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
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/rcupdate.h>
67 #include <linux/netdevice.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/mutex.h>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
88 #include <asm/uaccess.h>
89 #include <asm/unistd.h>
91 #include <net/compat.h>
94 #include <linux/netfilter.h>
96 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
97 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
98 unsigned long nr_segs
, loff_t pos
);
99 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
100 unsigned long nr_segs
, loff_t pos
);
101 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
103 static int sock_close(struct inode
*inode
, struct file
*file
);
104 static unsigned int sock_poll(struct file
*file
,
105 struct poll_table_struct
*wait
);
106 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
108 static long compat_sock_ioctl(struct file
*file
,
109 unsigned int cmd
, unsigned long arg
);
111 static int sock_fasync(int fd
, struct file
*filp
, int on
);
112 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
113 int offset
, size_t size
, loff_t
*ppos
, int more
);
114 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
115 struct pipe_inode_info
*pipe
, size_t len
,
119 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
120 * in the operation structures but are done directly via the socketcall() multiplexor.
123 static const struct file_operations socket_file_ops
= {
124 .owner
= THIS_MODULE
,
126 .aio_read
= sock_aio_read
,
127 .aio_write
= sock_aio_write
,
129 .unlocked_ioctl
= sock_ioctl
,
131 .compat_ioctl
= compat_sock_ioctl
,
134 .open
= sock_no_open
, /* special open code to disallow open via /proc */
135 .release
= sock_close
,
136 .fasync
= sock_fasync
,
137 .sendpage
= sock_sendpage
,
138 .splice_write
= generic_splice_sendpage
,
139 .splice_read
= sock_splice_read
,
143 * The protocol list. Each protocol is registered in here.
146 static DEFINE_SPINLOCK(net_family_lock
);
147 static const struct net_proto_family
*net_families
[NPROTO
] __read_mostly
;
150 * Statistics counters of the socket lists
153 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
157 * Move socket addresses back and forth across the kernel/user
158 * divide and look after the messy bits.
161 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
162 16 for IP, 16 for IPX,
165 must be at least one bigger than
166 the AF_UNIX size (see net/unix/af_unix.c
171 * move_addr_to_kernel - copy a socket address into kernel space
172 * @uaddr: Address in user space
173 * @kaddr: Address in kernel space
174 * @ulen: Length in user space
176 * The address is copied into kernel space. If the provided address is
177 * too long an error code of -EINVAL is returned. If the copy gives
178 * invalid addresses -EFAULT is returned. On a success 0 is returned.
181 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
183 if (ulen
< 0 || ulen
> MAX_SOCK_ADDR
)
187 if (copy_from_user(kaddr
, uaddr
, ulen
))
189 return audit_sockaddr(ulen
, kaddr
);
193 * move_addr_to_user - copy an address to user space
194 * @kaddr: kernel space address
195 * @klen: length of address in kernel
196 * @uaddr: user space address
197 * @ulen: pointer to user length field
199 * The value pointed to by ulen on entry is the buffer length available.
200 * This is overwritten with the buffer space used. -EINVAL is returned
201 * if an overlong buffer is specified or a negative buffer size. -EFAULT
202 * is returned if either the buffer or the length field are not
204 * After copying the data up to the limit the user specifies, the true
205 * length of the data is written over the length limit the user
206 * specified. Zero is returned for a success.
209 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
,
215 err
= get_user(len
, ulen
);
220 if (len
< 0 || len
> MAX_SOCK_ADDR
)
223 if (audit_sockaddr(klen
, kaddr
))
225 if (copy_to_user(uaddr
, kaddr
, len
))
229 * "fromlen shall refer to the value before truncation.."
232 return __put_user(klen
, ulen
);
235 #define SOCKFS_MAGIC 0x534F434B
237 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
239 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
241 struct socket_alloc
*ei
;
243 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
246 init_waitqueue_head(&ei
->socket
.wait
);
248 ei
->socket
.fasync_list
= NULL
;
249 ei
->socket
.state
= SS_UNCONNECTED
;
250 ei
->socket
.flags
= 0;
251 ei
->socket
.ops
= NULL
;
252 ei
->socket
.sk
= NULL
;
253 ei
->socket
.file
= NULL
;
255 return &ei
->vfs_inode
;
258 static void sock_destroy_inode(struct inode
*inode
)
260 kmem_cache_free(sock_inode_cachep
,
261 container_of(inode
, struct socket_alloc
, vfs_inode
));
264 static void init_once(void *foo
, struct kmem_cache
*cachep
, unsigned long flags
)
266 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
268 inode_init_once(&ei
->vfs_inode
);
271 static int init_inodecache(void)
273 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
274 sizeof(struct socket_alloc
),
276 (SLAB_HWCACHE_ALIGN
|
277 SLAB_RECLAIM_ACCOUNT
|
281 if (sock_inode_cachep
== NULL
)
286 static struct super_operations sockfs_ops
= {
287 .alloc_inode
= sock_alloc_inode
,
288 .destroy_inode
=sock_destroy_inode
,
289 .statfs
= simple_statfs
,
292 static int sockfs_get_sb(struct file_system_type
*fs_type
,
293 int flags
, const char *dev_name
, void *data
,
294 struct vfsmount
*mnt
)
296 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
300 static struct vfsmount
*sock_mnt __read_mostly
;
302 static struct file_system_type sock_fs_type
= {
304 .get_sb
= sockfs_get_sb
,
305 .kill_sb
= kill_anon_super
,
308 static int sockfs_delete_dentry(struct dentry
*dentry
)
311 * At creation time, we pretended this dentry was hashed
312 * (by clearing DCACHE_UNHASHED bit in d_flags)
313 * At delete time, we restore the truth : not hashed.
