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/mutex.h>
72 #include <linux/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/divert.h>
82 #include <linux/mount.h>
83 #include <linux/security.h>
84 #include <linux/syscalls.h>
85 #include <linux/compat.h>
86 #include <linux/kmod.h>
87 #include <linux/audit.h>
88 #include <linux/wireless.h>
90 #include <asm/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
96 #include <linux/netfilter.h>
98 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
99 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*buf
,
100 size_t size
, loff_t pos
);
101 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*buf
,
102 size_t size
, loff_t pos
);
103 static int sock_mmap(struct file
*file
, struct vm_area_struct
* vma
);
105 static int sock_close(struct inode
*inode
, struct file
*file
);
106 static unsigned int sock_poll(struct file
*file
,
107 struct poll_table_struct
*wait
);
108 static long sock_ioctl(struct file
*file
,
109 unsigned int cmd
, unsigned long arg
);
111 static long compat_sock_ioctl(struct file
*file
,
112 unsigned int cmd
, unsigned long arg
);
114 static int sock_fasync(int fd
, struct file
*filp
, int on
);
115 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
116 unsigned long count
, loff_t
*ppos
);
117 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
118 unsigned long count
, loff_t
*ppos
);
119 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
120 int offset
, size_t size
, loff_t
*ppos
, int more
);
124 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
125 * in the operation structures but are done directly via the socketcall() multiplexor.
128 static struct file_operations socket_file_ops
= {
129 .owner
= THIS_MODULE
,
131 .aio_read
= sock_aio_read
,
132 .aio_write
= sock_aio_write
,
134 .unlocked_ioctl
= sock_ioctl
,
136 .compat_ioctl
= compat_sock_ioctl
,
139 .open
= sock_no_open
, /* special open code to disallow open via /proc */
140 .release
= sock_close
,
141 .fasync
= sock_fasync
,
143 .writev
= sock_writev
,
144 .sendpage
= sock_sendpage
148 * The protocol list. Each protocol is registered in here.
151 static struct net_proto_family
*net_families
[NPROTO
];
153 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
154 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
155 static DEFINE_SPINLOCK(net_family_lock
);
157 /* The strategy is: modifications net_family vector are short, do not
158 sleep and veeery rare, but read access should be free of any exclusive
162 static void net_family_write_lock(void)
164 spin_lock(&net_family_lock
);
165 while (atomic_read(&net_family_lockct
) != 0) {
166 spin_unlock(&net_family_lock
);
170 spin_lock(&net_family_lock
);
174 static __inline__
void net_family_write_unlock(void)
176 spin_unlock(&net_family_lock
);
179 static __inline__
void net_family_read_lock(void)
181 atomic_inc(&net_family_lockct
);
182 spin_unlock_wait(&net_family_lock
);
185 static __inline__
void net_family_read_unlock(void)
187 atomic_dec(&net_family_lockct
);
191 #define net_family_write_lock() do { } while(0)
192 #define net_family_write_unlock() do { } while(0)
193 #define net_family_read_lock() do { } while(0)
194 #define net_family_read_unlock() do { } while(0)
199 * Statistics counters of the socket lists
202 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
205 * Support routines. Move socket addresses back and forth across the kernel/user
206 * divide and look after the messy bits.
209 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
210 16 for IP, 16 for IPX,
213 must be at least one bigger than
214 the AF_UNIX size (see net/unix/af_unix.c
219 * move_addr_to_kernel - copy a socket address into kernel space
220 * @uaddr: Address in user space
221 * @kaddr: Address in kernel space
222 * @ulen: Length in user space
224 * The address is copied into kernel space. If the provided address is
225 * too long an error code of -EINVAL is returned. If the copy gives
226 * invalid addresses -EFAULT is returned. On a success 0 is returned.
229 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
231 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
235 if(copy_from_user(kaddr
,uaddr
,ulen
))
237 return audit_sockaddr(ulen
, kaddr
);
241 * move_addr_to_user - copy an address to user space
242 * @kaddr: kernel space address
243 * @klen: length of address in kernel
244 * @uaddr: user space address
245 * @ulen: pointer to user length field
247 * The value pointed to by ulen on entry is the buffer length available.
248 * This is overwritten with the buffer space used. -EINVAL is returned
249 * if an overlong buffer is specified or a negative buffer size. -EFAULT
250 * is returned if either the buffer or the length field are not
252 * After copying the data up to the limit the user specifies, the true
253 * length of the data is written over the length limit the user
254 * specified. Zero is returned for a success.
257 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
, int __user
*ulen
)
262 if((err
=get_user(len
, ulen
)))
266 if(len
<0 || len
> MAX_SOCK_ADDR
)
270 if(copy_to_user(uaddr
,kaddr
,len
))
274 * "fromlen shall refer to the value before truncation.."
