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/smp_lock.h>
63 #include <linux/socket.h>
64 #include <linux/file.h>
65 #include <linux/net.h>
66 #include <linux/interrupt.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
, char __user
*buf
,
99 size_t size
, loff_t pos
);
100 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*buf
,
101 size_t size
, 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 struct net_proto_family
*net_families
[NPROTO
];
151 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
152 static atomic_t net_family_lockct
= ATOMIC_INIT(0);
153 static DEFINE_SPINLOCK(net_family_lock
);
155 /* The strategy is: modifications net_family vector are short, do not
156 sleep and veeery rare, but read access should be free of any exclusive
160 static void net_family_write_lock(void)
162 spin_lock(&net_family_lock
);
163 while (atomic_read(&net_family_lockct
) != 0) {
164 spin_unlock(&net_family_lock
);
168 spin_lock(&net_family_lock
);
172 static __inline__
void net_family_write_unlock(void)
174 spin_unlock(&net_family_lock
);
177 static __inline__
void net_family_read_lock(void)
179 atomic_inc(&net_family_lockct
);
180 spin_unlock_wait(&net_family_lock
);
183 static __inline__
void net_family_read_unlock(void)
185 atomic_dec(&net_family_lockct
);
189 #define net_family_write_lock() do { } while(0)
190 #define net_family_write_unlock() do { } while(0)
191 #define net_family_read_lock() do { } while(0)
192 #define net_family_read_unlock() do { } while(0)
196 * Statistics counters of the socket lists
199 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
203 * Move socket addresses back and forth across the kernel/user
204 * divide and look after the messy bits.
207 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
208 16 for IP, 16 for IPX,
211 must be at least one bigger than
212 the AF_UNIX size (see net/unix/af_unix.c
217 * move_addr_to_kernel - copy a socket address into kernel space
218 * @uaddr: Address in user space
219 * @kaddr: Address in kernel space
220 * @ulen: Length in user space
222 * The address is copied into kernel space. If the provided address is
223 * too long an error code of -EINVAL is returned. If the copy gives
224 * invalid addresses -EFAULT is returned. On a success 0 is returned.
227 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, void *kaddr
)
229 if (ulen
< 0 || ulen
> MAX_SOCK_ADDR
)
233 if (copy_from_user(kaddr
, uaddr
, ulen
))
235 return audit_sockaddr(ulen
, kaddr
);
239 * move_addr_to_user - copy an address to user space
240 * @kaddr: kernel space address
241 * @klen: length of address in kernel
242 * @uaddr: user space address
243 * @ulen: pointer to user length field
245 * The value pointed to by ulen on entry is the buffer length available.
246 * This is overwritten with the buffer space used. -EINVAL is returned
247 * if an overlong buffer is specified or a negative buffer size. -EFAULT
248 * is returned if either the buffer or the length field are not
250 * After copying the data up to the limit the user specifies, the true
251 * length of the data is written over the length limit the user
252 * specified. Zero is returned for a success.
255 int move_addr_to_user(void *kaddr
, int klen
, void __user
*uaddr
,
261 err
= get_user(len
, ulen
);
266 if (len
< 0 || len
> MAX_SOCK_ADDR
)
269 if (audit_sockaddr(klen
, kaddr
))
271 if (copy_to_user(uaddr
, kaddr
, len
))
275 * "fromlen shall refer to the value before truncation.."
278 return __put_user(klen
, ulen
);
281 #define SOCKFS_MAGIC 0x534F434B
283 static kmem_cache_t
*sock_inode_cachep __read_mostly
;
285 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
287 struct socket_alloc
*ei
;
289 ei
= kmem_cache_alloc(sock_inode_cachep
, SLAB_KERNEL
);
292 init_waitqueue_head(&ei
->socket
.wait
);
294 ei
->socket
.fasync_list
= NULL
;
295 ei
->socket
.state
= SS_UNCONNECTED
;
296 ei
->socket
.flags
= 0;
297 ei
->socket
.ops
= NULL
;
298 ei
->socket
.sk
= NULL
;
299 ei
->socket
.file
= NULL
;
300 ei
->socket
.flags
= 0;
302 return &ei
->vfs_inode
;
305 static void sock_destroy_inode(struct inode
*inode
)
307 kmem_cache_free(sock_inode_cachep
,
308 container_of(inode
, struct socket_alloc
, vfs_inode
));
311 static void init_once(void *foo
, kmem_cache_t
*cachep
, unsigned long flags
)
313 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
315 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
))
316 == SLAB_CTOR_CONSTRUCTOR
)
317 inode_init_once(&ei
->vfs_inode
);
320 static int init_inodecache(void)
322 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
323 sizeof(struct socket_alloc
),
325 (SLAB_HWCACHE_ALIGN
|
326 SLAB_RECLAIM_ACCOUNT
|
330 if (sock_inode_cachep
== NULL
)
335 static struct super_operations sockfs_ops
= {
336 .alloc_inode
= sock_alloc_inode
,
337 .destroy_inode
=sock_destroy_inode
,
338 .statfs
= simple_statfs
,
341 static int sockfs_get_sb(struct file_system_type
*fs_type
,
342 int flags
, const char *dev_name
, void *data
,
343 struct vfsmount
*mnt
)
345 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
349 static struct vfsmount
*sock_mnt __read_mostly
;
351 static struct file_system_type sock_fs_type
= {
353 .get_sb
= sockfs_get_sb
,
354 .kill_sb
= kill_anon_super
,
357 static int sockfs_delete_dentry(struct dentry
*dentry
)
361 static struct dentry_operations sockfs_dentry_operations
= {
362 .d_delete
= sockfs_delete_dentry
,
366 * Obtains the first available file descriptor and sets it up for use.
