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>
88 #ifdef CONFIG_NET_RADIO
89 #include <linux/wireless.h> /* Note : will define WIRELESS_EXT */
90 #endif /* CONFIG_NET_RADIO */
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
98 #include <linux/netfilter.h>
100 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
101 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*buf
,
102 size_t size
, loff_t pos
);
103 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*buf
,
104 size_t size
, loff_t pos
);
105 static int sock_mmap(struct file
*file
, struct vm_area_struct
* vma
);
107 static int sock_close(struct inode
*inode
, struct file
*file
);
108 static unsigned int sock_poll(struct file
*file
,
109 struct poll_table_struct
*wait
);
110 static long sock_ioctl(struct file
*file
,
111 unsigned int cmd
, unsigned long arg
);
112 static int sock_fasync(int fd
, struct file
*filp
, int on
);
113 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
114 unsigned long count
, loff_t
*ppos
);
115 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
116 unsigned long count
, loff_t
*ppos
);
117 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
118 int offset
, size_t size
, loff_t
*ppos
, int more
);
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
126 static struct file_operations socket_file_ops
= {
127 .owner
= THIS_MODULE
,
129 .aio_read
= sock_aio_read
,
130 .aio_write
= sock_aio_write
,
132 .unlocked_ioctl
= sock_ioctl
,
134 .open
= sock_no_open
, /* special open code to disallow open via /proc */
135 .release
= sock_close
,
136 .fasync
= sock_fasync
,
138 .writev
= sock_writev
,
139 .sendpage
= sock_sendpage
143 * The protocol list. Each protocol is registered in here.
146 static struct net_proto_family
*net_families
[NPROTO
];
148 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
149 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
150 static DEFINE_SPINLOCK(net_family_lock
);
152 /* The strategy is: modifications net_family vector are short, do not
153 sleep and veeery rare, but read access should be free of any exclusive
157 static void net_family_write_lock(void)
159 spin_lock(&net_family_lock
);
160 while (atomic_read(&net_family_lockct
) != 0) {
161 spin_unlock(&net_family_lock
);
165 spin_lock(&net_family_lock
);
169 static __inline__
void net_family_write_unlock(void)
171 spin_unlock(&net_family_lock
);
174 static __inline__
void net_family_read_lock(void)
176 atomic_inc(&net_family_lockct
);
177 spin_unlock_wait(&net_family_lock
);
180 static __inline__
void net_family_read_unlock(void)
182 atomic_dec(&net_family_lockct
);
186 #define net_family_write_lock() do { } while(0)
187 #define net_family_write_unlock() do { } while(0)
188 #define net_family_read_lock() do { } while(0)
189 #define net_family_read_unlock() do { } while(0)
194 * Statistics counters of the socket lists
197 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
200 * Support routines. Move socket addresses back and forth across the kernel/user
201 * divide and look after the messy bits.
204 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
205 16 for IP, 16 for IPX,
208 must be at least one bigger than
209 the AF_UNIX size (see net/unix/af_unix.c
214 * move_addr_to_kernel - copy a socket address into kernel space
215 * @uaddr: Address in user space
216 * @kaddr: Address in kernel space
217 * @ulen: Length in user space
219 * The address is copied into kernel space. If the provided address is
220 * too long an error code of -EINVAL is returned. If the copy gives
221 * invalid addresses -EFAULT is returned. On a success 0 is returned.
224 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
226 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
230 if(copy_from_user(kaddr
,uaddr
,ulen
))
232 return audit_sockaddr(ulen
, kaddr
);
236 * move_addr_to_user - copy an address to user space
237 * @kaddr: kernel space address
238 * @klen: length of address in kernel
239 * @uaddr: user space address
240 * @ulen: pointer to user length field
242 * The value pointed to by ulen on entry is the buffer length available.
243 * This is overwritten with the buffer space used. -EINVAL is returned
244 * if an overlong buffer is specified or a negative buffer size. -EFAULT
245 * is returned if either the buffer or the length field are not
247 * After copying the data up to the limit the user specifies, the true
248 * length of the data is written over the length limit the user
249 * specified. Zero is returned for a success.
252 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
, int __user
*ulen
)
257 if((err
=get_user(len
, ulen
)))
261 if(len
<0 || len
> MAX_SOCK_ADDR
)
265 if(copy_to_user(uaddr
,kaddr
,len
))
269 * "fromlen shall refer to the value before truncation.."