277 return __put_user(klen
, ulen
);
280 #define SOCKFS_MAGIC 0x534F434B
282 static kmem_cache_t
* sock_inode_cachep __read_mostly
;
284 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
286 struct socket_alloc
*ei
;
287 ei
= (struct socket_alloc
*)kmem_cache_alloc(sock_inode_cachep
, SLAB_KERNEL
);
290 init_waitqueue_head(&ei
->socket
.wait
);
292 ei
->socket
.fasync_list
= NULL
;
293 ei
->socket
.state
= SS_UNCONNECTED
;
294 ei
->socket
.flags
= 0;
295 ei
->socket
.ops
= NULL
;
296 ei
->socket
.sk
= NULL
;
297 ei
->socket
.file
= NULL
;
298 ei
->socket
.flags
= 0;
300 return &ei
->vfs_inode
;
303 static void sock_destroy_inode(struct inode
*inode
)
305 kmem_cache_free(sock_inode_cachep
,
306 container_of(inode
, struct socket_alloc
, vfs_inode
));
309 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
311 struct socket_alloc
*ei
= (struct socket_alloc
*) foo
;
313 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
314 SLAB_CTOR_CONSTRUCTOR
)
315 inode_init_once(&ei
->vfs_inode
);
318 static int init_inodecache(void)
320 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
321 sizeof(struct socket_alloc
),
322 0, SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
,
324 if (sock_inode_cachep
== NULL
)
329 static struct super_operations sockfs_ops
= {
330 .alloc_inode
= sock_alloc_inode
,
331 .destroy_inode
=sock_destroy_inode
,
332 .statfs
= simple_statfs
,
335 static struct super_block
*sockfs_get_sb(struct file_system_type
*fs_type
,
336 int flags
, const char *dev_name
, void *data
)
338 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
);
341 static struct vfsmount
*sock_mnt __read_mostly
;
343 static struct file_system_type sock_fs_type
= {
345 .get_sb
= sockfs_get_sb
,
346 .kill_sb
= kill_anon_super
,
348 static int sockfs_delete_dentry(struct dentry
*dentry
)
352 static struct dentry_operations sockfs_dentry_operations
= {
353 .d_delete
= sockfs_delete_dentry
,
357 * Obtains the first available file descriptor and sets it up for use.
359 * These functions create file structures and maps them to fd space
360 * of the current process. On success it returns file descriptor
361 * and file struct implicitly stored in sock->file.
362 * Note that another thread may close file descriptor before we return
363 * from this function. We use the fact that now we do not refer
364 * to socket after mapping. If one day we will need it, this
365 * function will increment ref. count on file by 1.
367 * In any case returned fd MAY BE not valid!
368 * This race condition is unavoidable
369 * with shared fd spaces, we cannot solve it inside kernel,
370 * but we take care of internal coherence yet.
373 static int sock_alloc_fd(struct file
**filep
)
377 fd
= get_unused_fd();
378 if (likely(fd
>= 0)) {
379 struct file
*file
= get_empty_filp();
382 if (unlikely(!file
)) {
391 static int sock_attach_fd(struct socket
*sock
, struct file
*file
)
396 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
398 this.hash
= SOCK_INODE(sock
)->i_ino
;
400 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
401 if (unlikely(!file
->f_dentry
))
404 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
405 d_add(file
->f_dentry
, SOCK_INODE(sock
));
406 file
->f_vfsmnt
= mntget(sock_mnt
);
407 file
->f_mapping
= file
->f_dentry
->d_inode
->i_mapping
;
410 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
411 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
412 file
->f_flags
= O_RDWR
;
414 file
->private_data
= sock
;
419 int sock_map_fd(struct socket
*sock
)
421 struct file
*newfile
;
422 int fd
= sock_alloc_fd(&newfile
);
424 if (likely(fd
>= 0)) {
425 int err
= sock_attach_fd(sock
, newfile
);
427 if (unlikely(err
< 0)) {
432 fd_install(fd
, newfile
);
437 static struct socket
*sock_from_file(struct file
*file
, int *err
)
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
)) {
451 sock
= SOCKET_I(inode
);
452 if (sock
->file
!= file
) {
453 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
460 * sockfd_lookup - Go from a file number to its socket slot
462 * @err: pointer to an error code return
464 * The file handle passed in is locked and the socket it is bound
465 * too is returned. If an error occurs the err pointer is overwritten
466 * with a negative errno code and NULL is returned. The function checks
467 * for both invalid handles and passing a handle which is not a socket.
469 * On a success the socket object pointer is returned.
472 struct socket
*sockfd_lookup(int fd
, int *err
)
477 if (!(file
= fget(fd
))) {
481 sock
= sock_from_file(file
, err
);
487 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
492 file
= fget_light(fd
, fput_needed
);
494 sock
= sock_from_file(file
, err
);
497 fput_light(file
, *fput_needed
);
503 * sock_alloc - allocate a socket
505 * Allocate a new inode and socket object. The two are bound together
506 * and initialised. The socket is then returned. If we are out of inodes
510 static struct socket
*sock_alloc(void)
512 struct inode
* inode
;
513 struct socket
* sock
;
515 inode
= new_inode(sock_mnt
->mnt_sb
);
519 sock
= SOCKET_I(inode
);
521 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
522 inode
->i_uid
= current
->fsuid
;
523 inode
->i_gid
= current
->fsgid
;
525 get_cpu_var(sockets_in_use
)++;
526 put_cpu_var(sockets_in_use
);
531 * In theory you can't get an open on this inode, but /proc provides
532 * a back door. Remember to keep it shut otherwise you'll let the
533 * creepy crawlies in.