368 * These functions create file structures and maps them to fd space
369 * of the current process. On success it returns file descriptor
370 * and file struct implicitly stored in sock->file.
371 * Note that another thread may close file descriptor before we return
372 * from this function. We use the fact that now we do not refer
373 * to socket after mapping. If one day we will need it, this
374 * function will increment ref. count on file by 1.
376 * In any case returned fd MAY BE not valid!
377 * This race condition is unavoidable
378 * with shared fd spaces, we cannot solve it inside kernel,
379 * but we take care of internal coherence yet.
382 static int sock_alloc_fd(struct file
**filep
)
386 fd
= get_unused_fd();
387 if (likely(fd
>= 0)) {
388 struct file
*file
= get_empty_filp();
391 if (unlikely(!file
)) {
400 static int sock_attach_fd(struct socket
*sock
, struct file
*file
)
405 this.len
= sprintf(name
, "[%lu]", SOCK_INODE(sock
)->i_ino
);
407 this.hash
= SOCK_INODE(sock
)->i_ino
;
409 file
->f_dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &this);
410 if (unlikely(!file
->f_dentry
))
413 file
->f_dentry
->d_op
= &sockfs_dentry_operations
;
414 d_add(file
->f_dentry
, SOCK_INODE(sock
));
415 file
->f_vfsmnt
= mntget(sock_mnt
);
416 file
->f_mapping
= file
->f_dentry
->d_inode
->i_mapping
;
419 file
->f_op
= SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
420 file
->f_mode
= FMODE_READ
| FMODE_WRITE
;
421 file
->f_flags
= O_RDWR
;
423 file
->private_data
= sock
;
428 int sock_map_fd(struct socket
*sock
)
430 struct file
*newfile
;
431 int fd
= sock_alloc_fd(&newfile
);
433 if (likely(fd
>= 0)) {
434 int err
= sock_attach_fd(sock
, newfile
);
436 if (unlikely(err
< 0)) {
441 fd_install(fd
, newfile
);
446 static struct socket
*sock_from_file(struct file
*file
, int *err
)
451 if (file
->f_op
== &socket_file_ops
)
452 return file
->private_data
; /* set in sock_map_fd */
454 inode
= file
->f_dentry
->d_inode
;
455 if (!S_ISSOCK(inode
->i_mode
)) {
460 sock
= SOCKET_I(inode
);
461 if (sock
->file
!= file
) {
462 printk(KERN_ERR
"socki_lookup: socket file changed!\n");
469 * sockfd_lookup - Go from a file number to its socket slot
471 * @err: pointer to an error code return
473 * The file handle passed in is locked and the socket it is bound
474 * too is returned. If an error occurs the err pointer is overwritten
475 * with a negative errno code and NULL is returned. The function checks
476 * for both invalid handles and passing a handle which is not a socket.
478 * On a success the socket object pointer is returned.
481 struct socket
*sockfd_lookup(int fd
, int *err
)
492 sock
= sock_from_file(file
, err
);
498 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
504 file
= fget_light(fd
, fput_needed
);
506 sock
= sock_from_file(file
, err
);
509 fput_light(file
, *fput_needed
);
515 * sock_alloc - allocate a socket
517 * Allocate a new inode and socket object. The two are bound together
518 * and initialised. The socket is then returned. If we are out of inodes
522 static struct socket
*sock_alloc(void)
527 inode
= new_inode(sock_mnt
->mnt_sb
);
531 sock
= SOCKET_I(inode
);
533 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
534 inode
->i_uid
= current
->fsuid
;
535 inode
->i_gid
= current
->fsgid
;
537 get_cpu_var(sockets_in_use
)++;
538 put_cpu_var(sockets_in_use
);
543 * In theory you can't get an open on this inode, but /proc provides
544 * a back door. Remember to keep it shut otherwise you'll let the
545 * creepy crawlies in.
548 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
553 const struct file_operations bad_sock_fops
= {
554 .owner
= THIS_MODULE
,
555 .open
= sock_no_open
,
559 * sock_release - close a socket
560 * @sock: socket to close
562 * The socket is released from the protocol stack if it has a release
563 * callback, and the inode is then released if the socket is bound to
564 * an inode not a file.