272 return __put_user(klen
, ulen
);
275 #define SOCKFS_MAGIC 0x534F434B
277 static kmem_cache_t
* sock_inode_cachep __read_mostly
;
279 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
281 struct socket_alloc
*ei
;
282 ei
= (struct socket_alloc
*)kmem_cache_alloc(sock_inode_cachep
, SLAB_KERNEL
);
285 init_waitqueue_head(&ei
->socket
.wait
);
287 ei
->socket
.fasync_list
= NULL
;
288 ei
->socket
.state
= SS_UNCONNECTED
;
289 ei
->socket
.flags
= 0;
290 ei
->socket
.ops
= NULL
;
291 ei
->socket
.sk
= NULL
;
292 ei
->socket
.file
= NULL
;
293 ei
->socket
.flags
= 0;
295 return &ei
->vfs_inode
;
298 static void sock_destroy_inode(struct inode
*inode
)
300 kmem_cache_free(sock_inode_cachep
,
301 container_of(inode
, struct socket_alloc
, vfs_inode
));
304 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
306 struct socket_alloc
*ei
= (struct socket_alloc
*) foo
;
308 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
309 SLAB_CTOR_CONSTRUCTOR
)
310 inode_init_once(&ei
->vfs_inode
);
313 static int init_inodecache(void)
315 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
316 sizeof(struct socket_alloc
),
317 0, SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
,
319 if (sock_inode_cachep
== NULL
)
324 static struct super_operations sockfs_ops
= {
325 .alloc_inode
= sock_alloc_inode
,
326 .destroy_inode
=sock_destroy_inode
,
327 .statfs
= simple_statfs
,
330 static struct super_block
*sockfs_get_sb(struct file_system_type
*fs_type
,
331 int flags
, const char *dev_name
, void *data
)
333 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
);
336 static struct vfsmount
*sock_mnt __read_mostly
;
338 static struct file_system_type sock_fs_type
= {
340 .get_sb
= sockfs_get_sb
,
341 .kill_sb
= kill_anon_super
,
343 static int sockfs_delete_dentry(struct dentry
*dentry
)
347 static struct dentry_operations sockfs_dentry_operations
= {
348 .d_delete
= sockfs_delete_dentry
,
352 * Obtains the first available file descriptor and sets it up for use.
354 * This function creates file structure and maps it to fd space
355 * of current process. On success it returns file descriptor
356 * and file struct implicitly stored in sock->file.
357 * Note that another thread may close file descriptor before we return
358 * from this function. We use the fact that now we do not refer
359 * to socket after mapping. If one day we will need it, this
360 * function will increment ref. count on file by 1.
362 * In any case returned fd MAY BE not valid!
363 * This race condition is unavoidable
364 * with shared fd spaces, we cannot solve it inside kernel,
365 * but we take care of internal coherence yet.
368 int sock_map_fd(struct socket
*sock
)
375 * Find a file descriptor suitable for return to the user.
378 fd
= get_unused_fd();
380 struct file
*file
= get_empty_filp();
388 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
390 this.hash
= SOCK_INODE(sock
)->i_ino
;
392 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
393 if (!file
->f_dentry
) {
399 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
400 d_add(file
->f_dentry
, SOCK_INODE(sock
));
401 file
->f_vfsmnt
= mntget(sock_mnt
);
402 file
->f_mapping
= file
->f_dentry
->d_inode
->i_mapping
;
405 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
406 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
407 file
->f_flags
= O_RDWR
;
409 file
->private_data
= sock
;
410 fd_install(fd
, file
);
418 * sockfd_lookup - Go from a file number to its socket slot
420 * @err: pointer to an error code return
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
427 * On a success the socket object pointer is returned.
430 struct socket
*sockfd_lookup(int fd
, int *err
)
436 if (!(file
= fget(fd
)))
442 if (file
->f_op
== &socket_file_ops
)
443 return file
->private_data
; /* set in sock_map_fd */
445 inode
= file
->f_dentry
->d_inode
;
446 if (!S_ISSOCK(inode
->i_mode
)) {
452 sock
= SOCKET_I(inode
);
453 if (sock
->file
!= file
) {
454 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
461 * sock_alloc - allocate a socket
463 * Allocate a new inode and socket object. The two are bound together
464 * and initialised. The socket is then returned. If we are out of inodes
468 static struct socket
*sock_alloc(void)
470 struct inode
* inode
;
471 struct socket
* sock
;
473 inode
= new_inode(sock_mnt
->mnt_sb
);
477 sock
= SOCKET_I(inode
);
479 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
480 inode
->i_uid
= current
->fsuid
;
481 inode
->i_gid
= current
->fsgid
;
483 get_cpu_var(sockets_in_use
)++;
484 put_cpu_var(sockets_in_use
);
489 * In theory you can't get an open on this inode, but /proc provides
490 * a back door. Remember to keep it shut otherwise you'll let the
491 * creepy crawlies in.
