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
);
110 static int sock_fasync(int fd
, struct file
*filp
, int on
);
111 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
112 unsigned long count
, loff_t
*ppos
);
113 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
114 unsigned long count
, loff_t
*ppos
);
115 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
116 int offset
, size_t size
, loff_t
*ppos
, int more
);
120 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
121 * in the operation structures but are done directly via the socketcall() multiplexor.
124 static struct file_operations socket_file_ops
= {
125 .owner
= THIS_MODULE
,
127 .aio_read
= sock_aio_read
,
128 .aio_write
= sock_aio_write
,
130 .unlocked_ioctl
= sock_ioctl
,
132 .open
= sock_no_open
, /* special open code to disallow open via /proc */
133 .release
= sock_close
,
134 .fasync
= sock_fasync
,
136 .writev
= sock_writev
,
137 .sendpage
= sock_sendpage
141 * The protocol list. Each protocol is registered in here.
144 static struct net_proto_family
*net_families
[NPROTO
];
146 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
147 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
148 static DEFINE_SPINLOCK(net_family_lock
);
150 /* The strategy is: modifications net_family vector are short, do not
151 sleep and veeery rare, but read access should be free of any exclusive
155 static void net_family_write_lock(void)
157 spin_lock(&net_family_lock
);
158 while (atomic_read(&net_family_lockct
) != 0) {
159 spin_unlock(&net_family_lock
);
163 spin_lock(&net_family_lock
);
167 static __inline__
void net_family_write_unlock(void)
169 spin_unlock(&net_family_lock
);
172 static __inline__
void net_family_read_lock(void)
174 atomic_inc(&net_family_lockct
);
175 spin_unlock_wait(&net_family_lock
);
178 static __inline__
void net_family_read_unlock(void)
180 atomic_dec(&net_family_lockct
);
184 #define net_family_write_lock() do { } while(0)
185 #define net_family_write_unlock() do { } while(0)
186 #define net_family_read_lock() do { } while(0)
187 #define net_family_read_unlock() do { } while(0)
192 * Statistics counters of the socket lists
195 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
198 * Support routines. Move socket addresses back and forth across the kernel/user
199 * divide and look after the messy bits.
202 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
203 16 for IP, 16 for IPX,
206 must be at least one bigger than
207 the AF_UNIX size (see net/unix/af_unix.c
212 * move_addr_to_kernel - copy a socket address into kernel space
213 * @uaddr: Address in user space
214 * @kaddr: Address in kernel space
215 * @ulen: Length in user space
217 * The address is copied into kernel space. If the provided address is
218 * too long an error code of -EINVAL is returned. If the copy gives
219 * invalid addresses -EFAULT is returned. On a success 0 is returned.
222 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
224 if(ulen
<0||ulen
>MAX_SOCK_ADDR
)
228 if(copy_from_user(kaddr
,uaddr
,ulen
))
230 return audit_sockaddr(ulen
, kaddr
);
234 * move_addr_to_user - copy an address to user space
235 * @kaddr: kernel space address
236 * @klen: length of address in kernel
237 * @uaddr: user space address
238 * @ulen: pointer to user length field
240 * The value pointed to by ulen on entry is the buffer length available.
241 * This is overwritten with the buffer space used. -EINVAL is returned
242 * if an overlong buffer is specified or a negative buffer size. -EFAULT
243 * is returned if either the buffer or the length field are not
245 * After copying the data up to the limit the user specifies, the true
246 * length of the data is written over the length limit the user
247 * specified. Zero is returned for a success.
250 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
, int __user
*ulen
)
255 if((err
=get_user(len
, ulen
)))
259 if(len
<0 || len
> MAX_SOCK_ADDR
)
263 if(copy_to_user(uaddr
,kaddr
,len
))
267 * "fromlen shall refer to the value before truncation.."
270 return __put_user(klen
, ulen
);
273 #define SOCKFS_MAGIC 0x534F434B
275 static kmem_cache_t
* sock_inode_cachep __read_mostly
;
277 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
279 struct socket_alloc
*ei
;
280 ei
= (struct socket_alloc
*)kmem_cache_alloc(sock_inode_cachep
, SLAB_KERNEL
);
283 init_waitqueue_head(&ei
->socket
.wait
);
285 ei
->socket
.fasync_list
= NULL
;
286 ei
->socket
.state
= SS_UNCONNECTED
;
287 ei
->socket
.flags
= 0;
288 ei
->socket
.ops
= NULL
;
289 ei
->socket
.sk
= NULL
;
290 ei
->socket
.file
= NULL
;
291 ei
->socket
.flags
= 0;
293 return &ei
->vfs_inode
;
296 static void sock_destroy_inode(struct inode
*inode
)
298 kmem_cache_free(sock_inode_cachep
,
299 container_of(inode
, struct socket_alloc
, vfs_inode
));
302 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
304 struct socket_alloc
*ei
= (struct socket_alloc
*) foo
;
306 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
307 SLAB_CTOR_CONSTRUCTOR
)
308 inode_init_once(&ei
->vfs_inode
);
311 static int init_inodecache(void)
313 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
314 sizeof(struct socket_alloc
),
315 0, SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
,
317 if (sock_inode_cachep
== NULL
)
322 static struct super_operations sockfs_ops
= {
323 .alloc_inode
= sock_alloc_inode
,
324 .destroy_inode
=sock_destroy_inode
,
325 .statfs
= simple_statfs
,
328 static struct super_block
*sockfs_get_sb(struct file_system_type
*fs_type
,
329 int flags
, const char *dev_name
, void *data
)
331 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
);
334 static struct vfsmount
*sock_mnt __read_mostly
;
336 static struct file_system_type sock_fs_type
= {
338 .get_sb
= sockfs_get_sb
,
339 .kill_sb
= kill_anon_super
,
341 static int sockfs_delete_dentry(struct dentry
*dentry
)
345 static struct dentry_operations sockfs_dentry_operations
= {
346 .d_delete
= sockfs_delete_dentry
,
350 * Obtains the first available file descriptor and sets it up for use.