314 * (so that dput() can proceed correctly)
316 dentry
->d_flags
|= DCACHE_UNHASHED
;
321 * sockfs_dname() is called from d_path().
323 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
325 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
326 dentry
->d_inode
->i_ino
);
329 static struct dentry_operations sockfs_dentry_operations
= {
330 .d_delete
= sockfs_delete_dentry
,
331 .d_dname
= sockfs_dname
,
335 * Obtains the first available file descriptor and sets it up for use.
337 * These functions create file structures and maps them to fd space
338 * of the current process. On success it returns file descriptor
339 * and file struct implicitly stored in sock->file.
340 * Note that another thread may close file descriptor before we return
341 * from this function. We use the fact that now we do not refer
342 * to socket after mapping. If one day we will need it, this
343 * function will increment ref. count on file by 1.
345 * In any case returned fd MAY BE not valid!
346 * This race condition is unavoidable
347 * with shared fd spaces, we cannot solve it inside kernel,
348 * but we take care of internal coherence yet.
351 static int sock_alloc_fd(struct file
**filep
)
355 fd
= get_unused_fd();
356 if (likely(fd
>= 0)) {
357 struct file
*file
= get_empty_filp();
360 if (unlikely(!file
)) {
369 static int sock_attach_fd(struct socket
*sock
, struct file
*file
)
371 struct qstr name
= { .name
= "" };
373 file
->f_path
.dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &name
);
374 if (unlikely(!file
->f_path
.dentry
))
377 file
->f_path
.dentry
->d_op
= &sockfs_dentry_operations
;
379 * We dont want to push this dentry into global dentry hash table.
380 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
381 * This permits a working /proc/$pid/fd/XXX on sockets
383 file
->f_path
.dentry
->d_flags
&= ~DCACHE_UNHASHED
;
384 d_instantiate(file
->f_path
.dentry
, SOCK_INODE(sock
));
385 file
->f_path
.mnt
= mntget(sock_mnt
);
386 file
->f_mapping
= file
->f_path
.dentry
->d_inode
->i_mapping
;
389 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
390 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
391 file
->f_flags
= O_RDWR
;
393 file
->private_data
= sock
;
398 int sock_map_fd(struct socket
*sock
)
400 struct file
*newfile
;
401 int fd
= sock_alloc_fd(&newfile
);
403 if (likely(fd
>= 0)) {
404 int err
= sock_attach_fd(sock
, newfile
);
406 if (unlikely(err
< 0)) {
411 fd_install(fd
, newfile
);
416 static struct socket
*sock_from_file(struct file
*file
, int *err
)
418 if (file
->f_op
== &socket_file_ops
)
419 return file
->private_data
; /* set in sock_map_fd */
426 * sockfd_lookup - Go from a file number to its socket slot
428 * @err: pointer to an error code return
430 * The file handle passed in is locked and the socket it is bound
431 * too is returned. If an error occurs the err pointer is overwritten
432 * with a negative errno code and NULL is returned. The function checks
433 * for both invalid handles and passing a handle which is not a socket.
435 * On a success the socket object pointer is returned.
438 struct socket
*sockfd_lookup(int fd
, int *err
)
449 sock
= sock_from_file(file
, err
);
455 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
461 file
= fget_light(fd
, fput_needed
);
463 sock
= sock_from_file(file
, err
);
466 fput_light(file
, *fput_needed
);
472 * sock_alloc - allocate a socket
474 * Allocate a new inode and socket object. The two are bound together
475 * and initialised. The socket is then returned. If we are out of inodes
479 static struct socket
*sock_alloc(void)
484 inode
= new_inode(sock_mnt
->mnt_sb
);
488 sock
= SOCKET_I(inode
);
490 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
491 inode
->i_uid
= current
->fsuid
;
492 inode
->i_gid
= current
->fsgid
;
494 get_cpu_var(sockets_in_use
)++;
495 put_cpu_var(sockets_in_use
);
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
505 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
510 const struct file_operations bad_sock_fops
= {
511 .owner
= THIS_MODULE
,
512 .open
= sock_no_open
,
516 * sock_release - close a socket
517 * @sock: socket to close
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
521 * an inode not a file.
