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/divert.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
98 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
99 unsigned long nr_segs
, loff_t pos
);
100 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
101 unsigned long nr_segs
, loff_t pos
);
102 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
104 static int sock_close(struct inode
*inode
, struct file
*file
);
105 static unsigned int sock_poll(struct file
*file
,
106 struct poll_table_struct
*wait
);
107 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
109 static long compat_sock_ioctl(struct file
*file
,
110 unsigned int cmd
, unsigned long arg
);
112 static int sock_fasync(int fd
, struct file
*filp
, int on
);
113 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*vector
,
114 unsigned long count
, loff_t
*ppos
);
115 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*vector
,
116 unsigned long count
, loff_t
*ppos
);
117 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
118 int offset
, size_t size
, loff_t
*ppos
, int more
);
121 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
122 * in the operation structures but are done directly via the socketcall() multiplexor.
125 static struct file_operations socket_file_ops
= {
126 .owner
= THIS_MODULE
,
128 .aio_read
= sock_aio_read
,
129 .aio_write
= sock_aio_write
,
131 .unlocked_ioctl
= sock_ioctl
,
133 .compat_ioctl
= compat_sock_ioctl
,
136 .open
= sock_no_open
, /* special open code to disallow open via /proc */
137 .release
= sock_close
,
138 .fasync
= sock_fasync
,
140 .writev
= sock_writev
,
141 .sendpage
= sock_sendpage
,
142 .splice_write
= generic_splice_sendpage
,
146 * The protocol list. Each protocol is registered in here.
149 static DEFINE_SPINLOCK(net_family_lock
);
150 static const struct net_proto_family
*net_families
[NPROTO
] __read_mostly
;
153 * Statistics counters of the socket lists
156 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
160 * Move socket addresses back and forth across the kernel/user
161 * divide and look after the messy bits.
164 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
165 16 for IP, 16 for IPX,
168 must be at least one bigger than
169 the AF_UNIX size (see net/unix/af_unix.c
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
186 if (ulen
< 0 || ulen
> MAX_SOCK_ADDR
)
190 if (copy_from_user(kaddr
, uaddr
, ulen
))
192 return audit_sockaddr(ulen
, kaddr
);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
,
218 err
= get_user(len
, ulen
);
223 if (len
< 0 || len
> MAX_SOCK_ADDR
)
226 if (audit_sockaddr(klen
, kaddr
))
228 if (copy_to_user(uaddr
, kaddr
, len
))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen
, ulen
);
238 #define SOCKFS_MAGIC 0x534F434B
240 static kmem_cache_t
*sock_inode_cachep __read_mostly
;
242 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
244 struct socket_alloc
*ei
;
246 ei
= kmem_cache_alloc(sock_inode_cachep
, SLAB_KERNEL
);
249 init_waitqueue_head(&ei
->socket
.wait
);
251 ei
->socket
.fasync_list
= NULL
;
252 ei
->socket
.state
= SS_UNCONNECTED
;
253 ei
->socket
.flags
= 0;
254 ei
->socket
.ops
= NULL
;
255 ei
->socket
.sk
= NULL
;
256 ei
->socket
.file
= NULL
;
258 return &ei
->vfs_inode
;
261 static void sock_destroy_inode(struct inode
*inode
)
263 kmem_cache_free(sock_inode_cachep
,
264 container_of(inode
, struct socket_alloc
, vfs_inode
));
267 static void init_once(void *foo
, kmem_cache_t
*cachep
, unsigned long flags
)
269 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
271 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
))
272 == SLAB_CTOR_CONSTRUCTOR
)
273 inode_init_once(&ei
->vfs_inode
);
276 static int init_inodecache(void)
278 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
279 sizeof(struct socket_alloc
),
281 (SLAB_HWCACHE_ALIGN
|
282 SLAB_RECLAIM_ACCOUNT
|
286 if (sock_inode_cachep
== NULL
)
291 static struct super_operations sockfs_ops
= {
292 .alloc_inode
= sock_alloc_inode
,
293 .destroy_inode
=sock_destroy_inode
,
294 .statfs
= simple_statfs
,
297 static int sockfs_get_sb(struct file_system_type
*fs_type
,
298 int flags
, const char *dev_name
, void *data
,
299 struct vfsmount
*mnt
)
301 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
305 static struct vfsmount
*sock_mnt __read_mostly
;
307 static struct file_system_type sock_fs_type
= {
309 .get_sb
= sockfs_get_sb
,
310 .kill_sb
= kill_anon_super
,
313 static int sockfs_delete_dentry(struct dentry
*dentry
)
317 static struct dentry_operations sockfs_dentry_operations
= {
318 .d_delete
= sockfs_delete_dentry
,
322 * Obtains the first available file descriptor and sets it up for use.
324 * These functions create file structures and maps them to fd space
325 * of the current process. On success it returns file descriptor
326 * and file struct implicitly stored in sock->file.
327 * Note that another thread may close file descriptor before we return
328 * from this function. We use the fact that now we do not refer
329 * to socket after mapping. If one day we will need it, this
330 * function will increment ref. count on file by 1.
332 * In any case returned fd MAY BE not valid!