494 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
499 struct file_operations bad_sock_fops
= {
500 .owner
= THIS_MODULE
,
501 .open
= sock_no_open
,
505 * sock_release - close a socket
506 * @sock: socket to close
508 * The socket is released from the protocol stack if it has a release
509 * callback, and the inode is then released if the socket is bound to
510 * an inode not a file.
513 void sock_release(struct socket
*sock
)
516 struct module
*owner
= sock
->ops
->owner
;
518 sock
->ops
->release(sock
);
523 if (sock
->fasync_list
)
524 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
526 get_cpu_var(sockets_in_use
)--;
527 put_cpu_var(sockets_in_use
);
529 iput(SOCK_INODE(sock
));
535 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
536 struct msghdr
*msg
, size_t size
)
538 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
546 err
= security_socket_sendmsg(sock
, msg
, size
);
550 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
553 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
556 struct sock_iocb siocb
;
559 init_sync_kiocb(&iocb
, NULL
);
560 iocb
.private = &siocb
;
561 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
562 if (-EIOCBQUEUED
== ret
)
563 ret
= wait_on_sync_kiocb(&iocb
);
567 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
568 struct kvec
*vec
, size_t num
, size_t size
)
570 mm_segment_t oldfs
= get_fs();
575 * the following is safe, since for compiler definitions of kvec and
576 * iovec are identical, yielding the same in-core layout and alignment
578 msg
->msg_iov
= (struct iovec
*)vec
,
579 msg
->msg_iovlen
= num
;
580 result
= sock_sendmsg(sock
, msg
, size
);
585 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
586 struct msghdr
*msg
, size_t size
, int flags
)
589 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
597 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
601 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
604 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
605 size_t size
, int flags
)
608 struct sock_iocb siocb
;
611 init_sync_kiocb(&iocb
, NULL
);
612 iocb
.private = &siocb
;
613 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
614 if (-EIOCBQUEUED
== ret
)
615 ret
= wait_on_sync_kiocb(&iocb
);
619 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
620 struct kvec
*vec
, size_t num
,
621 size_t size
, int flags
)
623 mm_segment_t oldfs
= get_fs();
628 * the following is safe, since for compiler definitions of kvec and
629 * iovec are identical, yielding the same in-core layout and alignment
631 msg
->msg_iov
= (struct iovec
*)vec
,
632 msg
->msg_iovlen
= num
;
633 result
= sock_recvmsg(sock
, msg
, size
, flags
);
638 static void sock_aio_dtor(struct kiocb
*iocb
)
640 kfree(iocb
->private);
643 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
644 int offset
, size_t size
, loff_t
*ppos
, int more
)
649 sock
= file
->private_data
;
651 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
655 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
658 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
659 char __user
*ubuf
, size_t size
, struct sock_iocb
*siocb
)
661 if (!is_sync_kiocb(iocb
)) {
662 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
665 iocb
->ki_dtor
= sock_aio_dtor
;
669 siocb
->async_iov
.iov_base
= ubuf
;
670 siocb
->async_iov
.iov_len
= size
;
672 iocb
->private = siocb
;
676 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
677 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
679 struct socket
*sock
= file
->private_data
;
683 for (i
= 0 ; i
< nr_segs
; i
++)
684 size
+= iov
[i
].iov_len
;
686 msg
->msg_name
= NULL
;
687 msg
->msg_namelen
= 0;
688 msg
->msg_control
= NULL
;
689 msg
->msg_controllen
= 0;
690 msg
->msg_iov
= (struct iovec
*) iov
;
691 msg
->msg_iovlen
= nr_segs
;
692 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
694 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
697 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
698 unsigned long nr_segs
, loff_t
*ppos
)
701 struct sock_iocb siocb
;
705 init_sync_kiocb(&iocb
, NULL
);
706 iocb
.private = &siocb
;
708 ret
= do_sock_read(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
709 if (-EIOCBQUEUED
== ret
)
710 ret
= wait_on_sync_kiocb(&iocb
);
714 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
715 size_t count
, loff_t pos
)
717 struct sock_iocb siocb
, *x
;
721 if (count
== 0) /* Match SYS5 behaviour */
724 x
= alloc_sock_iocb(iocb
, ubuf
, count
, &siocb
);
727 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
,
731 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
732 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
734 struct socket
*sock
= file
->private_data
;
738 for (i
= 0 ; i
< nr_segs
; i
++)
739 size
+= iov
[i
].iov_len
;
741 msg
->msg_name
= NULL
;
742 msg
->msg_namelen
= 0;
743 msg
->msg_control
= NULL
;
744 msg
->msg_controllen
= 0;
745 msg
->msg_iov
= (struct iovec
*) iov
;
746 msg
->msg_iovlen
= nr_segs
;
747 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
748 if (sock
->type
== SOCK_SEQPACKET
)
749 msg
->msg_flags
|= MSG_EOR
;
751 return __sock_sendmsg(iocb
, sock
, msg
, size
);
754 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
755 unsigned long nr_segs
, loff_t
*ppos
)
759 struct sock_iocb siocb
;
762 init_sync_kiocb(&iocb
, NULL
);
763 iocb
.private = &siocb
;
765 ret
= do_sock_write(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
766 if (-EIOCBQUEUED
== ret
)
767 ret
= wait_on_sync_kiocb(&iocb
);
771 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
772 size_t count
, loff_t pos
)
774 struct sock_iocb siocb
, *x
;
778 if (count
== 0) /* Match SYS5 behaviour */
781 x
= alloc_sock_iocb(iocb
, (void __user
*)ubuf
, count
, &siocb
);
785 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
,
791 * Atomic setting of ioctl hooks to avoid race
792 * with module unload.