567 void sock_release(struct socket
*sock
)
570 struct module
*owner
= sock
->ops
->owner
;
572 sock
->ops
->release(sock
);
577 if (sock
->fasync_list
)
578 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
580 get_cpu_var(sockets_in_use
)--;
581 put_cpu_var(sockets_in_use
);
583 iput(SOCK_INODE(sock
));
589 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
590 struct msghdr
*msg
, size_t size
)
592 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
600 err
= security_socket_sendmsg(sock
, msg
, size
);
604 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
607 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
610 struct sock_iocb siocb
;
613 init_sync_kiocb(&iocb
, NULL
);
614 iocb
.private = &siocb
;
615 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
616 if (-EIOCBQUEUED
== ret
)
617 ret
= wait_on_sync_kiocb(&iocb
);
621 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
622 struct kvec
*vec
, size_t num
, size_t size
)
624 mm_segment_t oldfs
= get_fs();
629 * the following is safe, since for compiler definitions of kvec and
630 * iovec are identical, yielding the same in-core layout and alignment
632 msg
->msg_iov
= (struct iovec
*)vec
;
633 msg
->msg_iovlen
= num
;
634 result
= sock_sendmsg(sock
, msg
, size
);
639 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
640 struct msghdr
*msg
, size_t size
, int flags
)
643 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
651 err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
655 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
658 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
659 size_t size
, int flags
)
662 struct sock_iocb siocb
;
665 init_sync_kiocb(&iocb
, NULL
);
666 iocb
.private = &siocb
;
667 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
668 if (-EIOCBQUEUED
== ret
)
669 ret
= wait_on_sync_kiocb(&iocb
);
673 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
674 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
676 mm_segment_t oldfs
= get_fs();
681 * the following is safe, since for compiler definitions of kvec and
682 * iovec are identical, yielding the same in-core layout and alignment
684 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
685 result
= sock_recvmsg(sock
, msg
, size
, flags
);
690 static void sock_aio_dtor(struct kiocb
*iocb
)
692 kfree(iocb
->private);
695 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
696 int offset
, size_t size
, loff_t
*ppos
, int more
)
701 sock
= file
->private_data
;
703 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
707 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
710 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
711 char __user
*ubuf
, size_t size
,
712 struct sock_iocb
*siocb
)
714 if (!is_sync_kiocb(iocb
)) {
715 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
718 iocb
->ki_dtor
= sock_aio_dtor
;
722 siocb
->async_iov
.iov_base
= ubuf
;
723 siocb
->async_iov
.iov_len
= size
;
725 iocb
->private = siocb
;
729 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
730 struct file
*file
, struct iovec
*iov
,
731 unsigned long nr_segs
)
733 struct socket
*sock
= file
->private_data
;
737 for (i
= 0; i
< nr_segs
; i
++)
738 size
+= iov
[i
].iov_len
;
740 msg
->msg_name
= NULL
;
741 msg
->msg_namelen
= 0;
742 msg
->msg_control
= NULL
;
743 msg
->msg_controllen
= 0;
744 msg
->msg_iov
= (struct iovec
*)iov
;
745 msg
->msg_iovlen
= nr_segs
;
746 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
748 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
751 static ssize_t
sock_readv(struct file
*file
, const struct iovec
*iov
,
752 unsigned long nr_segs
, loff_t
*ppos
)
755 struct sock_iocb siocb
;
759 init_sync_kiocb(&iocb
, NULL
);
760 iocb
.private = &siocb
;
762 ret
= do_sock_read(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
763 if (-EIOCBQUEUED
== ret
)
764 ret
= wait_on_sync_kiocb(&iocb
);
768 static ssize_t
sock_aio_read(struct kiocb
*iocb
, char __user
*ubuf
,
769 size_t count
, loff_t pos
)
771 struct sock_iocb siocb
, *x
;
775 if (count
== 0) /* Match SYS5 behaviour */
778 x
= alloc_sock_iocb(iocb
, ubuf
, count
, &siocb
);
781 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
,
785 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
786 struct file
*file
, struct iovec
*iov
,
787 unsigned long nr_segs
)
789 struct socket
*sock
= file
->private_data
;
793 for (i
= 0; i
< nr_segs
; i
++)
794 size
+= iov
[i
].iov_len
;
796 msg
->msg_name
= NULL
;
797 msg
->msg_namelen
= 0;
798 msg
->msg_control
= NULL
;
799 msg
->msg_controllen
= 0;
800 msg
->msg_iov
= (struct iovec
*)iov
;
801 msg
->msg_iovlen
= nr_segs
;
802 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
803 if (sock
->type
== SOCK_SEQPACKET
)
804 msg
->msg_flags
|= MSG_EOR
;
806 return __sock_sendmsg(iocb
, sock
, msg
, size
);
809 static ssize_t
sock_writev(struct file
*file
, const struct iovec
*iov
,
810 unsigned long nr_segs
, loff_t
*ppos
)
814 struct sock_iocb siocb
;
817 init_sync_kiocb(&iocb
, NULL
);
818 iocb
.private = &siocb
;
820 ret
= do_sock_write(&msg
, &iocb
, file
, (struct iovec
*)iov
, nr_segs
);
821 if (-EIOCBQUEUED
== ret
)
822 ret
= wait_on_sync_kiocb(&iocb
);
826 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const char __user
*ubuf
,
827 size_t count
, loff_t pos
)
829 struct sock_iocb siocb
, *x
;
833 if (count
== 0) /* Match SYS5 behaviour */
836 x
= alloc_sock_iocb(iocb
, (void __user
*)ubuf
, count
, &siocb
);
840 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
,
845 * Atomic setting of ioctl hooks to avoid race
846 * with module unload.
849 static DEFINE_MUTEX(br_ioctl_mutex
);
850 static int (*br_ioctl_hook
) (unsigned int cmd
, void __user
*arg
) = NULL
;
852 void brioctl_set(int (*hook
) (unsigned int, void __user
*))
854 mutex_lock(&br_ioctl_mutex
);
855 br_ioctl_hook
= hook
;
856 mutex_unlock(&br_ioctl_mutex
);
859 EXPORT_SYMBOL(brioctl_set
);
861 static DEFINE_MUTEX(vlan_ioctl_mutex
);
862 static int (*vlan_ioctl_hook
) (void __user
*arg
);
864 void vlan_ioctl_set(int (*hook
) (void __user
*))
866 mutex_lock(&vlan_ioctl_mutex
);
867 vlan_ioctl_hook
= hook
;
868 mutex_unlock(&vlan_ioctl_mutex
);
871 EXPORT_SYMBOL(vlan_ioctl_set
);
873 static DEFINE_MUTEX(dlci_ioctl_mutex
);
874 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
876 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
878 mutex_lock(&dlci_ioctl_mutex
);
879 dlci_ioctl_hook
= hook
;
880 mutex_unlock(&dlci_ioctl_mutex
);
883 EXPORT_SYMBOL(dlci_ioctl_set
);
886 * With an ioctl, arg may well be a user mode pointer, but we don't know
887 * what to do with it - that's up to the protocol still.