524 void sock_release(struct socket
*sock
)
527 struct module
*owner
= sock
->ops
->owner
;
529 sock
->ops
->release(sock
);
534 if (sock
->fasync_list
)
535 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
537 get_cpu_var(sockets_in_use
)--;
538 put_cpu_var(sockets_in_use
);
540 iput(SOCK_INODE(sock
));
546 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
547 struct msghdr
*msg
, size_t size
)
549 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
557 err
= security_socket_sendmsg(sock
, msg
, size
);
561 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
564 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
567 struct sock_iocb siocb
;
570 init_sync_kiocb(&iocb
, NULL
);
571 iocb
.private = &siocb
;
572 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
573 if (-EIOCBQUEUED
== ret
)
574 ret
= wait_on_sync_kiocb(&iocb
);
578 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
579 struct kvec
*vec
, size_t num
, size_t size
)
581 mm_segment_t oldfs
= get_fs();
586 * the following is safe, since for compiler definitions of kvec and
587 * iovec are identical, yielding the same in-core layout and alignment
589 msg
->msg_iov
= (struct iovec
*)vec
;
590 msg
->msg_iovlen
= num
;
591 result
= sock_sendmsg(sock
, msg
, size
);
597 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
599 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
602 ktime_t kt
= skb
->tstamp
;
604 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
606 /* Race occurred between timestamp enabling and packet
607 receiving. Fill in the current time for now. */
609 kt
= ktime_get_real();
611 tv
= ktime_to_timeval(kt
);
612 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
, sizeof(tv
), &tv
);
615 /* Race occurred between timestamp enabling and packet
616 receiving. Fill in the current time for now. */
618 kt
= ktime_get_real();
620 ts
= ktime_to_timespec(kt
);
621 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
, sizeof(ts
), &ts
);
625 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
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
, size_t size
, int flags
)
664 mm_segment_t oldfs
= get_fs();
669 * the following is safe, since for compiler definitions of kvec and
670 * iovec are identical, yielding the same in-core layout and alignment
672 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
673 result
= sock_recvmsg(sock
, msg
, size
, flags
);
678 static void sock_aio_dtor(struct kiocb
*iocb
)
680 kfree(iocb
->private);
683 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
684 int offset
, size_t size
, loff_t
*ppos
, int more
)
689 sock
= file
->private_data
;
691 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
695 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
698 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
699 struct pipe_inode_info
*pipe
, size_t len
,
702 struct socket
*sock
= file
->private_data
;
704 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
707 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
708 struct sock_iocb
*siocb
)
710 if (!is_sync_kiocb(iocb
)) {
711 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
714 iocb
->ki_dtor
= sock_aio_dtor
;
718 iocb
->private = siocb
;
722 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
723 struct file
*file
, const struct iovec
*iov
,
724 unsigned long nr_segs
)
726 struct socket
*sock
= file
->private_data
;
730 for (i
= 0; i
< nr_segs
; i
++)
731 size
+= iov
[i
].iov_len
;
733 msg
->msg_name
= NULL
;
734 msg
->msg_namelen
= 0;
735 msg
->msg_control
= NULL
;
736 msg
->msg_controllen
= 0;
737 msg
->msg_iov
= (struct iovec
*)iov
;
738 msg
->msg_iovlen
= nr_segs
;
739 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
741 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
744 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
745 unsigned long nr_segs
, loff_t pos
)
747 struct sock_iocb siocb
, *x
;
752 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
756 x
= alloc_sock_iocb(iocb
, &siocb
);
759 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
762 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
763 struct file
*file
, const struct iovec
*iov
,
764 unsigned long nr_segs
)
766 struct socket
*sock
= file
->private_data
;
770 for (i
= 0; i
< nr_segs
; i
++)
771 size
+= iov
[i
].iov_len
;
773 msg
->msg_name
= NULL
;
774 msg
->msg_namelen
= 0;
775 msg
->msg_control
= NULL
;
776 msg
->msg_controllen
= 0;
777 msg
->msg_iov
= (struct iovec
*)iov
;
778 msg
->msg_iovlen
= nr_segs
;
779 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
780 if (sock
->type
== SOCK_SEQPACKET
)
781 msg
->msg_flags
|= MSG_EOR
;
783 return __sock_sendmsg(iocb
, sock
, msg
, size
);
786 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
787 unsigned long nr_segs
, loff_t pos
)
789 struct sock_iocb siocb
, *x
;
794 x
= alloc_sock_iocb(iocb
, &siocb
);
798 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
802 * Atomic setting of ioctl hooks to avoid race
803 * with module unload.
806 static DEFINE_MUTEX(br_ioctl_mutex
);
807 static int (*br_ioctl_hook
) (unsigned int cmd
, void __user
*arg
) = NULL
;
809 void brioctl_set(int (*hook
) (unsigned int, void __user
*))
811 mutex_lock(&br_ioctl_mutex
);
812 br_ioctl_hook
= hook
;
813 mutex_unlock(&br_ioctl_mutex
);
816 EXPORT_SYMBOL(brioctl_set
);
818 static DEFINE_MUTEX(vlan_ioctl_mutex
);
819 static int (*vlan_ioctl_hook
) (void __user
*arg
);
821 void vlan_ioctl_set(int (*hook
) (void __user
*))
823 mutex_lock(&vlan_ioctl_mutex
);
824 vlan_ioctl_hook
= hook
;
825 mutex_unlock(&vlan_ioctl_mutex
);
828 EXPORT_SYMBOL(vlan_ioctl_set
);
830 static DEFINE_MUTEX(dlci_ioctl_mutex
);
831 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
833 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
835 mutex_lock(&dlci_ioctl_mutex
);
836 dlci_ioctl_hook
= hook
;
837 mutex_unlock(&dlci_ioctl_mutex
);
840 EXPORT_SYMBOL(dlci_ioctl_set
);
843 * With an ioctl, arg may well be a user mode pointer, but we don't know
844 * what to do with it - that's up to the protocol still.