795 static DECLARE_MUTEX(br_ioctl_mutex
);
796 static int (*br_ioctl_hook
)(unsigned int cmd
, void __user
*arg
) = NULL
;
798 void brioctl_set(int (*hook
)(unsigned int, void __user
*))
800 down(&br_ioctl_mutex
);
801 br_ioctl_hook
= hook
;
804 EXPORT_SYMBOL(brioctl_set
);
806 static DECLARE_MUTEX(vlan_ioctl_mutex
);
807 static int (*vlan_ioctl_hook
)(void __user
*arg
);
809 void vlan_ioctl_set(int (*hook
)(void __user
*))
811 down(&vlan_ioctl_mutex
);
812 vlan_ioctl_hook
= hook
;
813 up(&vlan_ioctl_mutex
);
815 EXPORT_SYMBOL(vlan_ioctl_set
);
817 static DECLARE_MUTEX(dlci_ioctl_mutex
);
818 static int (*dlci_ioctl_hook
)(unsigned int, void __user
*);
820 void dlci_ioctl_set(int (*hook
)(unsigned int, void __user
*))
822 down(&dlci_ioctl_mutex
);
823 dlci_ioctl_hook
= hook
;
824 up(&dlci_ioctl_mutex
);
826 EXPORT_SYMBOL(dlci_ioctl_set
);
829 * With an ioctl, arg may well be a user mode pointer, but we don't know
830 * what to do with it - that's up to the protocol still.
833 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
836 void __user
*argp
= (void __user
*)arg
;
839 sock
= file
->private_data
;
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
);
908 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
911 struct socket
*sock
= NULL
;
913 err
= security_socket_create(family
, type
, protocol
, 1);
923 security_socket_post_create(sock
, family
, type
, protocol
, 1);
930 /* No kernel lock held - perfect */
931 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
936 * We can't return errors to poll, so it's either yes or no.
938 sock
= file
->private_data
;
939 return sock
->ops
->poll(file
, sock
, wait
);
942 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
944 struct socket
*sock
= file
->private_data
;
946 return sock
->ops
->mmap(file
, sock
, vma
);
949 static int sock_close(struct inode
*inode
, struct file
*filp
)
952 * It was possible the inode is NULL we were
953 * closing an unfinished socket.
958 printk(KERN_DEBUG
"sock_close: NULL inode\n");
961 sock_fasync(-1, filp
, 0);
962 sock_release(SOCKET_I(inode
));
967 * Update the socket async list
969 * Fasync_list locking strategy.
971 * 1. fasync_list is modified only under process context socket lock
972 * i.e. under semaphore.
973 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
974 * or under socket lock.
975 * 3. fasync_list can be used from softirq context, so that
976 * modification under socket lock have to be enhanced with
977 * write_lock_bh(&sk->sk_callback_lock).