352 * These functions create file structures and maps them to fd space
353 * of the current process. On success it returns file descriptor
354 * and file struct implicitly stored in sock->file.
355 * Note that another thread may close file descriptor before we return
356 * from this function. We use the fact that now we do not refer
357 * to socket after mapping. If one day we will need it, this
358 * function will increment ref. count on file by 1.
360 * In any case returned fd MAY BE not valid!
361 * This race condition is unavoidable
362 * with shared fd spaces, we cannot solve it inside kernel,
363 * but we take care of internal coherence yet.
366 static int sock_alloc_fd(struct file
**filep
)
370 fd
= get_unused_fd();
371 if (likely(fd
>= 0)) {
372 struct file
*file
= get_empty_filp();
375 if (unlikely(!file
)) {
384 static int sock_attach_fd(struct socket
*sock
, struct file
*file
)
389 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
391 this.hash
= SOCK_INODE(sock
)->i_ino
;
393 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
394 if (unlikely(!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 file
->private_data
= sock
;
412 int sock_map_fd(struct socket
*sock
)
414 struct file
*newfile
;
415 int fd
= sock_alloc_fd(&newfile
);
417 if (likely(fd
>= 0)) {
418 int err
= sock_attach_fd(sock
, newfile
);
420 if (unlikely(err
< 0)) {
425 fd_install(fd
, newfile
);
430 static struct socket
*sock_from_file(struct file
*file
, int *err
)
435 if (file
->f_op
== &socket_file_ops
)
436 return file
->private_data
; /* set in sock_map_fd */
438 inode
= file
->f_dentry
->d_inode
;
439 if (!S_ISSOCK(inode
->i_mode
)) {
444 sock
= SOCKET_I(inode
);
445 if (sock
->file
!= file
) {
446 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
453 * sockfd_lookup - Go from a file number to its socket slot
455 * @err: pointer to an error code return
457 * The file handle passed in is locked and the socket it is bound
458 * too is returned. If an error occurs the err pointer is overwritten
459 * with a negative errno code and NULL is returned. The function checks
460 * for both invalid handles and passing a handle which is not a socket.
462 * On a success the socket object pointer is returned.
465 struct socket
*sockfd_lookup(int fd
, int *err
)
470 if (!(file
= fget(fd
))) {
474 sock
= sock_from_file(file
, err
);
480 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
485 file
= fget_light(fd
, fput_needed
);
487 sock
= sock_from_file(file
, err
);
490 fput_light(file
, *fput_needed
);
496 * sock_alloc - allocate a socket
498 * Allocate a new inode and socket object. The two are bound together
499 * and initialised. The socket is then returned. If we are out of inodes
503 static struct socket
*sock_alloc(void)
505 struct inode
* inode
;
506 struct socket
* sock
;
508 inode
= new_inode(sock_mnt
->mnt_sb
);
512 sock
= SOCKET_I(inode
);
514 inode
->i_mode
= S_IFSOCK
|S_IRWXUGO
;
515 inode
->i_uid
= current
->fsuid
;
516 inode
->i_gid
= current
->fsgid
;
518 get_cpu_var(sockets_in_use
)++;
519 put_cpu_var(sockets_in_use
);
524 * In theory you can't get an open on this inode, but /proc provides
525 * a back door. Remember to keep it shut otherwise you'll let the
526 * creepy crawlies in.
529 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
534 struct file_operations bad_sock_fops
= {
535 .owner
= THIS_MODULE
,
536 .open
= sock_no_open
,
540 * sock_release - close a socket
541 * @sock: socket to close
543 * The socket is released from the protocol stack if it has a release
544 * callback, and the inode is then released if the socket is bound to
545 * an inode not a file.