333 * This race condition is unavoidable
334 * with shared fd spaces, we cannot solve it inside kernel,
335 * but we take care of internal coherence yet.
338 static int sock_alloc_fd(struct file
**filep
)
342 fd
= get_unused_fd();
343 if (likely(fd
>= 0)) {
344 struct file
*file
= get_empty_filp();
347 if (unlikely(!file
)) {
356 static int sock_attach_fd(struct socket
*sock
, struct file
*file
)
361 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
363 this.hash
= SOCK_INODE(sock
)->i_ino
;
365 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
366 if (unlikely(!file
->f_dentry
))
369 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
370 d_add(file
->f_dentry
, SOCK_INODE(sock
));
371 file
->f_vfsmnt
= mntget(sock_mnt
);
372 file
->f_mapping
= file
->f_dentry
->d_inode
->i_mapping
;
375 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
376 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
377 file
->f_flags
= O_RDWR
;
379 file
->private_data
= sock
;
384 int sock_map_fd(struct socket
*sock
)
386 struct file
*newfile
;
387 int fd
= sock_alloc_fd(&newfile
);
389 if (likely(fd
>= 0)) {
390 int err
= sock_attach_fd(sock
, newfile
);
392 if (unlikely(err
< 0)) {
397 fd_install(fd
, newfile
);
402 static struct socket
*sock_from_file(struct file
*file
, int *err
)
407 if (file
->f_op
== &socket_file_ops
)
408 return file
->private_data
; /* set in sock_map_fd */
410 inode
= file
->f_dentry
->d_inode
;
411 if (!S_ISSOCK(inode
->i_mode
)) {
416 sock
= SOCKET_I(inode
);
417 if (sock
->file
!= file
) {
418 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
425 * sockfd_lookup - Go from a file number to its socket slot
427 * @err: pointer to an error code return
429 * The file handle passed in is locked and the socket it is bound
430 * too is returned. If an error occurs the err pointer is overwritten
431 * with a negative errno code and NULL is returned. The function checks
432 * for both invalid handles and passing a handle which is not a socket.
434 * On a success the socket object pointer is returned.
437 struct socket
*sockfd_lookup(int fd
, int *err
)
448 sock
= sock_from_file(file
, err
);
454 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
460 file
= fget_light(fd
, fput_needed
);
462 sock
= sock_from_file(file
, err
);
465 fput_light(file
, *fput_needed
);
471 * sock_alloc - allocate a socket
473 * Allocate a new inode and socket object. The two are bound together
474 * and initialised. The socket is then returned. If we are out of inodes
478 static struct socket
*sock_alloc(void)
483 inode
= new_inode(sock_mnt
->mnt_sb
);
487 sock
= SOCKET_I(inode
);
489 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
490 inode
->i_uid
= current
->fsuid
;
491 inode
->i_gid
= current
->fsgid
;
493 get_cpu_var(sockets_in_use
)++;
494 put_cpu_var(sockets_in_use
);
499 * In theory you can't get an open on this inode, but /proc provides
500 * a back door. Remember to keep it shut otherwise you'll let the
501 * creepy crawlies in.
504 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
509 const struct file_operations bad_sock_fops
= {
510 .owner
= THIS_MODULE
,
511 .open
= sock_no_open
,
515 * sock_release - close a socket
516 * @sock: socket to close
518 * The socket is released from the protocol stack if it has a release
519 * callback, and the inode is then released if the socket is bound to
520 * an inode not a file.
523 void sock_release(struct socket
*sock
)
526 struct module
*owner
= sock
->ops
->owner
;
528 sock
->ops
->release(sock
);
533 if (sock
->fasync_list
)
534 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
536 get_cpu_var(sockets_in_use
)--;
537 put_cpu_var(sockets_in_use
);
539 iput(SOCK_INODE(sock
));
545 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
546 struct msghdr
*msg
, size_t size
)
548 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
556 err
= security_socket_sendmsg(sock
, msg
, size
);
560 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
563 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
566 struct sock_iocb siocb
;
569 init_sync_kiocb(&iocb
, NULL
);
570 iocb
.private = &siocb
;
571 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
572 if (-EIOCBQUEUED
== ret
)
573 ret
= wait_on_sync_kiocb(&iocb
);
577 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
578 struct kvec
*vec
, size_t num
, size_t size
)
580 mm_segment_t oldfs
= get_fs();
585 * the following is safe, since for compiler definitions of kvec and
586 * iovec are identical, yielding the same in-core layout and alignment
588 msg
->msg_iov
= (struct iovec
*)vec
;
589 msg
->msg_iovlen
= num
;
590 result
= sock_sendmsg(sock
, msg
, size
);
595 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
596 struct msghdr
*msg
, size_t size
, int flags
)
599 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
607 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