890 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
893 void __user
*argp
= (void __user
*)arg
;
896 sock
= file
->private_data
;
897 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
898 err
= dev_ioctl(cmd
, argp
);
900 #ifdef CONFIG_WIRELESS_EXT
901 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
902 err
= dev_ioctl(cmd
, argp
);
904 #endif /* CONFIG_WIRELESS_EXT */
909 if (get_user(pid
, (int __user
*)argp
))
911 err
= f_setown(sock
->file
, pid
, 1);
915 err
= put_user(sock
->file
->f_owner
.pid
,
924 request_module("bridge");
926 mutex_lock(&br_ioctl_mutex
);
928 err
= br_ioctl_hook(cmd
, argp
);
929 mutex_unlock(&br_ioctl_mutex
);
934 if (!vlan_ioctl_hook
)
935 request_module("8021q");
937 mutex_lock(&vlan_ioctl_mutex
);
939 err
= vlan_ioctl_hook(argp
);
940 mutex_unlock(&vlan_ioctl_mutex
);
944 /* Convert this to call through a hook */
945 err
= divert_ioctl(cmd
, argp
);
950 if (!dlci_ioctl_hook
)
951 request_module("dlci");
953 if (dlci_ioctl_hook
) {
954 mutex_lock(&dlci_ioctl_mutex
);
955 err
= dlci_ioctl_hook(cmd
, argp
);
956 mutex_unlock(&dlci_ioctl_mutex
);
960 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
963 * If this ioctl is unknown try to hand it down
966 if (err
== -ENOIOCTLCMD
)
967 err
= dev_ioctl(cmd
, argp
);
973 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
976 struct socket
*sock
= NULL
;
978 err
= security_socket_create(family
, type
, protocol
, 1);
989 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1002 /* No kernel lock held - perfect */
1003 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1005 struct socket
*sock
;
1008 * We can't return errors to poll, so it's either yes or no.
1010 sock
= file
->private_data
;
1011 return sock
->ops
->poll(file
, sock
, wait
);
1014 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1016 struct socket
*sock
= file
->private_data
;
1018 return sock
->ops
->mmap(file
, sock
, vma
);
1021 static int sock_close(struct inode
*inode
, struct file
*filp
)
1024 * It was possible the inode is NULL we were
1025 * closing an unfinished socket.
1029 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1032 sock_fasync(-1, filp
, 0);
1033 sock_release(SOCKET_I(inode
));
1038 * Update the socket async list
1040 * Fasync_list locking strategy.
1042 * 1. fasync_list is modified only under process context socket lock
1043 * i.e. under semaphore.
1044 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1045 * or under socket lock.
1046 * 3. fasync_list can be used from softirq context, so that
1047 * modification under socket lock have to be enhanced with
1048 * write_lock_bh(&sk->sk_callback_lock).
1052 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1054 struct fasync_struct
*fa
, *fna
= NULL
, **prev
;
1055 struct socket
*sock
;
1059 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1064 sock
= filp
->private_data
;
1074 prev
= &(sock
->fasync_list
);
1076 for (fa
= *prev
; fa
!= NULL
; prev
= &fa
->fa_next
, fa
= *prev
)
1077 if (fa
->fa_file
== filp
)
1082 write_lock_bh(&sk
->sk_callback_lock
);
1084 write_unlock_bh(&sk
->sk_callback_lock
);
1089 fna
->fa_file
= filp
;
1091 fna
->magic
= FASYNC_MAGIC
;
1092 fna
->fa_next
= sock
->fasync_list
;
1093 write_lock_bh(&sk
->sk_callback_lock
);
1094 sock
->fasync_list
= fna
;
1095 write_unlock_bh(&sk
->sk_callback_lock
);
1098 write_lock_bh(&sk
->sk_callback_lock
);
1099 *prev
= fa
->fa_next
;
1100 write_unlock_bh(&sk
->sk_callback_lock
);
1106 release_sock(sock
->sk
);
1110 /* This function may be called only under socket lock or callback_lock */
1112 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1114 if (!sock
|| !sock
->fasync_list
)
1119 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1123 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1128 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1131 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1136 static int __sock_create(int family
, int type
, int protocol
,
1137 struct socket
**res
, int kern
)
1140 struct socket
*sock
;
1143 * Check protocol is in range
1145 if (family
< 0 || family
>= NPROTO
)
1146 return -EAFNOSUPPORT
;
1147 if (type
< 0 || type
>= SOCK_MAX
)
1152 This uglymoron is moved from INET layer to here to avoid
1153 deadlock in module load.
1155 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1159 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1165 err
= security_socket_create(family
, type
, protocol
, kern
);
1169 #if defined(CONFIG_KMOD)
1170 /* Attempt to load a protocol module if the find failed.
1172 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1173 * requested real, full-featured networking support upon configuration.
1174 * Otherwise module support will break!
1176 if (net_families
[family
] == NULL
) {
1177 request_module("net-pf-%d", family
);
1181 net_family_read_lock();
1182 if (net_families
[family
] == NULL
) {
1183 err
= -EAFNOSUPPORT
;
1188 * Allocate the socket and allow the family to set things up. if
1189 * the protocol is 0, the family is instructed to select an appropriate
1193 if (!(sock
= sock_alloc())) {
1194 if (net_ratelimit())
1195 printk(KERN_WARNING
"socket: no more sockets\n");
1196 err
= -ENFILE
; /* Not exactly a match, but its the
1197 closest posix thing */
1204 * We will call the ->create function, that possibly is in a loadable
1205 * module, so we have to bump that loadable module refcnt first.
