr_len
= COMPAT_NAMELEN(msg
);
1840 if (MSG_CMSG_COMPAT
& flags
) {
1841 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1843 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1848 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1849 msg_sys
.msg_flags
= 0;
1850 if (MSG_CMSG_COMPAT
& flags
)
1851 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1853 if (sock
->file
->f_flags
& O_NONBLOCK
)
1854 flags
|= MSG_DONTWAIT
;
1855 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1860 if (uaddr
!= NULL
) {
1861 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1865 err
= __put_user(msg_sys
.msg_flags
, COMPAT_FLAGS(msg
));
1868 if (MSG_CMSG_COMPAT
& flags
)
1869 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1870 &msg_compat
->msg_controllen
);
1872 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1873 &msg
->msg_controllen
);
1879 if (iov
!= iovstack
)
1880 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1887 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1889 /* Argument list sizes for sys_socketcall */
1890 #define AL(x) ((x) * sizeof(unsigned long))
1891 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1892 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1893 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1897 * System call vectors.
1899 * Argument checking cleaned up. Saved 20% in size.
1900 * This function doesn't need to set the kernel lock because
1901 * it is set by the callees.
1904 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1907 unsigned long a0
,a1
;
1910 if(call
<1||call
>SYS_RECVMSG
)
1913 /* copy_from_user should be SMP safe. */
1914 if (copy_from_user(a
, args
, nargs
[call
]))
1923 err
= sys_socket(a0
,a1
,a
[2]);
1926 err
= sys_bind(a0
,(struct sockaddr __user
*)a1
, a
[2]);
1929 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
1932 err
= sys_listen(a0
,a1
);
1935 err
= sys_accept(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1937 case SYS_GETSOCKNAME
:
1938 err
= sys_getsockname(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1940 case SYS_GETPEERNAME
:
1941 err
= sys_getpeername(a0
, (struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1943 case SYS_SOCKETPAIR
:
1944 err
= sys_socketpair(a0
,a1
, a
[2], (int __user
*)a
[3]);
1947 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
1950 err
= sys_sendto(a0
,(void __user
*)a1
, a
[2], a
[3],
1951 (struct sockaddr __user
*)a
[4], a
[5]);
1954 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
1957 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
1958 (struct sockaddr __user
*)a
[4], (int __user
*)a
[5]);
1961 err
= sys_shutdown(a0
,a1
);
1963 case SYS_SETSOCKOPT
:
1964 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
1966 case SYS_GETSOCKOPT
:
1967 err
= sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], (int __user
*)a
[4]);
1970 err
= sys_sendmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1973 err
= sys_recvmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1982 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
1985 * This function is called by a protocol handler that wants to
1986 * advertise its address family, and have it linked into the
1990 int sock_register(struct net_proto_family
*ops
)
1994 if (ops
->family
>= NPROTO
) {
1995 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
1998 net_family_write_lock();
2000 if (net_families
[ops
->family
] == NULL
) {
2001 net_families
[ops
->family
]=ops
;
2004 net_family_write_unlock();
2005 printk(KERN_INFO
"NET: Registered protocol family %d\n",
2011 * This function is called by a protocol handler that wants to
2012 * remove its address family, and have it unlinked from the
2016 int sock_unregister(int family
)
2018 if (family
< 0 || family
>= NPROTO
)
2021 net_family_write_lock();
2022 net_families
[family
]=NULL
;
2023 net_family_write_unlock();
2024 printk(KERN_INFO
"NET: Unregistered protocol family %d\n",
2030 extern void sk_init(void);
2032 void __init
sock_init(void)
2037 * Initialize all address (protocol) families.
2040 for (i
= 0; i
< NPROTO
; i
++)
2041 net_families
[i
] = NULL
;
2044 * Initialize sock SLAB cache.
2051 * Initialize skbuff SLAB cache
2057 * Initialize the protocols module.
2061 register_filesystem(&sock_fs_type
);
2062 sock_mnt
= kern_mount(&sock_fs_type
);
2063 /* The real protocol initialization is performed when
2064 * do_initcalls is run.
2067 #ifdef CONFIG_NETFILTER
2072 #ifdef CONFIG_PROC_FS
2073 void socket_seq_show(struct seq_file
*seq
)
2078 for (cpu
= 0; cpu
< NR_CPUS
; cpu
++)
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_alloc
);
2093 EXPORT_SYMBOL(sock_alloc_inode
);
2094 EXPORT_SYMBOL(sock_create
);
2095 EXPORT_SYMBOL(sock_create_kern
);
2096 EXPORT_SYMBOL(sock_create_lite
);
2097 EXPORT_SYMBOL(sock_map_fd
);
2098 EXPORT_SYMBOL(sock_recvmsg
);
2099 EXPORT_SYMBOL(sock_register
);
2100 EXPORT_SYMBOL(sock_release
);
2101 EXPORT_SYMBOL(sock_sendmsg
);
2102 EXPORT_SYMBOL(sock_unregister
);
2103 EXPORT_SYMBOL(sock_wake_async
);
2104 EXPORT_SYMBOL(sockfd_lookup
);
2105 EXPORT_SYMBOL(kernel_sendmsg
);
2106 EXPORT_SYMBOL(kernel_recvmsg
);