611 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
614 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
615 size_t size
, int flags
)
618 struct sock_iocb siocb
;
621 init_sync_kiocb(&iocb
, NULL
);
622 iocb
.private = &siocb
;
623 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
624 if (-EIOCBQUEUED
== ret
)
625 ret
= wait_on_sync_kiocb(&iocb
);
629 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
630 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
632 mm_segment_t oldfs
= get_fs();
637 * the following is safe, since for compiler definitions of kvec and
638 * iovec are identical, yielding the same in-core layout and alignment
640 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
641 result
= sock_recvmsg(sock
, msg
, size
, flags
);
646 static void sock_aio_dtor(struct kiocb
*iocb
)
648 kfree(iocb
->private);
651 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
652 int offset
, size_t size
, loff_t
*ppos
, int more
)
657 sock
= file
->private_data
;
659 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
663 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
666 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
667 struct sock_iocb
*siocb
)
669 if (!is_sync_kiocb(iocb
)) {
670 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
673 iocb
->ki_dtor
= sock_aio_dtor
;
677 iocb
->private = siocb
;
681 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
682 struct file
*file
, const struct iovec
*iov
,
683 unsigned long nr_segs
)
685 struct socket
*sock
= file
->private_data
;
689 for (i
= 0; i
< nr_segs
; i
++)
690 size
+= iov
[i
].iov_len
;
692 msg
->msg_name
= NULL
;
693 msg
->msg_namelen
= 0;
694 msg
->msg_control
= NULL
;
695 msg
->msg_controllen
= 0;
696 msg
->msg_iov
= (struct iovec
*)iov
;
697 msg
->msg_iovlen
= nr_segs
;
698 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
700 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
703 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
704 unsigned long nr_segs
, loff_t
*ppos
)
707 struct sock_iocb siocb
;
711 init_sync_kiocb(&iocb
, NULL
);
712 iocb
.private = &siocb
;
714 ret
= do_sock_read(&msg
, &iocb
, file
, iov
, nr_segs
);
715 if (-EIOCBQUEUED
== ret
)
716 ret
= wait_on_sync_kiocb(&iocb
);
720 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
721 unsigned long nr_segs
, loff_t pos
)
723 struct sock_iocb siocb
, *x
;
728 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
732 x
= alloc_sock_iocb(iocb
, &siocb
);
735 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
738 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
739 struct file
*file
, const struct iovec
*iov
,
740 unsigned long nr_segs
)
742 struct socket
*sock
= file
->private_data
;
746 for (i
= 0; i
< nr_segs
; i
++)
747 size
+= iov
[i
].iov_len
;
749 msg
->msg_name
= NULL
;
750 msg
->msg_namelen
= 0;
751 msg
->msg_control
= NULL
;
752 msg
->msg_controllen
= 0;
753 msg
->msg_iov
= (struct iovec
*)iov
;
754 msg
->msg_iovlen
= nr_segs
;
755 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
756 if (sock
->type
== SOCK_SEQPACKET
)
757 msg
->msg_flags
|= MSG_EOR
;
759 return __sock_sendmsg(iocb
, sock
, msg
, size
);
762 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
763 unsigned long nr_segs
, loff_t
*ppos
)
767 struct sock_iocb siocb
;
770 init_sync_kiocb(&iocb
, NULL
);
771 iocb
.private = &siocb
;
773 ret
= do_sock_write(&msg
, &iocb
, file
, iov
, nr_segs
);
774 if (-EIOCBQUEUED
== ret
)
775 ret
= wait_on_sync_kiocb(&iocb
);
779 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
780 unsigned long nr_segs
, loff_t pos
)
782 struct sock_iocb siocb
, *x
;
787 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
790 x
= alloc_sock_iocb(iocb
, &siocb
);
794 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
798 * Atomic setting of ioctl hooks to avoid race
799 * with module unload.
802 static DEFINE_MUTEX(br_ioctl_mutex
);
803 static int (*br_ioctl_hook
) (unsigned int cmd
, void __user
*arg
) = NULL
;
805 void brioctl_set(int (*hook
) (unsigned int, void __user
*))
807 mutex_lock(&br_ioctl_mutex
);
808 br_ioctl_hook
= hook
;
809 mutex_unlock(&br_ioctl_mutex
);
812 EXPORT_SYMBOL(brioctl_set
);
814 static DEFINE_MUTEX(vlan_ioctl_mutex
);
815 static int (*vlan_ioctl_hook
) (void __user
*arg
);
817 void vlan_ioctl_set(int (*hook
) (void __user
*))
819 mutex_lock(&vlan_ioctl_mutex
);
820 vlan_ioctl_hook
= hook
;
821 mutex_unlock(&vlan_ioctl_mutex
);
824 EXPORT_SYMBOL(vlan_ioctl_set
);
826 static DEFINE_MUTEX(dlci_ioctl_mutex
);
827 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
829 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
831 mutex_lock(&dlci_ioctl_mutex
);
832 dlci_ioctl_hook
= hook
;
833 mutex_unlock(&dlci_ioctl_mutex
);
836 EXPORT_SYMBOL(dlci_ioctl_set
);
839 * With an ioctl, arg may well be a user mode pointer, but we don't know
840 * what to do with it - that's up to the protocol still.