1207 err
= -EAFNOSUPPORT
;
1208 if (!try_module_get(net_families
[family
]->owner
))
1211 if ((err
= net_families
[family
]->create(sock
, protocol
)) < 0) {
1213 goto out_module_put
;
1217 * Now to bump the refcnt of the [loadable] module that owns this
1218 * socket at sock_release time we decrement its refcnt.
1220 if (!try_module_get(sock
->ops
->owner
)) {
1222 goto out_module_put
;
1225 * Now that we're done with the ->create function, the [loadable]
1226 * module can have its refcnt decremented
1228 module_put(net_families
[family
]->owner
);
1230 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1235 net_family_read_unlock();
1238 module_put(net_families
[family
]->owner
);
1244 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1246 return __sock_create(family
, type
, protocol
, res
, 0);
1249 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1251 return __sock_create(family
, type
, protocol
, res
, 1);
1254 asmlinkage
long sys_socket(int family
, int type
, int protocol
)
1257 struct socket
*sock
;
1259 retval
= sock_create(family
, type
, protocol
, &sock
);
1263 retval
= sock_map_fd(sock
);
1268 /* It may be already another descriptor 8) Not kernel problem. */
1277 * Create a pair of connected sockets.
1280 asmlinkage
long sys_socketpair(int family
, int type
, int protocol
,
1281 int __user
*usockvec
)
1283 struct socket
*sock1
, *sock2
;
1287 * Obtain the first socket and check if the underlying protocol
1288 * supports the socketpair call.
1291 err
= sock_create(family
, type
, protocol
, &sock1
);
1295 err
= sock_create(family
, type
, protocol
, &sock2
);
1299 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1301 goto out_release_both
;
1305 err
= sock_map_fd(sock1
);
1307 goto out_release_both
;
1310 err
= sock_map_fd(sock2
);
1315 /* fd1 and fd2 may be already another descriptors.
1316 * Not kernel problem.
1319 err
= put_user(fd1
, &usockvec
[0]);
1321 err
= put_user(fd2
, &usockvec
[1]);
1330 sock_release(sock2
);
1335 sock_release(sock2
);
1337 sock_release(sock1
);
1343 * Bind a name to a socket. Nothing much to do here since it's
1344 * the protocol's responsibility to handle the local address.
1346 * We move the socket address to kernel space before we call
1347 * the protocol layer (having also checked the address is ok).
1350 asmlinkage
long sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1352 struct socket
*sock
;
1353 char address
[MAX_SOCK_ADDR
];
1354 int err
, fput_needed
;
1356 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1358 err
= move_addr_to_kernel(umyaddr
, addrlen
, address
);
1360 err
= security_socket_bind(sock
,
1361 (struct sockaddr
*)address
,
1364 err
= sock
->ops
->bind(sock
,
1368 fput_light(sock
->file
, fput_needed
);
1374 * Perform a listen. Basically, we allow the protocol to do anything
1375 * necessary for a listen, and if that works, we mark the socket as
1376 * ready for listening.
1379 int sysctl_somaxconn
= SOMAXCONN
;
1381 asmlinkage
long sys_listen(int fd
, int backlog
)
1383 struct socket
*sock
;
1384 int err
, fput_needed
;
1386 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1388 if ((unsigned)backlog
> sysctl_somaxconn
)
1389 backlog
= sysctl_somaxconn
;
1391 err
= security_socket_listen(sock
, backlog
);
1393 err
= sock
->ops
->listen(sock
, backlog
);
1395 fput_light(sock
->file
, fput_needed
);
1401 * For accept, we attempt to create a new socket, set up the link
1402 * with the client, wake up the client, then return the new
1403 * connected fd. We collect the address of the connector in kernel
1404 * space and move it to user at the very end. This is unclean because
1405 * we open the socket then return an error.
1407 * 1003.1g adds the ability to recvmsg() to query connection pending
1408 * status to recvmsg. We need to add that support in a way thats
1409 * clean when we restucture accept also.
1412 asmlinkage
long sys_accept(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1413 int __user
*upeer_addrlen
)
1415 struct socket
*sock
, *newsock
;
1416 struct file
*newfile
;
1417 int err
, len
, newfd
, fput_needed
;
1418 char address
[MAX_SOCK_ADDR
];
1420 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1425 if (!(newsock
= sock_alloc()))
1428 newsock
->type
= sock
->type
;
1429 newsock
->ops
= sock
->ops
;
1432 * We don't need try_module_get here, as the listening socket (sock)
1433 * has the protocol module (sock->ops->owner) held.
1435 __module_get(newsock
->ops
->owner
);
1437 newfd
= sock_alloc_fd(&newfile
);
1438 if (unlikely(newfd
< 0)) {
1440 sock_release(newsock
);
1444 err
= sock_attach_fd(newsock
, newfile
);
1448 err
= security_socket_accept(sock
, newsock
);
1452 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1456 if (upeer_sockaddr
) {
1457 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)address
,
1459 err
= -ECONNABORTED
;
1462 err
= move_addr_to_user(address
, len
, upeer_sockaddr
,
1468 /* File flags are not inherited via accept() unlike another OSes. */
1470 fd_install(newfd
, newfile
);
1473 security_socket_post_accept(sock
, newsock
);
1476 fput_light(sock
->file
, fput_needed
);
1481 put_unused_fd(newfd
);
1486 * Attempt to connect to a socket with the server address. The address
1487 * is in user space so we verify it is OK and move it to kernel space.