548 void sock_release(struct socket
*sock
)
551 struct module
*owner
= sock
->ops
->owner
;
553 sock
->ops
->release(sock
);
558 if (sock
->fasync_list
)
559 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
561 get_cpu_var(sockets_in_use
)--;
562 put_cpu_var(sockets_in_use
);
564 iput(SOCK_INODE(sock
));
570 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
571 struct msghdr
*msg
, size_t size
)
573 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
581 err
= security_socket_sendmsg(sock
, msg
, size
);
585 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
588 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
591 struct sock_iocb siocb
;
594 init_sync_kiocb(&iocb
, NULL
);
595 iocb
.private = &siocb
;
596 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
597 if (-EIOCBQUEUED
== ret
)
598 ret
= wait_on_sync_kiocb(&iocb
);
602 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
603 struct kvec
*vec
, size_t num
, size_t size
)
605 mm_segment_t oldfs
= get_fs();
610 * the following is safe, since for compiler definitions of kvec and
611 * iovec are identical, yielding the same in-core layout and alignment
613 msg
->msg_iov
= (struct iovec
*)vec
,
614 msg
->msg_iovlen
= num
;
615 result
= sock_sendmsg(sock
, msg
, size
);
620 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
621 struct msghdr
*msg
, size_t size
, int flags
)
624 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
632 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
636 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
639 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
640 size_t size
, int flags
)
643 struct sock_iocb siocb
;
646 init_sync_kiocb(&iocb
, NULL
);
647 iocb
.private = &siocb
;
648 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
649 if (-EIOCBQUEUED
== ret
)
650 ret
= wait_on_sync_kiocb(&iocb
);
654 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
655 struct kvec
*vec
, size_t num
,
656 size_t size
, int flags
)
658 mm_segment_t oldfs
= get_fs();
663 * the following is safe, since for compiler definitions of kvec and
664 * iovec are identical, yielding the same in-core layout and alignment
666 msg
->msg_iov
= (struct iovec
*)vec
,
667 msg
->msg_iovlen
= num
;
668 result
= sock_recvmsg(sock
, msg
, size
, flags
);
673 static void sock_aio_dtor(struct kiocb
*iocb
)
675 kfree(iocb
->private);
678 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
679 int offset
, size_t size
, loff_t
*ppos
, int more
)
684 sock
= file
->private_data
;
686 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
690 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
693 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
694 char __user
*ubuf
, size_t size
, struct sock_iocb
*siocb
)
696 if (!is_sync_kiocb(iocb
)) {
697 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
700 iocb
->ki_dtor
= sock_aio_dtor
;
704 siocb
->async_iov
.iov_base
= ubuf
;
705 siocb
->async_iov
.iov_len
= size
;
707 iocb
->private = siocb
;
711 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
712 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
714 struct socket
*sock
= file
->private_data
;
718 for (i
= 0 ; i
< nr_segs
; i
++)
719 size
+= iov
[i
].iov_len
;
721 msg
->msg_name
= NULL
;
722 msg
->msg_namelen
= 0;
723 msg
->msg_control
= NULL
;
724 msg
->msg_controllen
= 0;
725 msg
->msg_iov
= (struct iovec
*) iov
;
726 msg
->msg_iovlen
= nr_segs
;
727 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
729 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
732 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
733 unsigned long nr_segs
, loff_t
*ppos
)
736 struct sock_iocb siocb
;
740 init_sync_kiocb(&iocb
, NULL
);
741 iocb
.private = &siocb
;
743 ret
= do_sock_read(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
744 if (-EIOCBQUEUED
== ret
)
745 ret
= wait_on_sync_kiocb(&iocb
);
749 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
750 size_t count
, loff_t pos
)
752 struct sock_iocb siocb
, *x
;
756 if (count
== 0) /* Match SYS5 behaviour */
759 x
= alloc_sock_iocb(iocb
, ubuf
, count
, &siocb
);
762 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
,
766 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
767 struct file
*file
, struct iovec
*iov
, unsigned long nr_segs
)
769 struct socket
*sock
= file
->private_data
;
773 for (i
= 0 ; i
< nr_segs
; i
++)
774 size
+= iov
[i
].iov_len
;
776 msg
->msg_name
= NULL
;
777 msg
->msg_namelen
= 0;
778 msg
->msg_control
= NULL
;
779 msg
->msg_controllen
= 0;
780 msg
->msg_iov
= (struct iovec
*) iov
;
781 msg
->msg_iovlen
= nr_segs
;
782 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
783 if (sock
->type
== SOCK_SEQPACKET
)
784 msg
->msg_flags
|= MSG_EOR
;
786 return __sock_sendmsg(iocb
, sock
, msg
, size
);
789 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
790 unsigned long nr_segs
, loff_t
*ppos
)
794 struct sock_iocb siocb
;
797 init_sync_kiocb(&iocb
, NULL
);
798 iocb
.private = &siocb
;
800 ret
= do_sock_write(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
801 if (-EIOCBQUEUED
== ret
)
802 ret
= wait_on_sync_kiocb(&iocb
);
806 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
807 size_t count
, loff_t pos
)
809 struct sock_iocb siocb
, *x
;
813 if (count
== 0) /* Match SYS5 behaviour */
816 x
= alloc_sock_iocb(iocb
, (void __user
*)ubuf
, count
, &siocb
);
820 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
,
826 * Atomic setting of ioctl hooks to avoid race
827 * with module unload.