536 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
541 struct file_operations bad_sock_fops
= {
542 .owner
= THIS_MODULE
,
543 .open
= sock_no_open
,
547 * sock_release - close a socket
548 * @sock: socket to close
550 * The socket is released from the protocol stack if it has a release
551 * callback, and the inode is then released if the socket is bound to
552 * an inode not a file.
555 void sock_release(struct socket
*sock
)
558 struct module
*owner
= sock
->ops
->owner
;
560 sock
->ops
->release(sock
);
565 if (sock
->fasync_list
)
566 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
568 get_cpu_var(sockets_in_use
)--;
569 put_cpu_var(sockets_in_use
);
571 iput(SOCK_INODE(sock
));
577 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
578 struct msghdr
*msg
, size_t size
)
580 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
588 err
= security_socket_sendmsg(sock
, msg
, size
);
592 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
595 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
598 struct sock_iocb siocb
;
601 init_sync_kiocb(&iocb
, NULL
);
602 iocb
.private = &siocb
;
603 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
604 if (-EIOCBQUEUED
== ret
)
605 ret
= wait_on_sync_kiocb(&iocb
);
609 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
610 struct kvec
*vec
, size_t num
, size_t size
)
612 mm_segment_t oldfs
= get_fs();
617 * the following is safe, since for compiler definitions of kvec and
618 * iovec are identical, yielding the same in-core layout and alignment
620 msg
->msg_iov
= (struct iovec
*)vec
,
621 msg
->msg_iovlen
= num
;
622 result
= sock_sendmsg(sock
, msg
, size
);
627 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
628 struct msghdr
*msg
, size_t size
, int flags
)
631 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
639 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
643 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
646 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
647 size_t size
, int flags
)
650 struct sock_iocb siocb
;
653 init_sync_kiocb(&iocb
, NULL
);
654 iocb
.private = &siocb
;
655 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
656 if (-EIOCBQUEUED
== ret
)
657 ret
= wait_on_sync_kiocb(&iocb
);
661 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
662 struct kvec
*vec
, size_t num
,
663 size_t size
, int flags
)
665 mm_segment_t oldfs
= get_fs();
670 * the following is safe, since for compiler definitions of kvec and
671 * iovec are identical, yielding the same in-core layout and alignment
673 msg
->msg_iov
= (struct iovec
*)vec
,
674 msg
->msg_iovlen
= num
;
675 result
= sock_recvmsg(sock
, msg
, size
, flags
);
680 static void sock_aio_dtor(struct kiocb
*iocb
)
682 kfree(iocb
->private);
685 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
686 int offset
, size_t size
, loff_t
*ppos
, int more
)
691 sock
= file
->private_data
;
693 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
697 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
700 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
701 char __user
*ubuf
, size_t size
, struct sock_iocb
*siocb
)
703 if (!is_sync_kiocb(iocb
)) {
704 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
707 iocb
->ki_dtor
= sock_aio_dtor
;
711 siocb
->async_iov
.iov_base
= ubuf
;
712 siocb
->async_iov
.iov_len
= size
;
714 iocb
->private = siocb
;
718 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
719 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
721 struct socket
*sock
= file
->private_data
;
725 for (i
= 0 ; i
< nr_segs
; i
++)
726 size
+= iov
[i
].iov_len
;
728 msg
->msg_name
= NULL
;
729 msg
->msg_namelen
= 0;
730 msg
->msg_control
= NULL
;
731 msg
->msg_controllen
= 0;
732 msg
->msg_iov
= (struct iovec
*) iov
;
733 msg
->msg_iovlen
= nr_segs
;
734 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
736 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
739 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
740 unsigned long nr_segs
, loff_t
*ppos
)
743 struct sock_iocb siocb
;
747 init_sync_kiocb(&iocb
, NULL
);
748 iocb
.private = &siocb
;
750 ret
= do_sock_read(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
751 if (-EIOCBQUEUED
== ret
)
752 ret
= wait_on_sync_kiocb(&iocb
);
756 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
757 size_t count
, loff_t pos
)
759 struct sock_iocb siocb
, *x
;
763 if (count
== 0) /* Match SYS5 behaviour */
766 x
= alloc_sock_iocb(iocb
, ubuf
, count
, &siocb
);
769 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
,
773 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
774 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
776 struct socket
*sock
= file
->private_data
;
780 for (i
= 0 ; i
< nr_segs
; i
++)
781 size
+= iov
[i
].iov_len
;
783 msg
->msg_name
= NULL
;
784 msg
->msg_namelen
= 0;
785 msg
->msg_control
= NULL
;
786 msg
->msg_controllen
= 0;
787 msg
->msg_iov
= (struct iovec
*) iov
;
788 msg
->msg_iovlen
= nr_segs
;
789 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
790 if (sock
->type
== SOCK_SEQPACKET
)
791 msg
->msg_flags
|= MSG_EOR
;
793 return __sock_sendmsg(iocb
, sock
, msg
, size
);
796 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
797 unsigned long nr_segs
, loff_t
*ppos
)
801 struct sock_iocb siocb
;
804 init_sync_kiocb(&iocb
, NULL
);
805 iocb
.private = &siocb
;
807 ret
= do_sock_write(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
808 if (-EIOCBQUEUED
== ret
)
809 ret
= wait_on_sync_kiocb(&iocb
);
813 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
814 size_t count
, loff_t pos
)
816 struct sock_iocb siocb
, *x
;
820 if (count
== 0) /* Match SYS5 behaviour */
823 x
= alloc_sock_iocb(iocb
, (void __user
*)ubuf
, count
, &siocb
);
827 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
,
833 * Atomic setting of ioctl hooks to avoid race
834 * with module unload.