843 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
846 void __user
*argp
= (void __user
*)arg
;
849 sock
= file
->private_data
;
850 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
851 err
= dev_ioctl(cmd
, argp
);
853 #ifdef CONFIG_WIRELESS_EXT
854 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
855 err
= dev_ioctl(cmd
, argp
);
857 #endif /* CONFIG_WIRELESS_EXT */
862 if (get_user(pid
, (int __user
*)argp
))
864 err
= f_setown(sock
->file
, pid
, 1);
868 err
= put_user(sock
->file
->f_owner
.pid
,
877 request_module("bridge");
879 mutex_lock(&br_ioctl_mutex
);
881 err
= br_ioctl_hook(cmd
, argp
);
882 mutex_unlock(&br_ioctl_mutex
);
887 if (!vlan_ioctl_hook
)
888 request_module("8021q");
890 mutex_lock(&vlan_ioctl_mutex
);
892 err
= vlan_ioctl_hook(argp
);
893 mutex_unlock(&vlan_ioctl_mutex
);
897 /* Convert this to call through a hook */
898 err
= divert_ioctl(cmd
, argp
);
903 if (!dlci_ioctl_hook
)
904 request_module("dlci");
906 if (dlci_ioctl_hook
) {
907 mutex_lock(&dlci_ioctl_mutex
);
908 err
= dlci_ioctl_hook(cmd
, argp
);
909 mutex_unlock(&dlci_ioctl_mutex
);
913 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
916 * If this ioctl is unknown try to hand it down
919 if (err
== -ENOIOCTLCMD
)
920 err
= dev_ioctl(cmd
, argp
);
926 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
929 struct socket
*sock
= NULL
;
931 err
= security_socket_create(family
, type
, protocol
, 1);
942 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
955 /* No kernel lock held - perfect */
956 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
961 * We can't return errors to poll, so it's either yes or no.
963 sock
= file
->private_data
;
964 return sock
->ops
->poll(file
, sock
, wait
);
967 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
969 struct socket
*sock
= file
->private_data
;
971 return sock
->ops
->mmap(file
, sock
, vma
);
974 static int sock_close(struct inode
*inode
, struct file
*filp
)
977 * It was possible the inode is NULL we were
978 * closing an unfinished socket.
982 printk(KERN_DEBUG
"sock_close: NULL inode\n");
985 sock_fasync(-1, filp
, 0);
986 sock_release(SOCKET_I(inode
));
991 * Update the socket async list
993 * Fasync_list locking strategy.
995 * 1. fasync_list is modified only under process context socket lock
996 * i.e. under semaphore.
997 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
998 * or under socket lock.
999 * 3. fasync_list can be used from softirq context, so that
1000 * modification under socket lock have to be enhanced with
1001 * write_lock_bh(&sk->sk_callback_lock).
1005 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1007 struct fasync_struct
*fa
, *fna
= NULL
, **prev
;
1008 struct socket
*sock
;
1012 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1017 sock
= filp
->private_data
;
1027 prev
= &(sock
->fasync_list
);
1029 for (fa
= *prev
; fa
!= NULL
; prev
= &fa
->fa_next
, fa
= *prev
)
1030 if (fa
->fa_file
== filp
)
1035 write_lock_bh(&sk
->sk_callback_lock
);
1037 write_unlock_bh(&sk
->sk_callback_lock
);
1042 fna
->fa_file
= filp
;
1044 fna
->magic
= FASYNC_MAGIC
;
1045 fna
->fa_next
= sock
->fasync_list
;
1046 write_lock_bh(&sk
->sk_callback_lock
);
1047 sock
->fasync_list
= fna
;
1048 write_unlock_bh(&sk
->sk_callback_lock
);
1051 write_lock_bh(&sk
->sk_callback_lock
);
1052 *prev
= fa
->fa_next
;
1053 write_unlock_bh(&sk
->sk_callback_lock
);
1059 release_sock(sock
->sk
);
1063 /* This function may be called only under socket lock or callback_lock */
1065 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1067 if (!sock
|| !sock
->fasync_list
)
1072 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1076 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1081 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1084 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1089 static int __sock_create(int family
, int type
, int protocol
,
1090 struct socket
**res
, int kern
)
1093 struct socket
*sock
;
1094 const struct net_proto_family
*pf
;
1097 * Check protocol is in range
1099 if (family
< 0 || family
>= NPROTO
)
1100 return -EAFNOSUPPORT
;
1101 if (type
< 0 || type
>= SOCK_MAX
)
1106 This uglymoron is moved from INET layer to here to avoid
1107 deadlock in module load.
1109 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1113 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1119 err
= security_socket_create(family
, type
, protocol
, kern
);
1124 * Allocate the socket and allow the family to set things up. if
1125 * the protocol is 0, the family is instructed to select an appropriate
1128 sock
= sock_alloc();
1130 if (net_ratelimit())
1131 printk(KERN_WARNING
"socket: no more sockets\n");
1132 return -ENFILE
; /* Not exactly a match, but its the
1133 closest posix thing */
1138 #if defined(CONFIG_KMOD)
1139 /* Attempt to load a protocol module if the find failed.
1141 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1142 * requested real, full-featured networking support upon configuration.
1143 * Otherwise module support will break!
1145 if (net_families
[family
] == NULL
)
1146 request_module("net-pf-%d", family
);
1150 pf
= rcu_dereference(net_families
[family
]);
1151 err
= -EAFNOSUPPORT
;
1156 * We will call the ->create function, that possibly is in a loadable
1157 * module, so we have to bump that loadable module refcnt first.