981 static int sock_fasync(int fd
, struct file
*filp
, int on
)
983 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
989 fna
=(struct fasync_struct
*)kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
994 sock
= filp
->private_data
;
996 if ((sk
=sock
->sk
) == NULL
) {
1003 prev
=&(sock
->fasync_list
);
1005 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
1006 if (fa
->fa_file
==filp
)
1013 write_lock_bh(&sk
->sk_callback_lock
);
1015 write_unlock_bh(&sk
->sk_callback_lock
);
1022 fna
->magic
=FASYNC_MAGIC
;
1023 fna
->fa_next
=sock
->fasync_list
;
1024 write_lock_bh(&sk
->sk_callback_lock
);
1025 sock
->fasync_list
=fna
;
1026 write_unlock_bh(&sk
->sk_callback_lock
);
1032 write_lock_bh(&sk
->sk_callback_lock
);
1034 write_unlock_bh(&sk
->sk_callback_lock
);
1040 release_sock(sock
->sk
);
1044 /* This function may be called only under socket lock or callback_lock */
1046 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1048 if (!sock
|| !sock
->fasync_list
)
1054 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1058 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1063 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1066 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1071 static int __sock_create(int family
, int type
, int protocol
, struct socket
**res
, int kern
)
1074 struct socket
*sock
;
1077 * Check protocol is in range
1079 if (family
< 0 || family
>= NPROTO
)
1080 return -EAFNOSUPPORT
;
1081 if (type
< 0 || type
>= SOCK_MAX
)
1086 This uglymoron is moved from INET layer to here to avoid
1087 deadlock in module load.
1089 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1093 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
1098 err
= security_socket_create(family
, type
, protocol
, kern
);
1102 #if defined(CONFIG_KMOD)
1103 /* Attempt to load a protocol module if the find failed.
1105 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1106 * requested real, full-featured networking support upon configuration.
1107 * Otherwise module support will break!
1109 if (net_families
[family
]==NULL
)
1111 request_module("net-pf-%d",family
);
1115 net_family_read_lock();
1116 if (net_families
[family
] == NULL
) {
1117 err
= -EAFNOSUPPORT
;
1122 * Allocate the socket and allow the family to set things up. if
1123 * the protocol is 0, the family is instructed to select an appropriate
1127 if (!(sock
= sock_alloc())) {
1128 printk(KERN_WARNING
"socket: no more sockets\n");
1129 err
= -ENFILE
; /* Not exactly a match, but its the
1130 closest posix thing */
1137 * We will call the ->create function, that possibly is in a loadable
1138 * module, so we have to bump that loadable module refcnt first.
1140 err
= -EAFNOSUPPORT
;
1141 if (!try_module_get(net_families
[family
]->owner
))
1144 if ((err
= net_families
[family
]->create(sock
, protocol
)) < 0) {
1146 goto out_module_put
;
1150 * Now to bump the refcnt of the [loadable] module that owns this
1151 * socket at sock_release time we decrement its refcnt.
1153 if (!try_module_get(sock
->ops
->owner
)) {
1155 goto out_module_put
;
1158 * Now that we're done with the ->create function, the [loadable]
1159 * module can have its refcnt decremented
1161 module_put(net_families
[family
]->owner
);
1163 security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1166 net_family_read_unlock();
1169 module_put(net_families
[family
]->owner
);
1175 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1177 return __sock_create(family
, type
, protocol
, res
, 0);
1180 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1182 return __sock_create(family
, type
, protocol
, res
, 1);
1185 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1188 struct socket
*sock
;
1190 retval
= sock_create(family
, type
, protocol
, &sock
);
1194 retval
= sock_map_fd(sock
);
1199 /* It may be already another descriptor 8) Not kernel problem. */
1208 * Create a pair of connected sockets.
1211 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1213 struct socket
*sock1
, *sock2
;
1217 * Obtain the first socket and check if the underlying protocol
1218 * supports the socketpair call.
1221 err
= sock_create(family
, type
, protocol
, &sock1
);
1225 err
= sock_create(family
, type
, protocol
, &sock2
);
1229 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1231 goto out_release_both
;
1235 err
= sock_map_fd(sock1
);
1237 goto out_release_both
;
1240 err
= sock_map_fd(sock2
);
1245 /* fd1 and fd2 may be already another descriptors.
1246 * Not kernel problem.
1249 err
= put_user(fd1
, &usockvec
[0]);
1251 err
= put_user(fd2
, &usockvec
[1]);
1260 sock_release(sock2
);
1265 sock_release(sock2
);
1267 sock_release(sock1
);
1274 * Bind a name to a socket. Nothing much to do here since it's
1275 * the protocol's responsibility to handle the local address.
1277 * We move the socket address to kernel space before we call
1278 * the protocol layer (having also checked the address is ok).
1281 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1283 struct socket
*sock
;
1284 char address
[MAX_SOCK_ADDR
];
1287 if((sock
= sockfd_lookup(fd
,&err
))!=NULL
)
1289 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0) {
1290 err
= security_socket_bind(sock
, (struct sockaddr
*)address
, addrlen
);
1295 err
= sock
->ops
->bind(sock
, (struct sockaddr
*)address
, addrlen
);
1304 * Perform a listen. Basically, we allow the protocol to do anything
1305 * necessary for a listen, and if that works, we mark the socket as
1306 * ready for listening.