837 static DEFINE_MUTEX(br_ioctl_mutex
);
838 static int (*br_ioctl_hook
)(unsigned int cmd
, void __user
*arg
) = NULL
;
840 void brioctl_set(int (*hook
)(unsigned int, void __user
*))
842 mutex_lock(&br_ioctl_mutex
);
843 br_ioctl_hook
= hook
;
844 mutex_unlock(&br_ioctl_mutex
);
846 EXPORT_SYMBOL(brioctl_set
);
848 static DEFINE_MUTEX(vlan_ioctl_mutex
);
849 static int (*vlan_ioctl_hook
)(void __user
*arg
);
851 void vlan_ioctl_set(int (*hook
)(void __user
*))
853 mutex_lock(&vlan_ioctl_mutex
);
854 vlan_ioctl_hook
= hook
;
855 mutex_unlock(&vlan_ioctl_mutex
);
857 EXPORT_SYMBOL(vlan_ioctl_set
);
859 static DEFINE_MUTEX(dlci_ioctl_mutex
);
860 static int (*dlci_ioctl_hook
)(unsigned int, void __user
*);
862 void dlci_ioctl_set(int (*hook
)(unsigned int, void __user
*))
864 mutex_lock(&dlci_ioctl_mutex
);
865 dlci_ioctl_hook
= hook
;
866 mutex_unlock(&dlci_ioctl_mutex
);
868 EXPORT_SYMBOL(dlci_ioctl_set
);
871 * With an ioctl, arg may well be a user mode pointer, but we don't know
872 * what to do with it - that's up to the protocol still.
875 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
878 void __user
*argp
= (void __user
*)arg
;
881 sock
= file
->private_data
;
882 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
883 err
= dev_ioctl(cmd
, argp
);
885 #ifdef CONFIG_WIRELESS_EXT
886 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
887 err
= dev_ioctl(cmd
, argp
);
889 #endif /* CONFIG_WIRELESS_EXT */
894 if (get_user(pid
, (int __user
*)argp
))
896 err
= f_setown(sock
->file
, pid
, 1);
900 err
= put_user(sock
->file
->f_owner
.pid
, (int __user
*)argp
);
908 request_module("bridge");
910 mutex_lock(&br_ioctl_mutex
);
912 err
= br_ioctl_hook(cmd
, argp
);
913 mutex_unlock(&br_ioctl_mutex
);
918 if (!vlan_ioctl_hook
)
919 request_module("8021q");
921 mutex_lock(&vlan_ioctl_mutex
);
923 err
= vlan_ioctl_hook(argp
);
924 mutex_unlock(&vlan_ioctl_mutex
);
928 /* Convert this to call through a hook */
929 err
= divert_ioctl(cmd
, argp
);
934 if (!dlci_ioctl_hook
)
935 request_module("dlci");
937 if (dlci_ioctl_hook
) {
938 mutex_lock(&dlci_ioctl_mutex
);
939 err
= dlci_ioctl_hook(cmd
, argp
);
940 mutex_unlock(&dlci_ioctl_mutex
);
944 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
947 * If this ioctl is unknown try to hand it down
950 if (err
== -ENOIOCTLCMD
)
951 err
= dev_ioctl(cmd
, argp
);
957 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
960 struct socket
*sock
= NULL
;
962 err
= security_socket_create(family
, type
, protocol
, 1);
972 security_socket_post_create(sock
, family
, type
, protocol
, 1);
979 /* No kernel lock held - perfect */
980 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
985 * We can't return errors to poll, so it's either yes or no.
987 sock
= file
->private_data
;
988 return sock
->ops
->poll(file
, sock
, wait
);
991 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
993 struct socket
*sock
= file
->private_data
;
995 return sock
->ops
->mmap(file
, sock
, vma
);
998 static int sock_close(struct inode
*inode
, struct file
*filp
)
1001 * It was possible the inode is NULL we were
1002 * closing an unfinished socket.
1007 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1010 sock_fasync(-1, filp
, 0);
1011 sock_release(SOCKET_I(inode
));
1016 * Update the socket async list
1018 * Fasync_list locking strategy.
1020 * 1. fasync_list is modified only under process context socket lock
1021 * i.e. under semaphore.
1022 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1023 * or under socket lock.
1024 * 3. fasync_list can be used from softirq context, so that
1025 * modification under socket lock have to be enhanced with
1026 * write_lock_bh(&sk->sk_callback_lock).