830 static DEFINE_MUTEX(br_ioctl_mutex
);
831 static int (*br_ioctl_hook
)(unsigned int cmd
, void __user
*arg
) = NULL
;
833 void brioctl_set(int (*hook
)(unsigned int, void __user
*))
835 mutex_lock(&br_ioctl_mutex
);
836 br_ioctl_hook
= hook
;
837 mutex_unlock(&br_ioctl_mutex
);
839 EXPORT_SYMBOL(brioctl_set
);
841 static DEFINE_MUTEX(vlan_ioctl_mutex
);
842 static int (*vlan_ioctl_hook
)(void __user
*arg
);
844 void vlan_ioctl_set(int (*hook
)(void __user
*))
846 mutex_lock(&vlan_ioctl_mutex
);
847 vlan_ioctl_hook
= hook
;
848 mutex_unlock(&vlan_ioctl_mutex
);
850 EXPORT_SYMBOL(vlan_ioctl_set
);
852 static DEFINE_MUTEX(dlci_ioctl_mutex
);
853 static int (*dlci_ioctl_hook
)(unsigned int, void __user
*);
855 void dlci_ioctl_set(int (*hook
)(unsigned int, void __user
*))
857 mutex_lock(&dlci_ioctl_mutex
);
858 dlci_ioctl_hook
= hook
;
859 mutex_unlock(&dlci_ioctl_mutex
);
861 EXPORT_SYMBOL(dlci_ioctl_set
);
864 * With an ioctl, arg may well be a user mode pointer, but we don't know
865 * what to do with it - that's up to the protocol still.
868 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
871 void __user
*argp
= (void __user
*)arg
;
874 sock
= file
->private_data
;
875 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
876 err
= dev_ioctl(cmd
, argp
);
878 #ifdef CONFIG_WIRELESS_EXT
879 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
880 err
= dev_ioctl(cmd
, argp
);
882 #endif /* CONFIG_WIRELESS_EXT */
887 if (get_user(pid
, (int __user
*)argp
))
889 err
= f_setown(sock
->file
, pid
, 1);
893 err
= put_user(sock
->file
->f_owner
.pid
, (int __user
*)argp
);
901 request_module("bridge");
903 mutex_lock(&br_ioctl_mutex
);
905 err
= br_ioctl_hook(cmd
, argp
);
906 mutex_unlock(&br_ioctl_mutex
);
911 if (!vlan_ioctl_hook
)
912 request_module("8021q");
914 mutex_lock(&vlan_ioctl_mutex
);
916 err
= vlan_ioctl_hook(argp
);
917 mutex_unlock(&vlan_ioctl_mutex
);
921 /* Convert this to call through a hook */
922 err
= divert_ioctl(cmd
, argp
);
927 if (!dlci_ioctl_hook
)
928 request_module("dlci");
930 if (dlci_ioctl_hook
) {
931 mutex_lock(&dlci_ioctl_mutex
);
932 err
= dlci_ioctl_hook(cmd
, argp
);
933 mutex_unlock(&dlci_ioctl_mutex
);
937 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
940 * If this ioctl is unknown try to hand it down
943 if (err
== -ENOIOCTLCMD
)
944 err
= dev_ioctl(cmd
, argp
);
950 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
953 struct socket
*sock
= NULL
;
955 err
= security_socket_create(family
, type
, protocol
, 1);
965 security_socket_post_create(sock
, family
, type
, protocol
, 1);
972 /* No kernel lock held - perfect */
973 static unsigned int sock_poll(struct file
*file
, poll_table
* wait
)
978 * We can't return errors to poll, so it's either yes or no.
980 sock
= file
->private_data
;
981 return sock
->ops
->poll(file
, sock
, wait
);
984 static int sock_mmap(struct file
* file
, struct vm_area_struct
* vma
)
986 struct socket
*sock
= file
->private_data
;
988 return sock
->ops
->mmap(file
, sock
, vma
);
991 static int sock_close(struct inode
*inode
, struct file
*filp
)
994 * It was possible the inode is NULL we were
995 * closing an unfinished socket.
1000 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1003 sock_fasync(-1, filp
, 0);
1004 sock_release(SOCKET_I(inode
));
1009 * Update the socket async list
1011 * Fasync_list locking strategy.
1013 * 1. fasync_list is modified only under process context socket lock
1014 * i.e. under semaphore.
1015 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1016 * or under socket lock.
1017 * 3. fasync_list can be used from softirq context, so that
1018 * modification under socket lock have to be enhanced with
1019 * write_lock_bh(&sk->sk_callback_lock).
1023 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1025 struct fasync_struct
*fa
, *fna
=NULL
, **prev
;
1026 struct socket
*sock
;
1031 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1036 sock
= filp
->private_data
;
1038 if ((sk
=sock
->sk
) == NULL
) {
1045 prev
=&(sock
->fasync_list
);
1047 for (fa
=*prev
; fa
!=NULL
; prev
=&fa
->fa_next
,fa
=*prev
)
1048 if (fa
->fa_file
==filp
)
1055 write_lock_bh(&sk
->sk_callback_lock
);
1057 write_unlock_bh(&sk
->sk_callback_lock
);
1064 fna
->magic
=FASYNC_MAGIC
;
1065 fna
->fa_next
=sock
->fasync_list
;
1066 write_lock_bh(&sk
->sk_callback_lock
);
1067 sock
->fasync_list
=fna
;
1068 write_unlock_bh(&sk
->sk_callback_lock
);
1074 write_lock_bh(&sk
->sk_callback_lock
);
1076 write_unlock_bh(&sk
->sk_callback_lock
);
1082 release_sock(sock
->sk
);
1086 /* This function may be called only under socket lock or callback_lock */
1088 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1090 if (!sock
|| !sock
->fasync_list
)
1096 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1100 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1105 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1108 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1113 static int __sock_create(int family
, int type
, int protocol
, struct socket
**res
, int kern
)
1116 struct socket
*sock
;
1119 * Check protocol is in range
1121 if (family
< 0 || family
>= NPROTO
)
1122 return -EAFNOSUPPORT
;
1123 if (type
< 0 || type
>= SOCK_MAX
)
1128 This uglymoron is moved from INET layer to here to avoid
1129 deadlock in module load.