847 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
850 void __user
*argp
= (void __user
*)arg
;
853 sock
= file
->private_data
;
854 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
855 err
= dev_ioctl(cmd
, argp
);
857 #ifdef CONFIG_WIRELESS_EXT
858 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
859 err
= dev_ioctl(cmd
, argp
);
861 #endif /* CONFIG_WIRELESS_EXT */
866 if (get_user(pid
, (int __user
*)argp
))
868 err
= f_setown(sock
->file
, pid
, 1);
872 err
= put_user(f_getown(sock
->file
),
881 request_module("bridge");
883 mutex_lock(&br_ioctl_mutex
);
885 err
= br_ioctl_hook(cmd
, argp
);
886 mutex_unlock(&br_ioctl_mutex
);
891 if (!vlan_ioctl_hook
)
892 request_module("8021q");
894 mutex_lock(&vlan_ioctl_mutex
);
896 err
= vlan_ioctl_hook(argp
);
897 mutex_unlock(&vlan_ioctl_mutex
);
902 if (!dlci_ioctl_hook
)
903 request_module("dlci");
905 if (dlci_ioctl_hook
) {
906 mutex_lock(&dlci_ioctl_mutex
);
907 err
= dlci_ioctl_hook(cmd
, argp
);
908 mutex_unlock(&dlci_ioctl_mutex
);
912 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
915 * If this ioctl is unknown try to hand it down
918 if (err
== -ENOIOCTLCMD
)
919 err
= dev_ioctl(cmd
, argp
);
925 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
928 struct socket
*sock
= NULL
;
930 err
= security_socket_create(family
, type
, protocol
, 1);
941 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
954 /* No kernel lock held - perfect */
955 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
960 * We can't return errors to poll, so it's either yes or no.
962 sock
= file
->private_data
;
963 return sock
->ops
->poll(file
, sock
, wait
);
966 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
968 struct socket
*sock
= file
->private_data
;
970 return sock
->ops
->mmap(file
, sock
, vma
);
973 static int sock_close(struct inode
*inode
, struct file
*filp
)
976 * It was possible the inode is NULL we were
977 * closing an unfinished socket.
981 printk(KERN_DEBUG
"sock_close: NULL inode\n");
984 sock_fasync(-1, filp
, 0);
985 sock_release(SOCKET_I(inode
));
990 * Update the socket async list
992 * Fasync_list locking strategy.
994 * 1. fasync_list is modified only under process context socket lock
995 * i.e. under semaphore.
996 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
997 * or under socket lock.
998 * 3. fasync_list can be used from softirq context, so that
999 * modification under socket lock have to be enhanced with
1000 * write_lock_bh(&sk->sk_callback_lock).
1004 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1006 struct fasync_struct
*fa
, *fna
= NULL
, **prev
;
1007 struct socket
*sock
;
1011 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1016 sock
= filp
->private_data
;
1026 prev
= &(sock
->fasync_list
);
1028 for (fa
= *prev
; fa
!= NULL
; prev
= &fa
->fa_next
, fa
= *prev
)
1029 if (fa
->fa_file
== filp
)
1034 write_lock_bh(&sk
->sk_callback_lock
);
1036 write_unlock_bh(&sk
->sk_callback_lock
);
1041 fna
->fa_file
= filp
;
1043 fna
->magic
= FASYNC_MAGIC
;
1044 fna
->fa_next
= sock
->fasync_list
;
1045 write_lock_bh(&sk
->sk_callback_lock
);
1046 sock
->fasync_list
= fna
;
1047 write_unlock_bh(&sk
->sk_callback_lock
);
1050 write_lock_bh(&sk
->sk_callback_lock
);
1051 *prev
= fa
->fa_next
;
1052 write_unlock_bh(&sk
->sk_callback_lock
);
1058 release_sock(sock
->sk
);
1062 /* This function may be called only under socket lock or callback_lock */
1064 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1066 if (!sock
|| !sock
->fasync_list
)
1071 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1075 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1080 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1083 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1088 static int __sock_create(int family
, int type
, int protocol
,
1089 struct socket
**res
, int kern
)
1092 struct socket
*sock
;
1093 const struct net_proto_family
*pf
;
1096 * Check protocol is in range
1098 if (family
< 0 || family
>= NPROTO
)
1099 return -EAFNOSUPPORT
;
1100 if (type
< 0 || type
>= SOCK_MAX
)
1105 This uglymoron is moved from INET layer to here to avoid
1106 deadlock in module load.
1108 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1112 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1118 err
= security_socket_create(family
, type
, protocol
, kern
);
1123 * Allocate the socket and allow the family to set things up. if
1124 * the protocol is 0, the family is instructed to select an appropriate
1127 sock
= sock_alloc();
1129 if (net_ratelimit())
1130 printk(KERN_WARNING
"socket: no more sockets\n");
1131 return -ENFILE
; /* Not exactly a match, but its the
1132 closest posix thing */
1137 #if defined(CONFIG_KMOD)
1138 /* Attempt to load a protocol module if the find failed.
1140 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1141 * requested real, full-featured networking support upon configuration.
1142 * Otherwise module support will break!
1144 if (net_families
[family
] == NULL
)
1145 request_module("net-pf-%d", family
);
1149 pf
= rcu_dereference(net_families
[family
]);
1150 err
= -EAFNOSUPPORT
;
1155 * We will call the ->create function, that possibly is in a loadable
1156 * module, so we have to bump that loadable module refcnt first.
1158 if (!try_module_get(pf
->owner
))
1161 /* Now protected by module ref count */
1164 err
= pf
->create(sock
, protocol
);
1166 goto out_module_put
;
1169 * Now to bump the refcnt of the [loadable] module that owns this
1170 * socket at sock_release time we decrement its refcnt.