1030 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1032 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
1033 struct socket
*sock
;
1038 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1043 sock
= filp
->private_data
;
1045 if ((sk
=sock
->sk
) == NULL
) {
1052 prev
=&(sock
->fasync_list
);
1054 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
1055 if (fa
->fa_file
==filp
)
1062 write_lock_bh(&sk
->sk_callback_lock
);
1064 write_unlock_bh(&sk
->sk_callback_lock
);
1071 fna
->magic
=FASYNC_MAGIC
;
1072 fna
->fa_next
=sock
->fasync_list
;
1073 write_lock_bh(&sk
->sk_callback_lock
);
1074 sock
->fasync_list
=fna
;
1075 write_unlock_bh(&sk
->sk_callback_lock
);
1081 write_lock_bh(&sk
->sk_callback_lock
);
1083 write_unlock_bh(&sk
->sk_callback_lock
);
1089 release_sock(sock
->sk
);
1093 /* This function may be called only under socket lock or callback_lock */
1095 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1097 if (!sock
|| !sock
->fasync_list
)
1103 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1107 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1112 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1115 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1120 static int __sock_create(int family
, int type
, int protocol
, struct socket
**res
, int kern
)
1123 struct socket
*sock
;
1126 * Check protocol is in range
1128 if (family
< 0 || family
>= NPROTO
)
1129 return -EAFNOSUPPORT
;
1130 if (type
< 0 || type
>= SOCK_MAX
)
1135 This uglymoron is moved from INET layer to here to avoid
1136 deadlock in module load.
1138 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1142 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
1147 err
= security_socket_create(family
, type
, protocol
, kern
);
1151 #if defined(CONFIG_KMOD)
1152 /* Attempt to load a protocol module if the find failed.
1154 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1155 * requested real, full-featured networking support upon configuration.
1156 * Otherwise module support will break!
1158 if (net_families
[family
]==NULL
)
1160 request_module("net-pf-%d",family
);
1164 net_family_read_lock();
1165 if (net_families
[family
] == NULL
) {
1166 err
= -EAFNOSUPPORT
;
1171 * Allocate the socket and allow the family to set things up. if
1172 * the protocol is 0, the family is instructed to select an appropriate
1176 if (!(sock
= sock_alloc())) {
1177 printk(KERN_WARNING
"socket: no more sockets\n");
1178 err
= -ENFILE
; /* Not exactly a match, but its the
1179 closest posix thing */
1186 * We will call the ->create function, that possibly is in a loadable
1187 * module, so we have to bump that loadable module refcnt first.
1189 err
= -EAFNOSUPPORT
;
1190 if (!try_module_get(net_families
[family
]->owner
))
1193 if ((err
= net_families
[family
]->create(sock
, protocol
)) < 0) {
1195 goto out_module_put
;
1199 * Now to bump the refcnt of the [loadable] module that owns this
1200 * socket at sock_release time we decrement its refcnt.
1202 if (!try_module_get(sock
->ops
->owner
)) {
1204 goto out_module_put
;
1207 * Now that we're done with the ->create function, the [loadable]
1208 * module can have its refcnt decremented
1210 module_put(net_families
[family
]->owner
);
1212 security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1215 net_family_read_unlock();
1218 module_put(net_families
[family
]->owner
);
1224 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1226 return __sock_create(family
, type
, protocol
, res
, 0);
1229 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1231 return __sock_create(family
, type
, protocol
, res
, 1);
1234 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1237 struct socket
*sock
;
1239 retval
= sock_create(family
, type
, protocol
, &sock
);
1243 retval
= sock_map_fd(sock
);
1248 /* It may be already another descriptor 8) Not kernel problem. */
1257 * Create a pair of connected sockets.
1260 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1262 struct socket
*sock1
, *sock2
;
1266 * Obtain the first socket and check if the underlying protocol
1267 * supports the socketpair call.
1270 err
= sock_create(family
, type
, protocol
, &sock1
);
1274 err
= sock_create(family
, type
, protocol
, &sock2
);
1278 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1280 goto out_release_both
;
1284 err
= sock_map_fd(sock1
);
1286 goto out_release_both
;
1289 err
= sock_map_fd(sock2
);
1294 /* fd1 and fd2 may be already another descriptors.
1295 * Not kernel problem.
1298 err
= put_user(fd1
, &usockvec
[0]);
1300 err
= put_user(fd2
, &usockvec
[1]);
1309 sock_release(sock2
);
1314 sock_release(sock2
);
1316 sock_release(sock1
);
1323 * Bind a name to a socket. Nothing much to do here since it's
1324 * the protocol's responsibility to handle the local address.
1326 * We move the socket address to kernel space before we call
1327 * the protocol layer (having also checked the address is ok).
1330 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1332 struct socket
*sock
;
1333 char address
[MAX_SOCK_ADDR
];
1334 int err
, fput_needed
;
1336 if((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
))!=NULL
)
1338 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0) {
1339 err
= security_socket_bind(sock
, (struct sockaddr
*)address
, addrlen
);
1341 err
= sock
->ops
->bind(sock
,
1342 (struct sockaddr
*)address
, addrlen
);
1344 fput_light(sock
->file
, fput_needed
);
1351 * Perform a listen. Basically, we allow the protocol to do anything
1352 * necessary for a listen, and if that works, we mark the socket as
1353 * ready for listening.