1489 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1492 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1493 * other SEQPACKET protocols that take time to connect() as it doesn't
1494 * include the -EINPROGRESS status for such sockets.
1497 asmlinkage
long sys_connect(int fd
, struct sockaddr __user
*uservaddr
,
1500 struct socket
*sock
;
1501 char address
[MAX_SOCK_ADDR
];
1502 int err
, fput_needed
;
1504 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1507 err
= move_addr_to_kernel(uservaddr
, addrlen
, address
);
1512 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1516 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
1517 sock
->file
->f_flags
);
1519 fput_light(sock
->file
, fput_needed
);
1525 * Get the local address ('name') of a socket object. Move the obtained
1526 * name to user space.
1529 asmlinkage
long sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
1530 int __user
*usockaddr_len
)
1532 struct socket
*sock
;
1533 char address
[MAX_SOCK_ADDR
];
1534 int len
, err
, fput_needed
;
1536 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1540 err
= security_socket_getsockname(sock
);
1544 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
, 0);
1547 err
= move_addr_to_user(address
, len
, usockaddr
, usockaddr_len
);
1550 fput_light(sock
->file
, fput_needed
);
1556 * Get the remote address ('name') of a socket object. Move the obtained
1557 * name to user space.
1560 asmlinkage
long sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
1561 int __user
*usockaddr_len
)
1563 struct socket
*sock
;
1564 char address
[MAX_SOCK_ADDR
];
1565 int len
, err
, fput_needed
;
1567 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1569 err
= security_socket_getpeername(sock
);
1571 fput_light(sock
->file
, fput_needed
);
1576 sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &len
,
1579 err
= move_addr_to_user(address
, len
, usockaddr
,
1581 fput_light(sock
->file
, fput_needed
);
1587 * Send a datagram to a given address. We move the address into kernel
1588 * space and check the user space data area is readable before invoking
1592 asmlinkage
long sys_sendto(int fd
, void __user
*buff
, size_t len
,
1593 unsigned flags
, struct sockaddr __user
*addr
,
1596 struct socket
*sock
;
1597 char address
[MAX_SOCK_ADDR
];
1602 struct file
*sock_file
;
1604 sock_file
= fget_light(fd
, &fput_needed
);
1608 sock
= sock_from_file(sock_file
, &err
);
1611 iov
.iov_base
= buff
;
1613 msg
.msg_name
= NULL
;
1616 msg
.msg_control
= NULL
;
1617 msg
.msg_controllen
= 0;
1618 msg
.msg_namelen
= 0;
1620 err
= move_addr_to_kernel(addr
, addr_len
, address
);
1623 msg
.msg_name
= address
;
1624 msg
.msg_namelen
= addr_len
;
1626 if (sock
->file
->f_flags
& O_NONBLOCK
)
1627 flags
|= MSG_DONTWAIT
;
1628 msg
.msg_flags
= flags
;
1629 err
= sock_sendmsg(sock
, &msg
, len
);
1632 fput_light(sock_file
, fput_needed
);
1637 * Send a datagram down a socket.
1640 asmlinkage
long sys_send(int fd
, void __user
*buff
, size_t len
, unsigned flags
)
1642 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1646 * Receive a frame from the socket and optionally record the address of the
1647 * sender. We verify the buffers are writable and if needed move the
1648 * sender address from kernel to user space.
1651 asmlinkage
long sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
,
1652 unsigned flags
, struct sockaddr __user
*addr
,
1653 int __user
*addr_len
)
1655 struct socket
*sock
;
1658 char address
[MAX_SOCK_ADDR
];
1660 struct file
*sock_file
;
1663 sock_file
= fget_light(fd
, &fput_needed
);
1667 sock
= sock_from_file(sock_file
, &err
);
1671 msg
.msg_control
= NULL
;
1672 msg
.msg_controllen
= 0;
1676 iov
.iov_base
= ubuf
;
1677 msg
.msg_name
= address
;
1678 msg
.msg_namelen
= MAX_SOCK_ADDR
;
1679 if (sock
->file
->f_flags
& O_NONBLOCK
)
1680 flags
|= MSG_DONTWAIT
;
1681 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1683 if (err
>= 0 && addr
!= NULL
) {
1684 err2
= move_addr_to_user(address
, msg
.msg_namelen
, addr
, addr_len
);
1689 fput_light(sock_file
, fput_needed
);
1694 * Receive a datagram from a socket.
1697 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1700 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1704 * Set a socket option. Because we don't know the option lengths we have
1705 * to pass the user mode parameter for the protocols to sort out.
1708 asmlinkage
long sys_setsockopt(int fd
, int level
, int optname
,
1709 char __user
*optval
, int optlen
)
1711 int err
, fput_needed
;
1712 struct socket
*sock
;
1717 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1719 err
= security_socket_setsockopt(sock
, level
, optname
);
1723 if (level
== SOL_SOCKET
)
1725 sock_setsockopt(sock
, level
, optname
, optval
,
1729 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1732 fput_light(sock
->file
, fput_needed
);
1738 * Get a socket option. Because we don't know the option lengths we have
1739 * to pass a user mode parameter for the protocols to sort out.
1742 asmlinkage
long sys_getsockopt(int fd
, int level
, int optname
,
1743 char __user
*optval
, int __user
*optlen
)
1745 int err
, fput_needed
;
1746 struct socket
*sock
;
1748 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1750 err
= security_socket_getsockopt(sock
, level
, optname
);
1754 if (level
== SOL_SOCKET
)
1756 sock_getsockopt(sock
, level
, optname
, optval
,
1760 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1763 fput_light(sock
->file
, fput_needed
);
1769 * Shutdown a socket.