1172 if (!try_module_get(sock
->ops
->owner
))
1173 goto out_module_busy
;
1176 * Now that we're done with the ->create function, the [loadable]
1177 * module can have its refcnt decremented
1179 module_put(pf
->owner
);
1180 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1182 goto out_sock_release
;
1188 err
= -EAFNOSUPPORT
;
1191 module_put(pf
->owner
);
1198 goto out_sock_release
;
1201 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1203 return __sock_create(family
, type
, protocol
, res
, 0);
1206 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1208 return __sock_create(family
, type
, protocol
, res
, 1);
1211 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1214 struct socket
*sock
;
1216 retval
= sock_create(family
, type
, protocol
, &sock
);
1220 retval
= sock_map_fd(sock
);
1225 /* It may be already another descriptor 8) Not kernel problem. */
1234 * Create a pair of connected sockets.
1237 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
,
1238 int __user
*usockvec
)
1240 struct socket
*sock1
, *sock2
;
1242 struct file
*newfile1
, *newfile2
;
1245 * Obtain the first socket and check if the underlying protocol
1246 * supports the socketpair call.
1249 err
= sock_create(family
, type
, protocol
, &sock1
);
1253 err
= sock_create(family
, type
, protocol
, &sock2
);
1257 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1259 goto out_release_both
;
1261 fd1
= sock_alloc_fd(&newfile1
);
1262 if (unlikely(fd1
< 0)) {
1264 goto out_release_both
;
1267 fd2
= sock_alloc_fd(&newfile2
);
1268 if (unlikely(fd2
< 0)) {
1272 goto out_release_both
;
1275 err
= sock_attach_fd(sock1
, newfile1
);
1276 if (unlikely(err
< 0)) {
1280 err
= sock_attach_fd(sock2
, newfile2
);
1281 if (unlikely(err
< 0)) {
1286 err
= audit_fd_pair(fd1
, fd2
);
1293 fd_install(fd1
, newfile1
);
1294 fd_install(fd2
, newfile2
);
1295 /* fd1 and fd2 may be already another descriptors.
1296 * Not kernel problem.
1299 err
= put_user(fd1
, &usockvec
[0]);
1301 err
= put_user(fd2
, &usockvec
[1]);
1310 sock_release(sock2
);
1312 sock_release(sock1
);
1318 sock_release(sock1
);
1321 sock_release(sock2
);
1329 * Bind a name to a socket. Nothing much to do here since it's
1330 * the protocol's responsibility to handle the local address.
1332 * We move the socket address to kernel space before we call
1333 * the protocol layer (having also checked the address is ok).
1336 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1338 struct socket
*sock
;
1339 char address
[MAX_SOCK_ADDR
];
1340 int err
, fput_needed
;
1342 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1344 err
= move_addr_to_kernel(umyaddr
, addrlen
, address
);
1346 err
= security_socket_bind(sock
,
1347 (struct sockaddr
*)address
,
1350 err
= sock
->ops
->bind(sock
,
1354 fput_light(sock
->file
, fput_needed
);
1360 * Perform a listen. Basically, we allow the protocol to do anything
1361 * necessary for a listen, and if that works, we mark the socket as
1362 * ready for listening.
1365 int sysctl_somaxconn __read_mostly
= SOMAXCONN
;
1367 asmlinkage
long sys_listen(int fd
, int backlog
)
1369 struct socket
*sock
;
1370 int err
, fput_needed
;
1372 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1374 if ((unsigned)backlog
> sysctl_somaxconn
)
1375 backlog
= sysctl_somaxconn
;
1377 err
= security_socket_listen(sock
, backlog
);
1379 err
= sock
->ops
->listen(sock
, backlog
);
1381 fput_light(sock
->file
, fput_needed
);
1387 * For accept, we attempt to create a new socket, set up the link
1388 * with the client, wake up the client, then return the new
1389 * connected fd. We collect the address of the connector in kernel
1390 * space and move it to user at the very end. This is unclean because
1391 * we open the socket then return an error.
1393 * 1003.1g adds the ability to recvmsg() to query connection pending
1394 * status to recvmsg. We need to add that support in a way thats
1395 * clean when we restucture accept also.
1398 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1399 int __user
*upeer_addrlen
)
1401 struct socket
*sock
, *newsock
;
1402 struct file
*newfile
;
1403 int err
, len
, newfd
, fput_needed
;
1404 char address
[MAX_SOCK_ADDR
];
1406 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1411 if (!(newsock
= sock_alloc()))
1414 newsock
->type
= sock
->type
;
1415 newsock
->ops
= sock
->ops
;
1418 * We don't need try_module_get here, as the listening socket (sock)
1419 * has the protocol module (sock->ops->owner) held.
1421 __module_get(newsock
->ops
->owner
);
1423 newfd
= sock_alloc_fd(&newfile
);
1424 if (unlikely(newfd
< 0)) {
1426 sock_release(newsock
);
1430 err
= sock_attach_fd(newsock
, newfile
);
1434 err
= security_socket_accept(sock
, newsock
);
1438 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1442 if (upeer_sockaddr
) {
1443 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
,
1445 err
= -ECONNABORTED
;
1448 err
= move_addr_to_user(address
, len
, upeer_sockaddr
,
1454 /* File flags are not inherited via accept() unlike another OSes. */
1456 fd_install(newfd
, newfile
);
1459 security_socket_post_accept(sock
, newsock
);
1462 fput_light(sock
->file
, fput_needed
);
1466 sock_release(newsock
);
1468 put_unused_fd(newfd
);
1472 put_unused_fd(newfd
);
1477 * Attempt to connect to a socket with the server address. The address
1478 * is in user space so we verify it is OK and move it to kernel space.