1309 int sysctl_somaxconn
= SOMAXCONN
;
1311 asmlinkage
long sys_listen(int fd
, int backlog
)
1313 struct socket
*sock
;
1316 if ((sock
= sockfd_lookup(fd
, &err
)) != NULL
) {
1317 if ((unsigned) backlog
> sysctl_somaxconn
)
1318 backlog
= sysctl_somaxconn
;
1320 err
= security_socket_listen(sock
, backlog
);
1326 err
=sock
->ops
->listen(sock
, backlog
);
1334 * For accept, we attempt to create a new socket, set up the link
1335 * with the client, wake up the client, then return the new
1336 * connected fd. We collect the address of the connector in kernel
1337 * space and move it to user at the very end. This is unclean because
1338 * we open the socket then return an error.
1340 * 1003.1g adds the ability to recvmsg() to query connection pending
1341 * status to recvmsg. We need to add that support in a way thats
1342 * clean when we restucture accept also.
1345 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
, int __user
*upeer_addrlen
)
1347 struct socket
*sock
, *newsock
;
1349 char address
[MAX_SOCK_ADDR
];
1351 sock
= sockfd_lookup(fd
, &err
);
1356 if (!(newsock
= sock_alloc()))
1359 newsock
->type
= sock
->type
;
1360 newsock
->ops
= sock
->ops
;
1363 * We don't need try_module_get here, as the listening socket (sock)
1364 * has the protocol module (sock->ops->owner) held.
1366 __module_get(newsock
->ops
->owner
);
1368 err
= security_socket_accept(sock
, newsock
);
1372 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1376 if (upeer_sockaddr
) {
1377 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1378 err
= -ECONNABORTED
;
1381 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1386 /* File flags are not inherited via accept() unlike another OSes. */
1388 if ((err
= sock_map_fd(newsock
)) < 0)
1391 security_socket_post_accept(sock
, newsock
);
1398 sock_release(newsock
);
1404 * Attempt to connect to a socket with the server address. The address
1405 * is in user space so we verify it is OK and move it to kernel space.
1407 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1410 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1411 * other SEQPACKET protocols that take time to connect() as it doesn't
1412 * include the -EINPROGRESS status for such sockets.
1415 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1417 struct socket
*sock
;
1418 char address
[MAX_SOCK_ADDR
];
1421 sock
= sockfd_lookup(fd
, &err
);
1424 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1428 err
= security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1432 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1433 sock
->file
->f_flags
);
1441 * Get the local address ('name') of a socket object. Move the obtained
1442 * name to user space.
1445 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1447 struct socket
*sock
;
1448 char address
[MAX_SOCK_ADDR
];
1451 sock
= sockfd_lookup(fd
, &err
);
1455 err
= security_socket_getsockname(sock
);
1459 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1462 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1471 * Get the remote address ('name') of a socket object. Move the obtained
1472 * name to user space.
1475 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1477 struct socket
*sock
;
1478 char address
[MAX_SOCK_ADDR
];
1481 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1483 err
= security_socket_getpeername(sock
);
1489 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1491 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1498 * Send a datagram to a given address. We move the address into kernel
1499 * space and check the user space data area is readable before invoking
1503 asmlinkage
long sys_sendto(int fd
, void __user
* buff
, size_t len
, unsigned flags
,
1504 struct sockaddr __user
*addr
, int addr_len
)
1506 struct socket
*sock
;
1507 char address
[MAX_SOCK_ADDR
];
1512 sock
= sockfd_lookup(fd
, &err
);
1520 msg
.msg_control
=NULL
;
1521 msg
.msg_controllen
=0;
1525 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1528 msg
.msg_name
=address
;
1529 msg
.msg_namelen
=addr_len
;
1531 if (sock
->file
->f_flags
& O_NONBLOCK
)
1532 flags
|= MSG_DONTWAIT
;
1533 msg
.msg_flags
= flags
;
1534 err
= sock_sendmsg(sock
, &msg
, len
);
1543 * Send a datagram down a socket.
1546 asmlinkage
long sys_send(int fd
, void __user
* buff
, size_t len
, unsigned flags
)
1548 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1552 * Receive a frame from the socket and optionally record the address of the
1553 * sender. We verify the buffers are writable and if needed move the
1554 * sender address from kernel to user space.