1356 int sysctl_somaxconn
= SOMAXCONN
;
1358 asmlinkage
long sys_listen(int fd
, int backlog
)
1360 struct socket
*sock
;
1361 int err
, fput_needed
;
1363 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1364 if ((unsigned) backlog
> sysctl_somaxconn
)
1365 backlog
= sysctl_somaxconn
;
1367 err
= security_socket_listen(sock
, backlog
);
1369 err
= sock
->ops
->listen(sock
, backlog
);
1371 fput_light(sock
->file
, fput_needed
);
1378 * For accept, we attempt to create a new socket, set up the link
1379 * with the client, wake up the client, then return the new
1380 * connected fd. We collect the address of the connector in kernel
1381 * space and move it to user at the very end. This is unclean because
1382 * we open the socket then return an error.
1384 * 1003.1g adds the ability to recvmsg() to query connection pending
1385 * status to recvmsg. We need to add that support in a way thats
1386 * clean when we restucture accept also.
1389 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
, int __user
*upeer_addrlen
)
1391 struct socket
*sock
, *newsock
;
1392 struct file
*newfile
;
1393 int err
, len
, newfd
, fput_needed
;
1394 char address
[MAX_SOCK_ADDR
];
1396 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1401 if (!(newsock
= sock_alloc()))
1404 newsock
->type
= sock
->type
;
1405 newsock
->ops
= sock
->ops
;
1408 * We don't need try_module_get here, as the listening socket (sock)
1409 * has the protocol module (sock->ops->owner) held.
1411 __module_get(newsock
->ops
->owner
);
1413 newfd
= sock_alloc_fd(&newfile
);
1414 if (unlikely(newfd
< 0)) {
1419 err
= sock_attach_fd(newsock
, newfile
);
1423 err
= security_socket_accept(sock
, newsock
);
1427 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1431 if (upeer_sockaddr
) {
1432 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1433 err
= -ECONNABORTED
;
1436 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1441 /* File flags are not inherited via accept() unlike another OSes. */
1443 fd_install(newfd
, newfile
);
1446 security_socket_post_accept(sock
, newsock
);
1449 fput_light(sock
->file
, fput_needed
);
1454 put_unused_fd(newfd
);
1456 sock_release(newsock
);
1462 * Attempt to connect to a socket with the server address. The address
1463 * is in user space so we verify it is OK and move it to kernel space.
1465 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1468 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1469 * other SEQPACKET protocols that take time to connect() as it doesn't
1470 * include the -EINPROGRESS status for such sockets.
1473 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1475 struct socket
*sock
;
1476 char address
[MAX_SOCK_ADDR
];
1477 int err
, fput_needed
;
1479 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1482 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1486 err
= security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1490 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1491 sock
->file
->f_flags
);
1493 fput_light(sock
->file
, fput_needed
);
1499 * Get the local address ('name') of a socket object. Move the obtained
1500 * name to user space.
1503 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1505 struct socket
*sock
;
1506 char address
[MAX_SOCK_ADDR
];
1507 int len
, err
, fput_needed
;
1509 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1513 err
= security_socket_getsockname(sock
);
1517 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1520 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1523 fput_light(sock
->file
, fput_needed
);
1529 * Get the remote address ('name') of a socket object. Move the obtained
1530 * name to user space.
1533 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1535 struct socket
*sock
;
1536 char address
[MAX_SOCK_ADDR
];
1537 int len
, err
, fput_needed
;
1539 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1540 err
= security_socket_getpeername(sock
);
1542 fput_light(sock
->file
, fput_needed
);
1546 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1548 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1549 fput_light(sock
->file
, fput_needed
);
1555 * Send a datagram to a given address. We move the address into kernel
1556 * space and check the user space data area is readable before invoking
1560 asmlinkage
long sys_sendto(int fd
, void __user
* buff
, size_t len
, unsigned flags
,
1561 struct sockaddr __user
*addr
, int addr_len
)
1563 struct socket
*sock
;
1564 char address
[MAX_SOCK_ADDR
];
1569 struct file
*sock_file
;
1571 sock_file
= fget_light(fd
, &fput_needed
);
1575 sock
= sock_from_file(sock_file
, &err
);
1583 msg
.msg_control
=NULL
;
1584 msg
.msg_controllen
=0;
1587 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1590 msg
.msg_name
=address
;
1591 msg
.msg_namelen
=addr_len
;
1593 if (sock
->file
->f_flags
& O_NONBLOCK
)
1594 flags
|= MSG_DONTWAIT
;
1595 msg
.msg_flags
= flags
;
1596 err
= sock_sendmsg(sock
, &msg
, len
);
1599 fput_light(sock_file
, fput_needed
);
1604 * Send a datagram down a socket.
1607 asmlinkage
long sys_send(int fd
, void __user
* buff
, size_t len
, unsigned flags
)
1609 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1613 * Receive a frame from the socket and optionally record the address of the
1614 * sender. We verify the buffers are writable and if needed move the
1615 * sender address from kernel to user space.
1618 asmlinkage
long sys_recvfrom(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
,
1619 struct sockaddr __user
*addr
, int __user
*addr_len
)
1621 struct socket
*sock
;
1624 char address
[MAX_SOCK_ADDR
];
1626 struct file
*sock_file
;
1629 sock_file
= fget_light(fd
, &fput_needed
);
1633 sock
= sock_from_file(sock_file
, &err
);
1637 msg
.msg_control
=NULL
;
1638 msg
.msg_controllen
=0;
1643 msg
.msg_name
=address
;
1644 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1645 if (sock
->file
->f_flags
& O_NONBLOCK
)
1646 flags
|= MSG_DONTWAIT
;
1647 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1649 if(err
>= 0 && addr
!= NULL
)
1651 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1656 fput_light(sock_file
, fput_needed
);
1661 * Receive a datagram from a socket.