1131 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1135 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n", current
->comm
);
1140 err
= security_socket_create(family
, type
, protocol
, kern
);
1144 #if defined(CONFIG_KMOD)
1145 /* Attempt to load a protocol module if the find failed.
1147 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1148 * requested real, full-featured networking support upon configuration.
1149 * Otherwise module support will break!
1151 if (net_families
[family
]==NULL
)
1153 request_module("net-pf-%d",family
);
1157 net_family_read_lock();
1158 if (net_families
[family
] == NULL
) {
1159 err
= -EAFNOSUPPORT
;
1164 * Allocate the socket and allow the family to set things up. if
1165 * the protocol is 0, the family is instructed to select an appropriate
1169 if (!(sock
= sock_alloc())) {
1170 printk(KERN_WARNING
"socket: no more sockets\n");
1171 err
= -ENFILE
; /* Not exactly a match, but its the
1172 closest posix thing */
1179 * We will call the ->create function, that possibly is in a loadable
1180 * module, so we have to bump that loadable module refcnt first.
1182 err
= -EAFNOSUPPORT
;
1183 if (!try_module_get(net_families
[family
]->owner
))
1186 if ((err
= net_families
[family
]->create(sock
, protocol
)) < 0) {
1188 goto out_module_put
;
1192 * Now to bump the refcnt of the [loadable] module that owns this
1193 * socket at sock_release time we decrement its refcnt.
1195 if (!try_module_get(sock
->ops
->owner
)) {
1197 goto out_module_put
;
1200 * Now that we're done with the ->create function, the [loadable]
1201 * module can have its refcnt decremented
1203 module_put(net_families
[family
]->owner
);
1205 security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1208 net_family_read_unlock();
1211 module_put(net_families
[family
]->owner
);
1217 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1219 return __sock_create(family
, type
, protocol
, res
, 0);
1222 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1224 return __sock_create(family
, type
, protocol
, res
, 1);
1227 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1230 struct socket
*sock
;
1232 retval
= sock_create(family
, type
, protocol
, &sock
);
1236 retval
= sock_map_fd(sock
);
1241 /* It may be already another descriptor 8) Not kernel problem. */
1250 * Create a pair of connected sockets.
1253 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1255 struct socket
*sock1
, *sock2
;
1259 * Obtain the first socket and check if the underlying protocol
1260 * supports the socketpair call.
1263 err
= sock_create(family
, type
, protocol
, &sock1
);
1267 err
= sock_create(family
, type
, protocol
, &sock2
);
1271 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1273 goto out_release_both
;
1277 err
= sock_map_fd(sock1
);
1279 goto out_release_both
;
1282 err
= sock_map_fd(sock2
);
1287 /* fd1 and fd2 may be already another descriptors.
1288 * Not kernel problem.
1291 err
= put_user(fd1
, &usockvec
[0]);
1293 err
= put_user(fd2
, &usockvec
[1]);
1302 sock_release(sock2
);
1307 sock_release(sock2
);
1309 sock_release(sock1
);
1316 * Bind a name to a socket. Nothing much to do here since it's
1317 * the protocol's responsibility to handle the local address.
1319 * We move the socket address to kernel space before we call
1320 * the protocol layer (having also checked the address is ok).
1323 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1325 struct socket
*sock
;
1326 char address
[MAX_SOCK_ADDR
];
1327 int err
, fput_needed
;
1329 if((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
))!=NULL
)
1331 if((err
=move_addr_to_kernel(umyaddr
,addrlen
,address
))>=0) {
1332 err
= security_socket_bind(sock
, (struct sockaddr
*)address
, addrlen
);
1334 err
= sock
->ops
->bind(sock
,
1335 (struct sockaddr
*)address
, addrlen
);
1337 fput_light(sock
->file
, fput_needed
);
1344 * Perform a listen. Basically, we allow the protocol to do anything
1345 * necessary for a listen, and if that works, we mark the socket as
1346 * ready for listening.
1349 int sysctl_somaxconn
= SOMAXCONN
;
1351 asmlinkage
long sys_listen(int fd
, int backlog
)
1353 struct socket
*sock
;
1354 int err
, fput_needed
;
1356 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1357 if ((unsigned) backlog
> sysctl_somaxconn
)
1358 backlog
= sysctl_somaxconn
;
1360 err
= security_socket_listen(sock
, backlog
);
1362 err
= sock
->ops
->listen(sock
, backlog
);
1364 fput_light(sock
->file
, fput_needed
);
1371 * For accept, we attempt to create a new socket, set up the link
1372 * with the client, wake up the client, then return the new
1373 * connected fd. We collect the address of the connector in kernel
1374 * space and move it to user at the very end. This is unclean because
1375 * we open the socket then return an error.
1377 * 1003.1g adds the ability to recvmsg() to query connection pending
1378 * status to recvmsg. We need to add that support in a way thats
1379 * clean when we restucture accept also.