1159 if (!try_module_get(pf
->owner
))
1162 /* Now protected by module ref count */
1165 err
= pf
->create(sock
, protocol
);
1167 goto out_module_put
;
1170 * Now to bump the refcnt of the [loadable] module that owns this
1171 * socket at sock_release time we decrement its refcnt.
1173 if (!try_module_get(sock
->ops
->owner
))
1174 goto out_module_busy
;
1177 * Now that we're done with the ->create function, the [loadable]
1178 * module can have its refcnt decremented
1180 module_put(pf
->owner
);
1181 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1189 err
= -EAFNOSUPPORT
;
1192 module_put(pf
->owner
);
1199 goto out_sock_release
;
1202 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1204 return __sock_create(family
, type
, protocol
, res
, 0);
1207 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1209 return __sock_create(family
, type
, protocol
, res
, 1);
1212 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1215 struct socket
*sock
;
1217 retval
= sock_create(family
, type
, protocol
, &sock
);
1221 retval
= sock_map_fd(sock
);
1226 /* It may be already another descriptor 8) Not kernel problem. */
1235 * Create a pair of connected sockets.
1238 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
,
1239 int __user
*usockvec
)
1241 struct socket
*sock1
, *sock2
;
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
;
1263 err
= sock_map_fd(sock1
);
1265 goto out_release_both
;
1268 err
= sock_map_fd(sock2
);
1273 /* fd1 and fd2 may be already another descriptors.
1274 * Not kernel problem.
1277 err
= put_user(fd1
, &usockvec
[0]);
1279 err
= put_user(fd2
, &usockvec
[1]);
1288 sock_release(sock2
);
1293 sock_release(sock2
);
1295 sock_release(sock1
);
1301 * Bind a name to a socket. Nothing much to do here since it's
1302 * the protocol's responsibility to handle the local address.
1304 * We move the socket address to kernel space before we call
1305 * the protocol layer (having also checked the address is ok).
1308 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1310 struct socket
*sock
;
1311 char address
[MAX_SOCK_ADDR
];
1312 int err
, fput_needed
;
1314 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1316 err
= move_addr_to_kernel(umyaddr
, addrlen
, address
);
1318 err
= security_socket_bind(sock
,
1319 (struct sockaddr
*)address
,
1322 err
= sock
->ops
->bind(sock
,
1326 fput_light(sock
->file
, fput_needed
);
1332 * Perform a listen. Basically, we allow the protocol to do anything
1333 * necessary for a listen, and if that works, we mark the socket as
1334 * ready for listening.
1337 int sysctl_somaxconn __read_mostly
= SOMAXCONN
;
1339 asmlinkage
long sys_listen(int fd
, int backlog
)
1341 struct socket
*sock
;
1342 int err
, fput_needed
;
1344 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1346 if ((unsigned)backlog
> sysctl_somaxconn
)
1347 backlog
= sysctl_somaxconn
;
1349 err
= security_socket_listen(sock
, backlog
);
1351 err
= sock
->ops
->listen(sock
, backlog
);
1353 fput_light(sock
->file
, fput_needed
);
1359 * For accept, we attempt to create a new socket, set up the link
1360 * with the client, wake up the client, then return the new
1361 * connected fd. We collect the address of the connector in kernel
1362 * space and move it to user at the very end. This is unclean because
1363 * we open the socket then return an error.
1365 * 1003.1g adds the ability to recvmsg() to query connection pending
1366 * status to recvmsg. We need to add that support in a way thats
1367 * clean when we restucture accept also.
1370 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1371 int __user
*upeer_addrlen
)
1373 struct socket
*sock
, *newsock
;
1374 struct file
*newfile
;
1375 int err
, len
, newfd
, fput_needed
;
1376 char address
[MAX_SOCK_ADDR
];
1378 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1383 if (!(newsock
= sock_alloc()))
1386 newsock
->type
= sock
->type
;
1387 newsock
->ops
= sock
->ops
;
1390 * We don't need try_module_get here, as the listening socket (sock)
1391 * has the protocol module (sock->ops->owner) held.
1393 __module_get(newsock
->ops
->owner
);
1395 newfd
= sock_alloc_fd(&newfile
);
1396 if (unlikely(newfd
< 0)) {
1398 sock_release(newsock
);
1402 err
= sock_attach_fd(newsock
, newfile
);
1406 err
= security_socket_accept(sock
, newsock
);
1410 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1414 if (upeer_sockaddr
) {
1415 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
,
1417 err
= -ECONNABORTED
;
1420 err
= move_addr_to_user(address
, len
, upeer_sockaddr
,
1426 /* File flags are not inherited via accept() unlike another OSes. */
1428 fd_install(newfd
, newfile
);
1431 security_socket_post_accept(sock
, newsock
);
1434 fput_light(sock
->file
, fput_needed
);
1439 put_unused_fd(newfd
);
1444 * Attempt to connect to a socket with the server address. The address
1445 * is in user space so we verify it is OK and move it to kernel space.
1447 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1450 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1451 * other SEQPACKET protocols that take time to connect() as it doesn't
1452 * include the -EINPROGRESS status for such sockets.