1664 asmlinkage
long sys_recv(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
)
1666 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1670 * Set a socket option. Because we don't know the option lengths we have
1671 * to pass the user mode parameter for the protocols to sort out.
1674 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char __user
*optval
, int optlen
)
1676 int err
, fput_needed
;
1677 struct socket
*sock
;
1682 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
)
1684 err
= security_socket_setsockopt(sock
,level
,optname
);
1688 if (level
== SOL_SOCKET
)
1689 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1691 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1693 fput_light(sock
->file
, fput_needed
);
1699 * Get a socket option. Because we don't know the option lengths we have
1700 * to pass a user mode parameter for the protocols to sort out.
1703 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
, int __user
*optlen
)
1705 int err
, fput_needed
;
1706 struct socket
*sock
;
1708 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1709 err
= security_socket_getsockopt(sock
, level
, optname
);
1713 if (level
== SOL_SOCKET
)
1714 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1716 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1718 fput_light(sock
->file
, fput_needed
);
1725 * Shutdown a socket.
1728 asmlinkage
long sys_shutdown(int fd
, int how
)
1730 int err
, fput_needed
;
1731 struct socket
*sock
;
1733 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
))!=NULL
)
1735 err
= security_socket_shutdown(sock
, how
);
1737 err
= sock
->ops
->shutdown(sock
, how
);
1738 fput_light(sock
->file
, fput_needed
);
1743 /* A couple of helpful macros for getting the address of the 32/64 bit
1744 * fields which are the same type (int / unsigned) on our platforms.
1746 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1747 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1748 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1752 * BSD sendmsg interface
1755 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1757 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1758 struct socket
*sock
;
1759 char address
[MAX_SOCK_ADDR
];
1760 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1761 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1762 __attribute__ ((aligned (sizeof(__kernel_size_t
))));
1763 /* 20 is size of ipv6_pktinfo */
1764 unsigned char *ctl_buf
= ctl
;
1765 struct msghdr msg_sys
;
1766 int err
, ctl_len
, iov_size
, total_len
;
1770 if (MSG_CMSG_COMPAT
& flags
) {
1771 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1773 } else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1776 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1780 /* do not move before msg_sys is valid */
1782 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1785 /* Check whether to allocate the iovec area*/
1787 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1788 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1789 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1794 /* This will also move the address data into kernel space */
1795 if (MSG_CMSG_COMPAT
& flags
) {
1796 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1798 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1805 if (msg_sys
.msg_controllen
> INT_MAX
)
1807 ctl_len
= msg_sys
.msg_controllen
;
1808 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1809 err
= cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
, sizeof(ctl
));
1812 ctl_buf
= msg_sys
.msg_control
;
1813 ctl_len
= msg_sys
.msg_controllen
;
1814 } else if (ctl_len
) {
1815 if (ctl_len
> sizeof(ctl
))
1817 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1818 if (ctl_buf
== NULL
)
1823 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1824 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1825 * checking falls down on this.
1827 if (copy_from_user(ctl_buf
, (void __user
*) msg_sys
.msg_control
, ctl_len
))
1829 msg_sys
.msg_control
= ctl_buf
;
1831 msg_sys
.msg_flags
= flags
;
1833 if (sock
->file
->f_flags
& O_NONBLOCK
)
1834 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1835 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1839 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1841 if (iov
!= iovstack
)
1842 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1844 fput_light(sock
->file
, fput_needed
);
1850 * BSD recvmsg interface
1853 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned int flags
)
1855 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1856 struct socket
*sock
;
1857 struct iovec iovstack
[UIO_FASTIOV
];
1858 struct iovec
*iov
=iovstack
;
1859 struct msghdr msg_sys
;
1860 unsigned long cmsg_ptr
;
1861 int err
, iov_size
, total_len
, len
;
1864 /* kernel mode address */
1865 char addr
[MAX_SOCK_ADDR
];
1867 /* user mode address pointers */
1868 struct sockaddr __user
*uaddr
;
1869 int __user
*uaddr_len
;
1871 if (MSG_CMSG_COMPAT
& flags
) {
1872 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1875 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1878 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1883 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1886 /* Check whether to allocate the iovec area*/
1888 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1889 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1890 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1896 * Save the user-mode address (verify_iovec will change the
1897 * kernel msghdr to use the kernel address space)
1900 uaddr
= (void __user
*) msg_sys
.msg_name
;
1901 uaddr_len
= COMPAT_NAMELEN(msg
);
1902 if (MSG_CMSG_COMPAT
& flags
) {
1903 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1905 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1910 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1911 msg_sys
.msg_flags
= 0;
1912 if (MSG_CMSG_COMPAT
& flags
)
1913 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1915 if (sock
->file
->f_flags
& O_NONBLOCK
)
1916 flags
|= MSG_DONTWAIT
;
1917 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1922 if (uaddr
!= NULL
) {
1923 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1927 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1931 if (MSG_CMSG_COMPAT
& flags
)
1932 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1933 &msg_compat
->msg_controllen
);
1935 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1936 &msg
->msg_controllen
);
1942 if (iov
!= iovstack
)
1943 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1945 fput_light(sock
->file
, fput_needed
);
1950 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1952 /* Argument list sizes for sys_socketcall */
1953 #define AL(x) ((x) * sizeof(unsigned long))
1954 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1955 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1956 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1960 * System call vectors.