1772 asmlinkage
long sys_shutdown(int fd
, int how
)
1774 int err
, fput_needed
;
1775 struct socket
*sock
;
1777 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1779 err
= security_socket_shutdown(sock
, how
);
1781 err
= sock
->ops
->shutdown(sock
, how
);
1782 fput_light(sock
->file
, fput_needed
);
1787 /* A couple of helpful macros for getting the address of the 32/64 bit
1788 * fields which are the same type (int / unsigned) on our platforms.
1790 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1791 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1792 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1795 * BSD sendmsg interface
1798 asmlinkage
long sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
1800 struct compat_msghdr __user
*msg_compat
=
1801 (struct compat_msghdr __user
*)msg
;
1802 struct socket
*sock
;
1803 char address
[MAX_SOCK_ADDR
];
1804 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1805 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1806 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1807 /* 20 is size of ipv6_pktinfo */
1808 unsigned char *ctl_buf
= ctl
;
1809 struct msghdr msg_sys
;
1810 int err
, ctl_len
, iov_size
, total_len
;
1814 if (MSG_CMSG_COMPAT
& flags
) {
1815 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1818 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1821 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1825 /* do not move before msg_sys is valid */
1827 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1830 /* Check whether to allocate the iovec area */
1832 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1833 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1834 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1839 /* This will also move the address data into kernel space */
1840 if (MSG_CMSG_COMPAT
& flags
) {
1841 err
= verify_compat_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1843 err
= verify_iovec(&msg_sys
, iov
, address
, VERIFY_READ
);
1850 if (msg_sys
.msg_controllen
> INT_MAX
)
1852 ctl_len
= msg_sys
.msg_controllen
;
1853 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1855 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1859 ctl_buf
= msg_sys
.msg_control
;
1860 ctl_len
= msg_sys
.msg_controllen
;
1861 } else if (ctl_len
) {
1862 if (ctl_len
> sizeof(ctl
)) {
1863 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1864 if (ctl_buf
== NULL
)
1869 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1870 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1871 * checking falls down on this.
1873 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1876 msg_sys
.msg_control
= ctl_buf
;
1878 msg_sys
.msg_flags
= flags
;
1880 if (sock
->file
->f_flags
& O_NONBLOCK
)
1881 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1882 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1886 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1888 if (iov
!= iovstack
)
1889 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1891 fput_light(sock
->file
, fput_needed
);
1897 * BSD recvmsg interface
1900 asmlinkage
long sys_recvmsg(int fd
, struct msghdr __user
*msg
,
1903 struct compat_msghdr __user
*msg_compat
=
1904 (struct compat_msghdr __user
*)msg
;
1905 struct socket
*sock
;
1906 struct iovec iovstack
[UIO_FASTIOV
];
1907 struct iovec
*iov
= iovstack
;
1908 struct msghdr msg_sys
;
1909 unsigned long cmsg_ptr
;
1910 int err
, iov_size
, total_len
, len
;
1913 /* kernel mode address */
1914 char addr
[MAX_SOCK_ADDR
];
1916 /* user mode address pointers */
1917 struct sockaddr __user
*uaddr
;
1918 int __user
*uaddr_len
;
1920 if (MSG_CMSG_COMPAT
& flags
) {
1921 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1924 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1927 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1932 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1935 /* Check whether to allocate the iovec area */
1937 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1938 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1939 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1945 * Save the user-mode address (verify_iovec will change the
1946 * kernel msghdr to use the kernel address space)
1949 uaddr
= (void __user
*)msg_sys
.msg_name
;
1950 uaddr_len
= COMPAT_NAMELEN(msg
);
1951 if (MSG_CMSG_COMPAT
& flags
) {
1952 err
= verify_compat_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1954 err
= verify_iovec(&msg_sys
, iov
, addr
, VERIFY_WRITE
);
1959 cmsg_ptr
= (unsigned long)msg_sys
.msg_control
;
1960 msg_sys
.msg_flags
= 0;
1961 if (MSG_CMSG_COMPAT
& flags
)
1962 msg_sys
.msg_flags
= MSG_CMSG_COMPAT
;
1964 if (sock
->file
->f_flags
& O_NONBLOCK
)
1965 flags
|= MSG_DONTWAIT
;
1966 err
= sock_recvmsg(sock
, &msg_sys
, total_len
, flags
);
1971 if (uaddr
!= NULL
) {
1972 err
= move_addr_to_user(addr
, msg_sys
.msg_namelen
, uaddr
,
1977 err
= __put_user((msg_sys
.msg_flags
& ~MSG_CMSG_COMPAT
),
1981 if (MSG_CMSG_COMPAT
& flags
)
1982 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1983 &msg_compat
->msg_controllen
);
1985 err
= __put_user((unsigned long)msg_sys
.msg_control
- cmsg_ptr
,
1986 &msg
->msg_controllen
);
1992 if (iov
!= iovstack
)
1993 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1995 fput_light(sock
->file
, fput_needed
);
2000 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2002 /* Argument list sizes for sys_socketcall */
2003 #define AL(x) ((x) * sizeof(unsigned long))
2004 static const unsigned char nargs
[18]={
2005 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2006 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2007 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
2013 * System call vectors.
2015 * Argument checking cleaned up. Saved 20% in size.
2016 * This function doesn't need to set the kernel lock because
2017 * it is set by the callees.