1382 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
, int __user
*upeer_addrlen
)
1384 struct socket
*sock
, *newsock
;
1385 struct file
*newfile
;
1386 int err
, len
, newfd
, fput_needed
;
1387 char address
[MAX_SOCK_ADDR
];
1389 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1394 if (!(newsock
= sock_alloc()))
1397 newsock
->type
= sock
->type
;
1398 newsock
->ops
= sock
->ops
;
1401 * We don't need try_module_get here, as the listening socket (sock)
1402 * has the protocol module (sock->ops->owner) held.
1404 __module_get(newsock
->ops
->owner
);
1406 newfd
= sock_alloc_fd(&newfile
);
1407 if (unlikely(newfd
< 0)) {
1412 err
= sock_attach_fd(newsock
, newfile
);
1416 err
= security_socket_accept(sock
, newsock
);
1420 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1424 if (upeer_sockaddr
) {
1425 if(newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
, &len
, 2)<0) {
1426 err
= -ECONNABORTED
;
1429 err
= move_addr_to_user(address
, len
, upeer_sockaddr
, upeer_addrlen
);
1434 /* File flags are not inherited via accept() unlike another OSes. */
1436 fd_install(newfd
, newfile
);
1439 security_socket_post_accept(sock
, newsock
);
1442 fput_light(sock
->file
, fput_needed
);
1447 put_unused_fd(newfd
);
1449 sock_release(newsock
);
1455 * Attempt to connect to a socket with the server address. The address
1456 * is in user space so we verify it is OK and move it to kernel space.
1458 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1461 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1462 * other SEQPACKET protocols that take time to connect() as it doesn't
1463 * include the -EINPROGRESS status for such sockets.
1466 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1468 struct socket
*sock
;
1469 char address
[MAX_SOCK_ADDR
];
1470 int err
, fput_needed
;
1472 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1475 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1479 err
= security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1483 err
= sock
->ops
->connect(sock
, (struct sockaddr
*) address
, addrlen
,
1484 sock
->file
->f_flags
);
1486 fput_light(sock
->file
, fput_needed
);
1492 * Get the local address ('name') of a socket object. Move the obtained
1493 * name to user space.
1496 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1498 struct socket
*sock
;
1499 char address
[MAX_SOCK_ADDR
];
1500 int len
, err
, fput_needed
;
1502 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1506 err
= security_socket_getsockname(sock
);
1510 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1513 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1516 fput_light(sock
->file
, fput_needed
);
1522 * Get the remote address ('name') of a socket object. Move the obtained
1523 * name to user space.
1526 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
, int __user
*usockaddr_len
)
1528 struct socket
*sock
;
1529 char address
[MAX_SOCK_ADDR
];
1530 int len
, err
, fput_needed
;
1532 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1533 err
= security_socket_getpeername(sock
);
1535 fput_light(sock
->file
, fput_needed
);
1539 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 1);
1541 err
=move_addr_to_user(address
,len
, usockaddr
, usockaddr_len
);
1542 fput_light(sock
->file
, fput_needed
);
1548 * Send a datagram to a given address. We move the address into kernel
1549 * space and check the user space data area is readable before invoking
1553 asmlinkage
long sys_sendto(int fd
, void __user
* buff
, size_t len
, unsigned flags
,
1554 struct sockaddr __user
*addr
, int addr_len
)
1556 struct socket
*sock
;
1557 char address
[MAX_SOCK_ADDR
];
1562 struct file
*sock_file
;
1564 sock_file
= fget_light(fd
, &fput_needed
);
1568 sock
= sock_from_file(sock_file
, &err
);
1576 msg
.msg_control
=NULL
;
1577 msg
.msg_controllen
=0;
1580 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1583 msg
.msg_name
=address
;
1584 msg
.msg_namelen
=addr_len
;
1586 if (sock
->file
->f_flags
& O_NONBLOCK
)
1587 flags
|= MSG_DONTWAIT
;
1588 msg
.msg_flags
= flags
;
1589 err
= sock_sendmsg(sock
, &msg
, len
);
1592 fput_light(sock_file
, fput_needed
);
1597 * Send a datagram down a socket.
1600 asmlinkage
long sys_send(int fd
, void __user
* buff
, size_t len
, unsigned flags
)
1602 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1606 * Receive a frame from the socket and optionally record the address of the
1607 * sender. We verify the buffers are writable and if needed move the
1608 * sender address from kernel to user space.
1611 asmlinkage
long sys_recvfrom(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
,
1612 struct sockaddr __user
*addr
, int __user
*addr_len
)
1614 struct socket
*sock
;
1617 char address
[MAX_SOCK_ADDR
];
1619 struct file
*sock_file
;
1622 sock_file
= fget_light(fd
, &fput_needed
);
1626 sock
= sock_from_file(sock_file
, &err
);
1630 msg
.msg_control
=NULL
;
1631 msg
.msg_controllen
=0;
1636 msg
.msg_name
=address
;
1637 msg
.msg_namelen
=MAX_SOCK_ADDR
;
1638 if (sock
->file
->f_flags
& O_NONBLOCK
)
1639 flags
|= MSG_DONTWAIT
;
1640 err
=sock_recvmsg(sock
, &msg
, size
, flags
);
1642 if(err
>= 0 && addr
!= NULL
)
1644 err2
=move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1649 fput_light(sock_file
, fput_needed
);
1654 * Receive a datagram from a socket.