1455 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
,
1458 struct socket
*sock
;
1459 char address
[MAX_SOCK_ADDR
];
1460 int err
, fput_needed
;
1462 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1465 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1470 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1474 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
1475 sock
->file
->f_flags
);
1477 fput_light(sock
->file
, fput_needed
);
1483 * Get the local address ('name') of a socket object. Move the obtained
1484 * name to user space.
1487 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
1488 int __user
*usockaddr_len
)
1490 struct socket
*sock
;
1491 char address
[MAX_SOCK_ADDR
];
1492 int len
, err
, fput_needed
;
1494 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1498 err
= security_socket_getsockname(sock
);
1502 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1505 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1508 fput_light(sock
->file
, fput_needed
);
1514 * Get the remote address ('name') of a socket object. Move the obtained
1515 * name to user space.
1518 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
1519 int __user
*usockaddr_len
)
1521 struct socket
*sock
;
1522 char address
[MAX_SOCK_ADDR
];
1523 int len
, err
, fput_needed
;
1525 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1527 err
= security_socket_getpeername(sock
);
1529 fput_light(sock
->file
, fput_needed
);
1534 sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
,
1537 err
= move_addr_to_user(address
, len
, usockaddr
,
1539 fput_light(sock
->file
, fput_needed
);
1545 * Send a datagram to a given address. We move the address into kernel
1546 * space and check the user space data area is readable before invoking
1550 asmlinkage
long sys_sendto(int fd
, void __user
*buff
, size_t len
,
1551 unsigned flags
, struct sockaddr __user
*addr
,
1554 struct socket
*sock
;
1555 char address
[MAX_SOCK_ADDR
];
1560 struct file
*sock_file
;
1562 sock_file
= fget_light(fd
, &fput_needed
);
1566 sock
= sock_from_file(sock_file
, &err
);
1569 iov
.iov_base
= buff
;
1571 msg
.msg_name
= NULL
;
1574 msg
.msg_control
= NULL
;
1575 msg
.msg_controllen
= 0;
1576 msg
.msg_namelen
= 0;
1578 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1581 msg
.msg_name
= address
;
1582 msg
.msg_namelen
= addr_len
;
1584 if (sock
->file
->f_flags
& O_NONBLOCK
)
1585 flags
|= MSG_DONTWAIT
;
1586 msg
.msg_flags
= flags
;
1587 err
= sock_sendmsg(sock
, &msg
, len
);
1590 fput_light(sock_file
, fput_needed
);
1595 * Send a datagram down a socket.
1598 asmlinkage
long sys_send(int fd
, void __user
*buff
, size_t len
, unsigned flags
)
1600 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1604 * Receive a frame from the socket and optionally record the address of the
1605 * sender. We verify the buffers are writable and if needed move the
1606 * sender address from kernel to user space.
1609 asmlinkage
long sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
,
1610 unsigned flags
, struct sockaddr __user
*addr
,
1611 int __user
*addr_len
)
1613 struct socket
*sock
;
1616 char address
[MAX_SOCK_ADDR
];
1618 struct file
*sock_file
;
1621 sock_file
= fget_light(fd
, &fput_needed
);
1625 sock
= sock_from_file(sock_file
, &err
);
1629 msg
.msg_control
= NULL
;
1630 msg
.msg_controllen
= 0;
1634 iov
.iov_base
= ubuf
;
1635 msg
.msg_name
= address
;
1636 msg
.msg_namelen
= MAX_SOCK_ADDR
;
1637 if (sock
->file
->f_flags
& O_NONBLOCK
)
1638 flags
|= MSG_DONTWAIT
;
1639 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1641 if (err
>= 0 && addr
!= NULL
) {
1642 err2
= move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1647 fput_light(sock_file
, fput_needed
);
1652 * Receive a datagram from a socket.
1655 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1658 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1662 * Set a socket option. Because we don't know the option lengths we have
1663 * to pass the user mode parameter for the protocols to sort out.
1666 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
,
1667 char __user
*optval
, int optlen
)
1669 int err
, fput_needed
;
1670 struct socket
*sock
;
1675 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1677 err
= security_socket_setsockopt(sock
, level
, optname
);
1681 if (level
== SOL_SOCKET
)
1683 sock_setsockopt(sock
, level
, optname
, optval
,
1687 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1690 fput_light(sock
->file
, fput_needed
);
1696 * Get a socket option. Because we don't know the option lengths we have
1697 * to pass a user mode parameter for the protocols to sort out.
1700 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
,
1701 char __user
*optval
, int __user
*optlen
)
1703 int err
, fput_needed
;
1704 struct socket
*sock
;
1706 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1708 err
= security_socket_getsockopt(sock
, level
, optname
);
1712 if (level
== SOL_SOCKET
)
1714 sock_getsockopt(sock
, level
, optname
, optval
,
1718 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1721 fput_light(sock
->file
, fput_needed
);
1727 * Shutdown a socket.
1730 asmlinkage
long sys_shutdown(int fd
, int how
)
1732 int err
, fput_needed
;
1733 struct socket
*sock
;
1735 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1737 err
= security_socket_shutdown(sock
, how
);
1739 err
= sock
->ops
->shutdown(sock
, how
);
1740 fput_light(sock
->file
, fput_needed
);
1745 /* A couple of helpful macros for getting the address of the 32/64 bit
1746 * fields which are the same type (int / unsigned) on our platforms.