1480 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1483 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1484 * other SEQPACKET protocols that take time to connect() as it doesn't
1485 * include the -EINPROGRESS status for such sockets.
1488 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
,
1491 struct socket
*sock
;
1492 char address
[MAX_SOCK_ADDR
];
1493 int err
, fput_needed
;
1495 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1498 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1503 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1507 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
1508 sock
->file
->f_flags
);
1510 fput_light(sock
->file
, fput_needed
);
1516 * Get the local address ('name') of a socket object. Move the obtained
1517 * name to user space.
1520 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
1521 int __user
*usockaddr_len
)
1523 struct socket
*sock
;
1524 char address
[MAX_SOCK_ADDR
];
1525 int len
, err
, fput_needed
;
1527 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1531 err
= security_socket_getsockname(sock
);
1535 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1538 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1541 fput_light(sock
->file
, fput_needed
);
1547 * Get the remote address ('name') of a socket object. Move the obtained
1548 * name to user space.
1551 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
1552 int __user
*usockaddr_len
)
1554 struct socket
*sock
;
1555 char address
[MAX_SOCK_ADDR
];
1556 int len
, err
, fput_needed
;
1558 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1560 err
= security_socket_getpeername(sock
);
1562 fput_light(sock
->file
, fput_needed
);
1567 sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
,
1570 err
= move_addr_to_user(address
, len
, usockaddr
,
1572 fput_light(sock
->file
, fput_needed
);
1578 * Send a datagram to a given address. We move the address into kernel
1579 * space and check the user space data area is readable before invoking
1583 asmlinkage
long sys_sendto(int fd
, void __user
*buff
, size_t len
,
1584 unsigned flags
, struct sockaddr __user
*addr
,
1587 struct socket
*sock
;
1588 char address
[MAX_SOCK_ADDR
];
1593 struct file
*sock_file
;
1595 sock_file
= fget_light(fd
, &fput_needed
);
1600 sock
= sock_from_file(sock_file
, &err
);
1603 iov
.iov_base
= buff
;
1605 msg
.msg_name
= NULL
;
1608 msg
.msg_control
= NULL
;
1609 msg
.msg_controllen
= 0;
1610 msg
.msg_namelen
= 0;
1612 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1615 msg
.msg_name
= address
;
1616 msg
.msg_namelen
= addr_len
;
1618 if (sock
->file
->f_flags
& O_NONBLOCK
)
1619 flags
|= MSG_DONTWAIT
;
1620 msg
.msg_flags
= flags
;
1621 err
= sock_sendmsg(sock
, &msg
, len
);
1624 fput_light(sock_file
, fput_needed
);
1630 * Send a datagram down a socket.
1633 asmlinkage
long sys_send(int fd
, void __user
*buff
, size_t len
, unsigned flags
)
1635 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1639 * Receive a frame from the socket and optionally record the address of the
1640 * sender. We verify the buffers are writable and if needed move the
1641 * sender address from kernel to user space.
1644 asmlinkage
long sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
,
1645 unsigned flags
, struct sockaddr __user
*addr
,
1646 int __user
*addr_len
)
1648 struct socket
*sock
;
1651 char address
[MAX_SOCK_ADDR
];
1653 struct file
*sock_file
;
1656 sock_file
= fget_light(fd
, &fput_needed
);
1661 sock
= sock_from_file(sock_file
, &err
);
1665 msg
.msg_control
= NULL
;
1666 msg
.msg_controllen
= 0;
1670 iov
.iov_base
= ubuf
;
1671 msg
.msg_name
= address
;
1672 msg
.msg_namelen
= MAX_SOCK_ADDR
;
1673 if (sock
->file
->f_flags
& O_NONBLOCK
)
1674 flags
|= MSG_DONTWAIT
;
1675 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1677 if (err
>= 0 && addr
!= NULL
) {
1678 err2
= move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1683 fput_light(sock_file
, fput_needed
);
1689 * Receive a datagram from a socket.
1692 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1695 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1699 * Set a socket option. Because we don't know the option lengths we have
1700 * to pass the user mode parameter for the protocols to sort out.
1703 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
,
1704 char __user
*optval
, int optlen
)
1706 int err
, fput_needed
;
1707 struct socket
*sock
;
1712 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1714 err
= security_socket_setsockopt(sock
, level
, optname
);
1718 if (level
== SOL_SOCKET
)
1720 sock_setsockopt(sock
, level
, optname
, optval
,
1724 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1727 fput_light(sock
->file
, fput_needed
);
1733 * Get a socket option. Because we don't know the option lengths we have
1734 * to pass a user mode parameter for the protocols to sort out.
1737 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
,
1738 char __user
*optval
, int __user
*optlen
)
1740 int err
, fput_needed
;
1741 struct socket
*sock
;
1743 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1745 err
= security_socket_getsockopt(sock
, level
, optname
);
1749 if (level
== SOL_SOCKET
)
1751 sock_getsockopt(sock
, level
, optname
, optval
,
1755 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1758 fput_light(sock
->file
, fput_needed
);
1764 * Shutdown a socket.