1657 asmlinkage
long sys_recv(int fd
, void __user
* ubuf
, size_t size
, unsigned flags
)
1659 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1663 * Set a socket option. Because we don't know the option lengths we have
1664 * to pass the user mode parameter for the protocols to sort out.
1667 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
, char __user
*optval
, int optlen
)
1669 int err
, fput_needed
;
1670 struct socket
*sock
;
1675 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
)
1677 err
= security_socket_setsockopt(sock
,level
,optname
);
1681 if (level
== SOL_SOCKET
)
1682 err
=sock_setsockopt(sock
,level
,optname
,optval
,optlen
);
1684 err
=sock
->ops
->setsockopt(sock
, level
, optname
, optval
, optlen
);
1686 fput_light(sock
->file
, fput_needed
);
1692 * Get a socket option. Because we don't know the option lengths we have
1693 * to pass a user mode parameter for the protocols to sort out.
1696 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
, int __user
*optlen
)
1698 int err
, fput_needed
;
1699 struct socket
*sock
;
1701 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
)) != NULL
) {
1702 err
= security_socket_getsockopt(sock
, level
, optname
);
1706 if (level
== SOL_SOCKET
)
1707 err
=sock_getsockopt(sock
,level
,optname
,optval
,optlen
);
1709 err
=sock
->ops
->getsockopt(sock
, level
, optname
, optval
, optlen
);
1711 fput_light(sock
->file
, fput_needed
);
1718 * Shutdown a socket.
1721 asmlinkage
long sys_shutdown(int fd
, int how
)
1723 int err
, fput_needed
;
1724 struct socket
*sock
;
1726 if ((sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
))!=NULL
)
1728 err
= security_socket_shutdown(sock
, how
);
1730 err
= sock
->ops
->shutdown(sock
, how
);
1731 fput_light(sock
->file
, fput_needed
);
1736 /* A couple of helpful macros for getting the address of the 32/64 bit
1737 * fields which are the same type (int / unsigned) on our platforms.
1739 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1740 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1741 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1745 * BSD sendmsg interface
1748 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1750 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1751 struct socket
*sock
;
1752 char address
[MAX_SOCK_ADDR
];
1753 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1754 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1755 __attribute__ ((aligned (sizeof(__kernel_size_t
))));
1756 /* 20 is size of ipv6_pktinfo */
1757 unsigned char *ctl_buf
= ctl
;
1758 struct msghdr msg_sys
;
1759 int err
, ctl_len
, iov_size
, total_len
;
1763 if (MSG_CMSG_COMPAT
& flags
) {
1764 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1766 } else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1769 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1773 /* do not move before msg_sys is valid */
1775 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1778 /* Check whether to allocate the iovec area*/
1780 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1781 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1782 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1787 /* This will also move the address data into kernel space */
1788 if (MSG_CMSG_COMPAT
& flags
) {
1789 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1791 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1798 if (msg_sys
.msg_controllen
> INT_MAX
)
1800 ctl_len
= msg_sys
.msg_controllen
;
1801 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1802 err
= cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
, sizeof(ctl
));
1805 ctl_buf
= msg_sys
.msg_control
;
1806 ctl_len
= msg_sys
.msg_controllen
;
1807 } else if (ctl_len
) {
1808 if (ctl_len
> sizeof(ctl
))
1810 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1811 if (ctl_buf
== NULL
)
1816 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1817 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1818 * checking falls down on this.
1820 if (copy_from_user(ctl_buf
, (void __user
*) msg_sys
.msg_control
, ctl_len
))
1822 msg_sys
.msg_control
= ctl_buf
;
1824 msg_sys
.msg_flags
= flags
;
1826 if (sock
->file
->f_flags
& O_NONBLOCK
)
1827 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1828 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1832 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1834 if (iov
!= iovstack
)
1835 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1837 fput_light(sock
->file
, fput_needed
);
1843 * BSD recvmsg interface
1846 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned int flags
)
1848 struct compat_msghdr __user
*msg_compat
= (struct compat_msghdr __user
*)msg
;
1849 struct socket
*sock
;
1850 struct iovec iovstack
[UIO_FASTIOV
];
1851 struct iovec
*iov
=iovstack
;
1852 struct msghdr msg_sys
;
1853 unsigned long cmsg_ptr
;
1854 int err
, iov_size
, total_len
, len
;
1857 /* kernel mode address */
1858 char addr
[MAX_SOCK_ADDR
];
1860 /* user mode address pointers */
1861 struct sockaddr __user
*uaddr
;
1862 int __user
*uaddr_len
;
1864 if (MSG_CMSG_COMPAT
& flags
) {
1865 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1868 if (copy_from_user(&msg_sys
,msg
,sizeof(struct msghdr
)))
1871 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1876 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1879 /* Check whether to allocate the iovec area*/
1881 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1882 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1883 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1889 * Save the user-mode address (verify_iovec will change the
1890 * kernel msghdr to use the kernel address space)
1893 uaddr
= (void __user
*) msg_sys
.msg_name
;
1894 uaddr_len
= COMPAT_NAMELEN(msg
);
1895 if (MSG_CMSG_COMPAT
& flags
) {
1896 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1898 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1903 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1904 msg_sys
.msg_flags
= 0;
1905 if (MSG_CMSG_COMPAT
& flags
)
1906 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1908 if (sock
->file
->f_flags
& O_NONBLOCK
)
1909 flags
|= MSG_DONTWAIT
;
1910 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1915 if (uaddr
!= NULL
) {
1916 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
, uaddr_len
);
1920 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1924 if (MSG_CMSG_COMPAT
& flags
)
1925 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1926 &msg_compat
->msg_controllen
);
1928 err
= __put_user((unsigned long)msg_sys
.msg_control
-cmsg_ptr
,
1929 &msg
->msg_controllen
);
1935 if (iov
!= iovstack
)
1936 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1938 fput_light(sock
->file
, fput_needed
);
1943 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1945 /* Argument list sizes for sys_socketcall */
1946 #define AL(x) ((x) * sizeof(unsigned long))
1947 static unsigned char nargs
[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1948 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1949 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1953 * System call vectors.