1748 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1749 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1750 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1753 * BSD sendmsg interface
1756 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1758 struct compat_msghdr __user
*msg_compat
=
1759 (struct compat_msghdr __user
*)msg
;
1760 struct socket
*sock
;
1761 char address
[MAX_SOCK_ADDR
];
1762 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1763 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1764 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1765 /* 20 is size of ipv6_pktinfo */
1766 unsigned char *ctl_buf
= ctl
;
1767 struct msghdr msg_sys
;
1768 int err
, ctl_len
, iov_size
, total_len
;
1772 if (MSG_CMSG_COMPAT
& flags
) {
1773 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1776 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1779 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1783 /* do not move before msg_sys is valid */
1785 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1788 /* Check whether to allocate the iovec area */
1790 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1791 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1792 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1797 /* This will also move the address data into kernel space */
1798 if (MSG_CMSG_COMPAT
& flags
) {
1799 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1801 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1808 if (msg_sys
.msg_controllen
> INT_MAX
)
1810 ctl_len
= msg_sys
.msg_controllen
;
1811 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1813 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1817 ctl_buf
= msg_sys
.msg_control
;
1818 ctl_len
= msg_sys
.msg_controllen
;
1819 } else if (ctl_len
) {
1820 if (ctl_len
> sizeof(ctl
)) {
1821 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1822 if (ctl_buf
== NULL
)
1827 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1828 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1829 * checking falls down on this.
1831 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1834 msg_sys
.msg_control
= ctl_buf
;
1836 msg_sys
.msg_flags
= flags
;
1838 if (sock
->file
->f_flags
& O_NONBLOCK
)
1839 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1840 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1844 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1846 if (iov
!= iovstack
)
1847 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1849 fput_light(sock
->file
, fput_needed
);
1855 * BSD recvmsg interface
1858 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
,
1861 struct compat_msghdr __user
*msg_compat
=
1862 (struct compat_msghdr __user
*)msg
;
1863 struct socket
*sock
;
1864 struct iovec iovstack
[UIO_FASTIOV
];
1865 struct iovec
*iov
= iovstack
;
1866 struct msghdr msg_sys
;
1867 unsigned long cmsg_ptr
;
1868 int err
, iov_size
, total_len
, len
;
1871 /* kernel mode address */
1872 char addr
[MAX_SOCK_ADDR
];
1874 /* user mode address pointers */
1875 struct sockaddr __user
*uaddr
;
1876 int __user
*uaddr_len
;
1878 if (MSG_CMSG_COMPAT
& flags
) {
1879 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1882 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1885 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1890 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1893 /* Check whether to allocate the iovec area */
1895 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1896 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1897 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1903 * Save the user-mode address (verify_iovec will change the
1904 * kernel msghdr to use the kernel address space)
1907 uaddr
= (void __user
*)msg_sys
.msg_name
;
1908 uaddr_len
= COMPAT_NAMELEN(msg
);
1909 if (MSG_CMSG_COMPAT
& flags
) {
1910 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1912 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1917 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1918 msg_sys
.msg_flags
= 0;
1919 if (MSG_CMSG_COMPAT
& flags
)
1920 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1922 if (sock
->file
->f_flags
& O_NONBLOCK
)
1923 flags
|= MSG_DONTWAIT
;
1924 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1929 if (uaddr
!= NULL
) {
1930 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
,
1935 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1939 if (MSG_CMSG_COMPAT
& flags
)
1940 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1941 &msg_compat
->msg_controllen
);
1943 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1944 &msg
->msg_controllen
);
1950 if (iov
!= iovstack
)
1951 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1953 fput_light(sock
->file
, fput_needed
);
1958 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1960 /* Argument list sizes for sys_socketcall */
1961 #define AL(x) ((x) * sizeof(unsigned long))
1962 static const unsigned char nargs
[18]={
1963 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1964 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1965 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1971 * System call vectors.
1973 * Argument checking cleaned up. Saved 20% in size.
1974 * This function doesn't need to set the kernel lock because
1975 * it is set by the callees.
1978 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
1981 unsigned long a0
, a1
;
1984 if (call
< 1 || call
> SYS_RECVMSG
)
1987 /* copy_from_user should be SMP safe. */
1988 if (copy_from_user(a
, args
, nargs
[call
]))
1991 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2000 err
= sys_socket(a0
, a1
, a
[2]);
2003 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2006 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2009 err
= sys_listen(a0
, a1
);
2013 sys_accept(a0
, (struct sockaddr __user
*)a1
,
2014 (int __user
*)a
[2]);
2016 case SYS_GETSOCKNAME
:
2018 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2019 (int __user
*)a
[2]);
2021 case SYS_GETPEERNAME
:
2023 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2024 (int __user
*)a
[2]);
2026 case SYS_SOCKETPAIR
:
2027 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2030 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2033 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2034 (struct sockaddr __user
*)a
[4], a
[5]);
2037 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2040 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2041 (struct sockaddr __user
*)a
[4],
2042 (int __user
*)a
[5]);
2045 err
= sys_shutdown(a0
, a1
);
2047 case SYS_SETSOCKOPT
:
2048 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2050 case SYS_GETSOCKOPT
:
2052 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2053 (int __user
*)a
[4]);
2056 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2059 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2068 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2071 * sock_register - add a socket protocol handler
2072 * @ops: description of protocol
2074 * This function is called by a protocol handler that wants to
2075 * advertise its address family, and have it linked into the
2076 * socket interface. The value ops->family coresponds to the
2077 * socket system call protocol family.