2020 asmlinkage
long sys_socketcall(int call
, unsigned long __user
*args
)
2023 unsigned long a0
, a1
;
2026 if (call
< 1 || call
> SYS_RECVMSG
)
2029 /* copy_from_user should be SMP safe. */
2030 if (copy_from_user(a
, args
, nargs
[call
]))
2033 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2042 err
= sys_socket(a0
, a1
, a
[2]);
2045 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2048 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2051 err
= sys_listen(a0
, a1
);
2055 sys_accept(a0
, (struct sockaddr __user
*)a1
,
2056 (int __user
*)a
[2]);
2058 case SYS_GETSOCKNAME
:
2060 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2061 (int __user
*)a
[2]);
2063 case SYS_GETPEERNAME
:
2065 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2066 (int __user
*)a
[2]);
2068 case SYS_SOCKETPAIR
:
2069 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2072 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2075 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2076 (struct sockaddr __user
*)a
[4], a
[5]);
2079 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2082 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2083 (struct sockaddr __user
*)a
[4],
2084 (int __user
*)a
[5]);
2087 err
= sys_shutdown(a0
, a1
);
2089 case SYS_SETSOCKOPT
:
2090 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2092 case SYS_GETSOCKOPT
:
2094 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2095 (int __user
*)a
[4]);
2098 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2101 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2110 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2113 * This function is called by a protocol handler that wants to
2114 * advertise its address family, and have it linked into the
2118 int sock_register(struct net_proto_family
*ops
)
2122 if (ops
->family
>= NPROTO
) {
2123 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2127 net_family_write_lock();
2129 if (net_families
[ops
->family
] == NULL
) {
2130 net_families
[ops
->family
] = ops
;
2133 net_family_write_unlock();
2134 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2139 * This function is called by a protocol handler that wants to
2140 * remove its address family, and have it unlinked from the
2144 int sock_unregister(int family
)
2146 if (family
< 0 || family
>= NPROTO
)
2149 net_family_write_lock();
2150 net_families
[family
] = NULL
;
2151 net_family_write_unlock();
2152 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2156 static int __init
sock_init(void)
2159 * Initialize sock SLAB cache.
2165 * Initialize skbuff SLAB cache
2170 * Initialize the protocols module.
2174 register_filesystem(&sock_fs_type
);
2175 sock_mnt
= kern_mount(&sock_fs_type
);
2177 /* The real protocol initialization is performed in later initcalls.
2180 #ifdef CONFIG_NETFILTER
2187 core_initcall(sock_init
); /* early initcall */
2189 #ifdef CONFIG_PROC_FS
2190 void socket_seq_show(struct seq_file
*seq
)
2195 for_each_possible_cpu(cpu
)
2196 counter
+= per_cpu(sockets_in_use
, cpu
);
2198 /* It can be negative, by the way. 8) */
2202 seq_printf(seq
, "sockets: used %d\n", counter
);
2204 #endif /* CONFIG_PROC_FS */
2206 #ifdef CONFIG_COMPAT
2207 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2210 struct socket
*sock
= file
->private_data
;
2211 int ret
= -ENOIOCTLCMD
;
2213 if (sock
->ops
->compat_ioctl
)
2214 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2220 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
2222 return sock
->ops
->bind(sock
, addr
, addrlen
);
2225 int kernel_listen(struct socket
*sock
, int backlog
)
2227 return sock
->ops
->listen(sock
, backlog
);
2230 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
2232 struct sock
*sk
= sock
->sk
;
2235 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
2240 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
2242 sock_release(*newsock
);
2246 (*newsock
)->ops
= sock
->ops
;
2252 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
2255 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
2258 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
2261 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
2264 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
2267 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
2270 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
2271 char *optval
, int *optlen
)
2273 mm_segment_t oldfs
= get_fs();
2277 if (level
== SOL_SOCKET
)
2278 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2280 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2286 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
2287 char *optval
, int optlen
)
2289 mm_segment_t oldfs
= get_fs();
2293 if (level
== SOL_SOCKET
)
2294 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2296 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2302 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
2303 size_t size
, int flags
)
2305 if (sock
->ops
->sendpage
)
2306 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
2308 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
2311 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
2313 mm_segment_t oldfs
= get_fs();
2317 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
2323 /* ABI emulation layers need these two */
2324 EXPORT_SYMBOL(move_addr_to_kernel
);
2325 EXPORT_SYMBOL(move_addr_to_user
);
2326 EXPORT_SYMBOL(sock_create
);
2327 EXPORT_SYMBOL(sock_create_kern
);
2328 EXPORT_SYMBOL(sock_create_lite
);
2329 EXPORT_SYMBOL(sock_map_fd
);
2330 EXPORT_SYMBOL(sock_recvmsg
);
2331 EXPORT_SYMBOL(sock_register
);
2332 EXPORT_SYMBOL(sock_release
);
2333 EXPORT_SYMBOL(sock_sendmsg
);
2334 EXPORT_SYMBOL(sock_unregister
);
2335 EXPORT_SYMBOL(sock_wake_async
);
2336 EXPORT_SYMBOL(sockfd_lookup
);
2337 EXPORT_SYMBOL(kernel_sendmsg
);
2338 EXPORT_SYMBOL(kernel_recvmsg
);
2339 EXPORT_SYMBOL(kernel_bind
);
2340 EXPORT_SYMBOL(kernel_listen
);
2341 EXPORT_SYMBOL(kernel_accept
);
2342 EXPORT_SYMBOL(kernel_connect
);
2343 EXPORT_SYMBOL(kernel_getsockname
);
2344 EXPORT_SYMBOL(kernel_getpeername
);
2345 EXPORT_SYMBOL(kernel_getsockopt
);
2346 EXPORT_SYMBOL(kernel_setsockopt
);
2347 EXPORT_SYMBOL(kernel_sendpage
);
2348 EXPORT_SYMBOL(kernel_sock_ioctl
);