1955 * Argument checking cleaned up. Saved 20% in size.
1956 * This function doesn't need to set the kernel lock because
1957 * it is set by the callees.
1960 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1963 unsigned long a0
,a1
;
1966 if(call
<1||call
>SYS_RECVMSG
)
1969 /* copy_from_user should be SMP safe. */
1970 if (copy_from_user(a
, args
, nargs
[call
]))
1973 err
= audit_socketcall(nargs
[call
]/sizeof(unsigned long), a
);
1983 err
= sys_socket(a0
,a1
,a
[2]);
1986 err
= sys_bind(a0
,(struct sockaddr __user
*)a1
, a
[2]);
1989 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
1992 err
= sys_listen(a0
,a1
);
1995 err
= sys_accept(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
1997 case SYS_GETSOCKNAME
:
1998 err
= sys_getsockname(a0
,(struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
2000 case SYS_GETPEERNAME
:
2001 err
= sys_getpeername(a0
, (struct sockaddr __user
*)a1
, (int __user
*)a
[2]);
2003 case SYS_SOCKETPAIR
:
2004 err
= sys_socketpair(a0
,a1
, a
[2], (int __user
*)a
[3]);
2007 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2010 err
= sys_sendto(a0
,(void __user
*)a1
, a
[2], a
[3],
2011 (struct sockaddr __user
*)a
[4], a
[5]);
2014 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2017 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2018 (struct sockaddr __user
*)a
[4], (int __user
*)a
[5]);
2021 err
= sys_shutdown(a0
,a1
);
2023 case SYS_SETSOCKOPT
:
2024 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2026 case SYS_GETSOCKOPT
:
2027 err
= sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], (int __user
*)a
[4]);
2030 err
= sys_sendmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
2033 err
= sys_recvmsg(a0
, (struct msghdr __user
*) a1
, a
[2]);
2042 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2045 * This function is called by a protocol handler that wants to
2046 * advertise its address family, and have it linked into the
2050 int sock_register(struct net_proto_family
*ops
)
2054 if (ops
->family
>= NPROTO
) {
2055 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2058 net_family_write_lock();
2060 if (net_families
[ops
->family
] == NULL
) {
2061 net_families
[ops
->family
]=ops
;
2064 net_family_write_unlock();
2065 printk(KERN_INFO
"NET: Registered protocol family %d\n",
2071 * This function is called by a protocol handler that wants to
2072 * remove its address family, and have it unlinked from the
2076 int sock_unregister(int family
)
2078 if (family
< 0 || family
>= NPROTO
)
2081 net_family_write_lock();
2082 net_families
[family
]=NULL
;
2083 net_family_write_unlock();
2084 printk(KERN_INFO
"NET: Unregistered protocol family %d\n",
2089 static int __init
sock_init(void)
2092 * Initialize sock SLAB cache.
2098 * Initialize skbuff SLAB cache
2103 * Initialize the protocols module.
2107 register_filesystem(&sock_fs_type
);
2108 sock_mnt
= kern_mount(&sock_fs_type
);
2110 /* The real protocol initialization is performed in later initcalls.
2113 #ifdef CONFIG_NETFILTER
2120 core_initcall(sock_init
); /* early initcall */
2122 #ifdef CONFIG_PROC_FS
2123 void socket_seq_show(struct seq_file
*seq
)
2129 counter
+= per_cpu(sockets_in_use
, cpu
);
2131 /* It can be negative, by the way. 8) */
2135 seq_printf(seq
, "sockets: used %d\n", counter
);
2137 #endif /* CONFIG_PROC_FS */
2139 /* ABI emulation layers need these two */
2140 EXPORT_SYMBOL(move_addr_to_kernel
);
2141 EXPORT_SYMBOL(move_addr_to_user
);
2142 EXPORT_SYMBOL(sock_create
);
2143 EXPORT_SYMBOL(sock_create_kern
);
2144 EXPORT_SYMBOL(sock_create_lite
);
2145 EXPORT_SYMBOL(sock_map_fd
);
2146 EXPORT_SYMBOL(sock_recvmsg
);
2147 EXPORT_SYMBOL(sock_register
);
2148 EXPORT_SYMBOL(sock_release
);
2149 EXPORT_SYMBOL(sock_sendmsg
);
2150 EXPORT_SYMBOL(sock_unregister
);
2151 EXPORT_SYMBOL(sock_wake_async
);
2152 EXPORT_SYMBOL(sockfd_lookup
);
2153 EXPORT_SYMBOL(kernel_sendmsg
);
2154 EXPORT_SYMBOL(kernel_recvmsg
);