2079 int sock_register(const struct net_proto_family
*ops
)
2083 if (ops
->family
>= NPROTO
) {
2084 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2089 spin_lock(&net_family_lock
);
2090 if (net_families
[ops
->family
])
2093 net_families
[ops
->family
] = ops
;
2096 spin_unlock(&net_family_lock
);
2098 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2103 * sock_unregister - remove a protocol handler
2104 * @family: protocol family to remove
2106 * This function is called by a protocol handler that wants to
2107 * remove its address family, and have it unlinked from the
2108 * new socket creation.
2110 * If protocol handler is a module, then it can use module reference
2111 * counts to protect against new references. If protocol handler is not
2112 * a module then it needs to provide its own protection in
2113 * the ops->create routine.
2115 void sock_unregister(int family
)
2117 BUG_ON(family
< 0 || family
>= NPROTO
);
2119 spin_lock(&net_family_lock
);
2120 net_families
[family
] = NULL
;
2121 spin_unlock(&net_family_lock
);
2125 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2128 static int __init
sock_init(void)
2131 * Initialize sock SLAB cache.
2137 * Initialize skbuff SLAB cache
2142 * Initialize the protocols module.
2146 register_filesystem(&sock_fs_type
);
2147 sock_mnt
= kern_mount(&sock_fs_type
);
2149 /* The real protocol initialization is performed in later initcalls.
2152 #ifdef CONFIG_NETFILTER
2159 core_initcall(sock_init
); /* early initcall */
2161 #ifdef CONFIG_PROC_FS
2162 void socket_seq_show(struct seq_file
*seq
)
2167 for_each_possible_cpu(cpu
)
2168 counter
+= per_cpu(sockets_in_use
, cpu
);
2170 /* It can be negative, by the way. 8) */
2174 seq_printf(seq
, "sockets: used %d\n", counter
);
2176 #endif /* CONFIG_PROC_FS */
2178 #ifdef CONFIG_COMPAT
2179 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2182 struct socket
*sock
= file
->private_data
;
2183 int ret
= -ENOIOCTLCMD
;
2185 if (sock
->ops
->compat_ioctl
)
2186 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2192 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
2194 return sock
->ops
->bind(sock
, addr
, addrlen
);
2197 int kernel_listen(struct socket
*sock
, int backlog
)
2199 return sock
->ops
->listen(sock
, backlog
);
2202 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
2204 struct sock
*sk
= sock
->sk
;
2207 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
2212 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
2214 sock_release(*newsock
);
2218 (*newsock
)->ops
= sock
->ops
;
2224 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
2227 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
2230 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
2233 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
2236 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
2239 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
2242 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
2243 char *optval
, int *optlen
)
2245 mm_segment_t oldfs
= get_fs();
2249 if (level
== SOL_SOCKET
)
2250 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2252 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2258 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
2259 char *optval
, int optlen
)
2261 mm_segment_t oldfs
= get_fs();
2265 if (level
== SOL_SOCKET
)
2266 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2268 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2274 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
2275 size_t size
, int flags
)
2277 if (sock
->ops
->sendpage
)
2278 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
2280 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
2283 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
2285 mm_segment_t oldfs
= get_fs();
2289 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
2295 /* ABI emulation layers need these two */
2296 EXPORT_SYMBOL(move_addr_to_kernel
);
2297 EXPORT_SYMBOL(move_addr_to_user
);
2298 EXPORT_SYMBOL(sock_create
);
2299 EXPORT_SYMBOL(sock_create_kern
);
2300 EXPORT_SYMBOL(sock_create_lite
);
2301 EXPORT_SYMBOL(sock_map_fd
);
2302 EXPORT_SYMBOL(sock_recvmsg
);
2303 EXPORT_SYMBOL(sock_register
);
2304 EXPORT_SYMBOL(sock_release
);
2305 EXPORT_SYMBOL(sock_sendmsg
);
2306 EXPORT_SYMBOL(sock_unregister
);
2307 EXPORT_SYMBOL(sock_wake_async
);
2308 EXPORT_SYMBOL(sockfd_lookup
);
2309 EXPORT_SYMBOL(kernel_sendmsg
);
2310 EXPORT_SYMBOL(kernel_recvmsg
);
2311 EXPORT_SYMBOL(kernel_bind
);
2312 EXPORT_SYMBOL(kernel_listen
);
2313 EXPORT_SYMBOL(kernel_accept
);
2314 EXPORT_SYMBOL(kernel_connect
);
2315 EXPORT_SYMBOL(kernel_getsockname
);
2316 EXPORT_SYMBOL(kernel_getpeername
);
2317 EXPORT_SYMBOL(kernel_getsockopt
);
2318 EXPORT_SYMBOL(kernel_setsockopt
);
2319 EXPORT_SYMBOL(kernel_sendpage
);
2320 EXPORT_SYMBOL(kernel_sock_ioctl
);