1557 asmlinkage
long sys_recvfrom(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
,
1558 struct sockaddr __user
*addr
, int __user
*addr_len
)
1560 struct socket
*sock
;
1563 char address
[MAX_SOCK_ADDR
];
1566 sock
= sockfd_lookup(fd
, &err
);
1570 msg
.msg_control
=NULL
;
1571 msg
.msg_controllen
=0;
1576 msg
.msg_name
=address
;
1577 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1578 if (sock
->file
->f_flags
& O_NONBLOCK
)
1579 flags
|= MSG_DONTWAIT
;
1580 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1582 if(err
>= 0 && addr
!= NULL
)
1584 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1594 * Receive a datagram from a socket.
1597 asmlinkage
long sys_recv(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
)
1599 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1603 * Set a socket option. Because we don't know the option lengths we have
1604 * to pass the user mode parameter for the protocols to sort out.
1607 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char __user
*optval
, int optlen
)
1610 struct socket
*sock
;
1615 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1617 err
= security_socket_setsockopt(sock
,level
,optname
);
1623 if (level
== SOL_SOCKET
)
1624 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1626 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1633 * Get a socket option. Because we don't know the option lengths we have
1634 * to pass a user mode parameter for the protocols to sort out.
1637 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
, int __user
*optlen
)
1640 struct socket
*sock
;
1642 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1644 err
= security_socket_getsockopt(sock
, level
,
1651 if (level
== SOL_SOCKET
)
1652 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1654 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1662 * Shutdown a socket.
1665 asmlinkage
long sys_shutdown(int fd
, int how
)
1668 struct socket
*sock
;
1670 if ((sock
= sockfd_lookup(fd
, &err
))!=NULL
)
1672 err
= security_socket_shutdown(sock
, how
);
1678 err
=sock
->ops
->shutdown(sock
, how
);
1684 /* A couple of helpful macros for getting the address of the 32/64 bit
1685 * fields which are the same type (int / unsigned) on our platforms.
1687 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1688 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1689 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1693 * BSD sendmsg interface
1696 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1698 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1699 struct socket
*sock
;
1700 char address
[MAX_SOCK_ADDR
];
1701 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1702 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1703 __attribute__ ((aligned (sizeof(__kernel_size_t
))));
1704 /* 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
, sock
->sk
, ctl
, sizeof(ctl
));
1752 ctl_buf
= msg_sys
.msg_control
;
1753 ctl_len
= msg_sys
.msg_controllen
;
1754 } else if (ctl_len
) {
1755 if (ctl_len
> sizeof(ctl
))
1757 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1758 if (ctl_buf
== NULL
)
1763 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1764 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1765 * checking falls down on this.
1767 if (copy_from_user(ctl_buf
, (void __user
*) msg_sys
.msg_control
, ctl_len
))
1769 msg_sys
.msg_control
= ctl_buf
;
1771 msg_sys
.msg_flags
= flags
;
1773 if (sock
->file
->f_flags
& O_NONBLOCK
)
1774 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1775 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1779 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1781 if (iov
!= iovstack
)
1782 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1790 * BSD recvmsg interface
1793 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned int flags
)
1795 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1796 struct socket
*sock
;
1797 struct iovec iovstack
[UIO_FASTIOV
];
1798 struct iovec
*iov
=iovstack
;
1799 struct msghdr msg_sys
;
1800 unsigned long cmsg_ptr
;
1801 int err
, iov_size
, total_len
, len
;
1803 /* kernel mode address */
1804 char addr
[MAX_SOCK_ADDR
];
1806 /* user mode address pointers */
1807 struct sockaddr __user
*uaddr
;
1808 int __user
*uaddr_len
;
1810 if (MSG_CMSG_COMPAT
& flags
) {
1811 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1814 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1817 sock
= sockfd_lookup(fd
, &err
);
1822 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1825 /* Check whether to allocate the iovec area*/
1827 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1828 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1829 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1835 * Save the user-mode address (verify_iovec will change the
1836 * kernel msghdr to use the kernel address space)
1839 uaddr
= (void __user
*) msg_sys
.msg_name
;
1840 uaddr_len
= COMPAT_NAMELEN(msg
);
1841 if (MSG_CMSG_COMPAT
& flags
) {
1842 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1844 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1849 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1850 msg_sys
.msg_flags
= 0;
1851 if (MSG_CMSG_COMPAT
& flags
)
1852 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1854 if (sock
->file
->f_flags
& O_NONBLOCK
)
1855 flags
|= MSG_DONTWAIT
;
1856 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1861 if (uaddr
!= NULL
) {
1862 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1866 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1870 if (MSG_CMSG_COMPAT
& flags
)
1871 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1872 &msg_compat
->msg_controllen
);
1874 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1875 &msg
->msg_controllen
);
1881 if (iov
!= iovstack
)
1882 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1889 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1891 /* Argument list sizes for sys_socketcall */
1892 #define AL(x) ((x) * sizeof(unsigned long))
1893 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1894 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1895 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1899 * System call vectors.