1767 asmlinkage
long sys_shutdown(int fd
, int how
)
1769 int err
, fput_needed
;
1770 struct socket
*sock
;
1772 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1774 err
= security_socket_shutdown(sock
, how
);
1776 err
= sock
->ops
->shutdown(sock
, how
);
1777 fput_light(sock
->file
, fput_needed
);
1782 /* A couple of helpful macros for getting the address of the 32/64 bit
1783 * fields which are the same type (int / unsigned) on our platforms.
1785 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1786 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1787 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1790 * BSD sendmsg interface
1793 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1795 struct compat_msghdr __user
*msg_compat
=
1796 (struct compat_msghdr __user
*)msg
;
1797 struct socket
*sock
;
1798 char address
[MAX_SOCK_ADDR
];
1799 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1800 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1801 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1802 /* 20 is size of ipv6_pktinfo */
1803 unsigned char *ctl_buf
= ctl
;
1804 struct msghdr msg_sys
;
1805 int err
, ctl_len
, iov_size
, total_len
;
1809 if (MSG_CMSG_COMPAT
& flags
) {
1810 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1813 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1816 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1820 /* do not move before msg_sys is valid */
1822 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1825 /* Check whether to allocate the iovec area */
1827 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1828 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1829 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1834 /* This will also move the address data into kernel space */
1835 if (MSG_CMSG_COMPAT
& flags
) {
1836 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1838 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1845 if (msg_sys
.msg_controllen
> INT_MAX
)
1847 ctl_len
= msg_sys
.msg_controllen
;
1848 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1850 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1854 ctl_buf
= msg_sys
.msg_control
;
1855 ctl_len
= msg_sys
.msg_controllen
;
1856 } else if (ctl_len
) {
1857 if (ctl_len
> sizeof(ctl
)) {
1858 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1859 if (ctl_buf
== NULL
)
1864 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1865 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1866 * checking falls down on this.
1868 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1871 msg_sys
.msg_control
= ctl_buf
;
1873 msg_sys
.msg_flags
= flags
;
1875 if (sock
->file
->f_flags
& O_NONBLOCK
)
1876 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1877 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1881 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1883 if (iov
!= iovstack
)
1884 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1886 fput_light(sock
->file
, fput_needed
);
1892 * BSD recvmsg interface
1895 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
,
1898 struct compat_msghdr __user
*msg_compat
=
1899 (struct compat_msghdr __user
*)msg
;
1900 struct socket
*sock
;
1901 struct iovec iovstack
[UIO_FASTIOV
];
1902 struct iovec
*iov
= iovstack
;
1903 struct msghdr msg_sys
;
1904 unsigned long cmsg_ptr
;
1905 int err
, iov_size
, total_len
, len
;
1908 /* kernel mode address */
1909 char addr
[MAX_SOCK_ADDR
];
1911 /* user mode address pointers */
1912 struct sockaddr __user
*uaddr
;
1913 int __user
*uaddr_len
;
1915 if (MSG_CMSG_COMPAT
& flags
) {
1916 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1919 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1922 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1927 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1930 /* Check whether to allocate the iovec area */
1932 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1933 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1934 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1940 * Save the user-mode address (verify_iovec will change the
1941 * kernel msghdr to use the kernel address space)
1944 uaddr
= (void __user
*)msg_sys
.msg_name
;
1945 uaddr_len
= COMPAT_NAMELEN(msg
);
1946 if (MSG_CMSG_COMPAT
& flags
) {
1947 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1949 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1954 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1955 msg_sys
.msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
1957 if (sock
->file
->f_flags
& O_NONBLOCK
)
1958 flags
|= MSG_DONTWAIT
;
1959 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1964 if (uaddr
!= NULL
) {
1965 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
,
1970 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1974 if (MSG_CMSG_COMPAT
& flags
)
1975 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1976 &msg_compat
->msg_controllen
);
1978 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1979 &msg
->msg_controllen
);
1985 if (iov
!= iovstack
)
1986 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1988 fput_light(sock
->file
, fput_needed
);
1993 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1995 /* Argument list sizes for sys_socketcall */
1996 #define AL(x) ((x) * sizeof(unsigned long))
1997 static const unsigned char nargs
[18]={
1998 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1999 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2000 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
2006 * System call vectors.
2008 * Argument checking cleaned up. Saved 20% in size.
2009 * This function doesn't need to set the kernel lock because
2010 * it is set by the callees.
2013 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
2016 unsigned long a0
, a1
;
2019 if (call
< 1 || call
> SYS_RECVMSG
)
2022 /* copy_from_user should be SMP safe. */
2023 if (copy_from_user(a
, args
, nargs
[call
]))
2026 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2035 err
= sys_socket(a0
, a1
, a
[2]);
2038 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2041 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2044 err
= sys_listen(a0
, a1
);
2048 sys_accept(a0
, (struct sockaddr __user
*)a1
,
2049 (int __user
*)a
[2]);
2051 case SYS_GETSOCKNAME
:
2053 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2054 (int __user
*)a
[2]);
2056 case SYS_GETPEERNAME
:
2058 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2059 (int __user
*)a
[2]);
2061 case SYS_SOCKETPAIR
:
2062 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2065 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2068 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2069 (struct sockaddr __user
*)a
[4], a
[5]);
2072 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2075 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2076 (struct sockaddr __user
*)a
[4],
2077 (int __user
*)a
[5]);
2080 err
= sys_shutdown(a0
, a1
);
2082 case SYS_SETSOCKOPT
:
2083 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2085 case SYS_GETSOCKOPT
:
2087 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2088 (int __user
*)a
[4]);
2091 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2094 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2103 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2106 * sock_register - add a socket protocol handler
2107 * @ops: description of protocol
2109 * This function is called by a protocol handler that wants to
2110 * advertise its address family, and have it linked into the
2111 * socket interface. The value ops->family coresponds to the
2112 * socket system call protocol family.