1962 * Argument checking cleaned up. Saved 20% in size.
1963 * This function doesn't need to set the kernel lock because
1964 * it is set by the callees.
1967 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1970 unsigned long a0
,a1
;
1973 if(call
<1||call
>SYS_RECVMSG
)
1976 /* copy_from_user should be SMP safe. */
1977 if (copy_from_user(a
, args
, nargs
[call
]))
1980 err
= audit_socketcall(nargs
[call
]/sizeof(unsigned long), a
);
1990 err
= sys_socket(a0
,a1
,a
[2]);
1993 err
= sys_bind(a0
,(struct sockaddr __user
*)a1
, a
[2]);
1996 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
1999 err
= sys_listen(a0
,a1
);
2002 err
= sys_accept(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
2004 case SYS_GETSOCKNAME
:
2005 err
= sys_getsockname(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
2007 case SYS_GETPEERNAME
:
2008 err
= sys_getpeername(a0
, (struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
2010 case SYS_SOCKETPAIR
:
2011 err
= sys_socketpair(a0
,a1
, a
[2], (int __user
*)a
[3]);
2014 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2017 err
= sys_sendto(a0
,(void __user
*)a1
, a
[2], a
[3],
2018 (struct sockaddr __user
*)a
[4], a
[5]);
2021 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2024 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2025 (struct sockaddr __user
*)a
[4], (int __user
*)a
[5]);
2028 err
= sys_shutdown(a0
,a1
);
2030 case SYS_SETSOCKOPT
:
2031 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2033 case SYS_GETSOCKOPT
:
2034 err
= sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], (int __user
*)a
[4]);
2037 err
= sys_sendmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
2040 err
= sys_recvmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
2049 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2052 * This function is called by a protocol handler that wants to
2053 * advertise its address family, and have it linked into the
2057 int sock_register(struct net_proto_family
*ops
)
2061 if (ops
->family
>= NPROTO
) {
2062 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2065 net_family_write_lock();
2067 if (net_families
[ops
->family
] == NULL
) {
2068 net_families
[ops
->family
]=ops
;
2071 net_family_write_unlock();
2072 printk(KERN_INFO
"NET: Registered protocol family %d\n",
2078 * This function is called by a protocol handler that wants to
2079 * remove its address family, and have it unlinked from the
2083 int sock_unregister(int family
)
2085 if (family
< 0 || family
>= NPROTO
)
2088 net_family_write_lock();
2089 net_families
[family
]=NULL
;
2090 net_family_write_unlock();
2091 printk(KERN_INFO
"NET: Unregistered protocol family %d\n",
2096 static int __init
sock_init(void)
2099 * Initialize sock SLAB cache.
2105 * Initialize skbuff SLAB cache
2110 * Initialize the protocols module.
2114 register_filesystem(&sock_fs_type
);
2115 sock_mnt
= kern_mount(&sock_fs_type
);
2117 /* The real protocol initialization is performed in later initcalls.
2120 #ifdef CONFIG_NETFILTER
2127 core_initcall(sock_init
); /* early initcall */
2129 #ifdef CONFIG_PROC_FS
2130 void socket_seq_show(struct seq_file
*seq
)
2136 counter
+= per_cpu(sockets_in_use
, cpu
);
2138 /* It can be negative, by the way. 8) */
2142 seq_printf(seq
, "sockets: used %d\n", counter
);
2144 #endif /* CONFIG_PROC_FS */
2146 #ifdef CONFIG_COMPAT
2147 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2150 struct socket
*sock
= file
->private_data
;
2151 int ret
= -ENOIOCTLCMD
;
2153 if (sock
->ops
->compat_ioctl
)
2154 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2160 /* ABI emulation layers need these two */
2161 EXPORT_SYMBOL(move_addr_to_kernel
);
2162 EXPORT_SYMBOL(move_addr_to_user
);
2163 EXPORT_SYMBOL(sock_create
);
2164 EXPORT_SYMBOL(sock_create_kern
);
2165 EXPORT_SYMBOL(sock_create_lite
);
2166 EXPORT_SYMBOL(sock_map_fd
);
2167 EXPORT_SYMBOL(sock_recvmsg
);
2168 EXPORT_SYMBOL(sock_register
);
2169 EXPORT_SYMBOL(sock_release
);
2170 EXPORT_SYMBOL(sock_sendmsg
);
2171 EXPORT_SYMBOL(sock_unregister
);
2172 EXPORT_SYMBOL(sock_wake_async
);
2173 EXPORT_SYMBOL(sockfd_lookup
);
2174 EXPORT_SYMBOL(kernel_sendmsg
);
2175 EXPORT_SYMBOL(kernel_recvmsg
);