1901 * Argument checking cleaned up. Saved 20% in size.
1902 * This function doesn't need to set the kernel lock because
1903 * it is set by the callees.
1906 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1909 unsigned long a0
,a1
;
1912 if(call
<1||call
>SYS_RECVMSG
)
1915 /* copy_from_user should be SMP safe. */
1916 if (copy_from_user(a
, args
, nargs
[call
]))
1919 err
= audit_socketcall(nargs
[call
]/sizeof(unsigned long), a
);
1929 err
= sys_socket(a0
,a1
,a
[2]);
1932 err
= sys_bind(a0
,(struct sockaddr __user
*)a1
, a
[2]);
1935 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
1938 err
= sys_listen(a0
,a1
);
1941 err
= sys_accept(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1943 case SYS_GETSOCKNAME
:
1944 err
= sys_getsockname(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1946 case SYS_GETPEERNAME
:
1947 err
= sys_getpeername(a0
, (struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1949 case SYS_SOCKETPAIR
:
1950 err
= sys_socketpair(a0
,a1
, a
[2], (int __user
*)a
[3]);
1953 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
1956 err
= sys_sendto(a0
,(void __user
*)a1
, a
[2], a
[3],
1957 (struct sockaddr __user
*)a
[4], a
[5]);
1960 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
1963 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
1964 (struct sockaddr __user
*)a
[4], (int __user
*)a
[5]);
1967 err
= sys_shutdown(a0
,a1
);
1969 case SYS_SETSOCKOPT
:
1970 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
1972 case SYS_GETSOCKOPT
:
1973 err
= sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], (int __user
*)a
[4]);
1976 err
= sys_sendmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1979 err
= sys_recvmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
1988 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
1991 * This function is called by a protocol handler that wants to
1992 * advertise its address family, and have it linked into the
1996 int sock_register(struct net_proto_family
*ops
)
2000 if (ops
->family
>= NPROTO
) {
2001 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2004 net_family_write_lock();
2006 if (net_families
[ops
->family
] == NULL
) {
2007 net_families
[ops
->family
]=ops
;
2010 net_family_write_unlock();
2011 printk(KERN_INFO
"NET: Registered protocol family %d\n",
2017 * This function is called by a protocol handler that wants to
2018 * remove its address family, and have it unlinked from the
2022 int sock_unregister(int family
)
2024 if (family
< 0 || family
>= NPROTO
)
2027 net_family_write_lock();
2028 net_families
[family
]=NULL
;
2029 net_family_write_unlock();
2030 printk(KERN_INFO
"NET: Unregistered protocol family %d\n",
2035 static int __init
sock_init(void)
2038 * Initialize sock SLAB cache.
2044 * Initialize skbuff SLAB cache
2049 * Initialize the protocols module.
2053 register_filesystem(&sock_fs_type
);
2054 sock_mnt
= kern_mount(&sock_fs_type
);
2056 /* The real protocol initialization is performed in later initcalls.
2059 #ifdef CONFIG_NETFILTER
2066 core_initcall(sock_init
); /* early initcall */
2068 #ifdef CONFIG_PROC_FS
2069 void socket_seq_show(struct seq_file
*seq
)
2074 for (cpu
= 0; cpu
< NR_CPUS
; cpu
++)
2075 counter
+= per_cpu(sockets_in_use
, cpu
);
2077 /* It can be negative, by the way. 8) */
2081 seq_printf(seq
, "sockets: used %d\n", counter
);
2083 #endif /* CONFIG_PROC_FS */
2085 /* ABI emulation layers need these two */
2086 EXPORT_SYMBOL(move_addr_to_kernel
);
2087 EXPORT_SYMBOL(move_addr_to_user
);
2088 EXPORT_SYMBOL(sock_create
);
2089 EXPORT_SYMBOL(sock_create_kern
);
2090 EXPORT_SYMBOL(sock_create_lite
);
2091 EXPORT_SYMBOL(sock_map_fd
);
2092 EXPORT_SYMBOL(sock_recvmsg
);
2093 EXPORT_SYMBOL(sock_register
);
2094 EXPORT_SYMBOL(sock_release
);
2095 EXPORT_SYMBOL(sock_sendmsg
);
2096 EXPORT_SYMBOL(sock_unregister
);
2097 EXPORT_SYMBOL(sock_wake_async
);
2098 EXPORT_SYMBOL(sockfd_lookup
);
2099 EXPORT_SYMBOL(kernel_sendmsg
);
2100 EXPORT_SYMBOL(kernel_recvmsg
);