2114 int sock_register(const struct net_proto_family
*ops
)
2118 if (ops
->family
>= NPROTO
) {
2119 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2124 spin_lock(&net_family_lock
);
2125 if (net_families
[ops
->family
])
2128 net_families
[ops
->family
] = ops
;
2131 spin_unlock(&net_family_lock
);
2133 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2138 * sock_unregister - remove a protocol handler
2139 * @family: protocol family to remove
2141 * This function is called by a protocol handler that wants to
2142 * remove its address family, and have it unlinked from the
2143 * new socket creation.
2145 * If protocol handler is a module, then it can use module reference
2146 * counts to protect against new references. If protocol handler is not
2147 * a module then it needs to provide its own protection in
2148 * the ops->create routine.
2150 void sock_unregister(int family
)
2152 BUG_ON(family
< 0 || family
>= NPROTO
);
2154 spin_lock(&net_family_lock
);
2155 net_families
[family
] = NULL
;
2156 spin_unlock(&net_family_lock
);
2160 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2163 static int __init
sock_init(void)
2166 * Initialize sock SLAB cache.
2172 * Initialize skbuff SLAB cache
2177 * Initialize the protocols module.
2181 register_filesystem(&sock_fs_type
);
2182 sock_mnt
= kern_mount(&sock_fs_type
);
2184 /* The real protocol initialization is performed in later initcalls.
2187 #ifdef CONFIG_NETFILTER
2194 core_initcall(sock_init
); /* early initcall */
2196 #ifdef CONFIG_PROC_FS
2197 void socket_seq_show(struct seq_file
*seq
)
2202 for_each_possible_cpu(cpu
)
2203 counter
+= per_cpu(sockets_in_use
, cpu
);
2205 /* It can be negative, by the way. 8) */
2209 seq_printf(seq
, "sockets: used %d\n", counter
);
2211 #endif /* CONFIG_PROC_FS */
2213 #ifdef CONFIG_COMPAT
2214 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2217 struct socket
*sock
= file
->private_data
;
2218 int ret
= -ENOIOCTLCMD
;
2220 if (sock
->ops
->compat_ioctl
)
2221 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2227 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
2229 return sock
->ops
->bind(sock
, addr
, addrlen
);
2232 int kernel_listen(struct socket
*sock
, int backlog
)
2234 return sock
->ops
->listen(sock
, backlog
);
2237 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
2239 struct sock
*sk
= sock
->sk
;
2242 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
2247 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
2249 sock_release(*newsock
);
2254 (*newsock
)->ops
= sock
->ops
;
2260 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
2263 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
2266 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
2269 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
2272 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
2275 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
2278 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
2279 char *optval
, int *optlen
)
2281 mm_segment_t oldfs
= get_fs();
2285 if (level
== SOL_SOCKET
)
2286 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2288 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2294 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
2295 char *optval
, int optlen
)
2297 mm_segment_t oldfs
= get_fs();
2301 if (level
== SOL_SOCKET
)
2302 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2304 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2310 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
2311 size_t size
, int flags
)
2313 if (sock
->ops
->sendpage
)
2314 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
2316 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
2319 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
2321 mm_segment_t oldfs
= get_fs();
2325 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
2331 /* ABI emulation layers need these two */
2332 EXPORT_SYMBOL(move_addr_to_kernel
);
2333 EXPORT_SYMBOL(move_addr_to_user
);
2334 EXPORT_SYMBOL(sock_create
);
2335 EXPORT_SYMBOL(sock_create_kern
);
2336 EXPORT_SYMBOL(sock_create_lite
);
2337 EXPORT_SYMBOL(sock_map_fd
);
2338 EXPORT_SYMBOL(sock_recvmsg
);
2339 EXPORT_SYMBOL(sock_register
);
2340 EXPORT_SYMBOL(sock_release
);
2341 EXPORT_SYMBOL(sock_sendmsg
);
2342 EXPORT_SYMBOL(sock_unregister
);
2343 EXPORT_SYMBOL(sock_wake_async
);
2344 EXPORT_SYMBOL(sockfd_lookup
);
2345 EXPORT_SYMBOL(kernel_sendmsg
);
2346 EXPORT_SYMBOL(kernel_recvmsg
);
2347 EXPORT_SYMBOL(kernel_bind
);
2348 EXPORT_SYMBOL(kernel_listen
);
2349 EXPORT_SYMBOL(kernel_accept
);
2350 EXPORT_SYMBOL(kernel_connect
);
2351 EXPORT_SYMBOL(kernel_getsockname
);
2352 EXPORT_SYMBOL(kernel_getpeername
);
2353 EXPORT_SYMBOL(kernel_getsockopt
);
2354 EXPORT_SYMBOL(kernel_setsockopt
);
2355 EXPORT_SYMBOL(kernel_sendpage
);
2356 EXPORT_SYMBOL(kernel_sock_ioctl
);