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/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/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>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
108 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
109 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
110 unsigned long nr_segs
, loff_t pos
);
111 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
112 unsigned long nr_segs
, loff_t pos
);
113 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
115 static int sock_close(struct inode
*inode
, struct file
*file
);
116 static unsigned int sock_poll(struct file
*file
,
117 struct poll_table_struct
*wait
);
118 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
120 static long compat_sock_ioctl(struct file
*file
,
121 unsigned int cmd
, unsigned long arg
);
123 static int sock_fasync(int fd
, struct file
*filp
, int on
);
124 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
125 int offset
, size_t size
, loff_t
*ppos
, int more
);
126 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
127 struct pipe_inode_info
*pipe
, size_t len
,
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
135 static const struct file_operations socket_file_ops
= {
136 .owner
= THIS_MODULE
,
138 .aio_read
= sock_aio_read
,
139 .aio_write
= sock_aio_write
,
141 .unlocked_ioctl
= sock_ioctl
,
143 .compat_ioctl
= compat_sock_ioctl
,
146 .open
= sock_no_open
, /* special open code to disallow open via /proc */
147 .release
= sock_close
,
148 .fasync
= sock_fasync
,
149 .sendpage
= sock_sendpage
,
150 .splice_write
= generic_splice_sendpage
,
151 .splice_read
= sock_splice_read
,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock
);
159 static const struct net_proto_family
*net_families
[NPROTO
] __read_mostly
;
162 * Statistics counters of the socket lists
165 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
173 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
174 16 for IP, 16 for IPX,
177 must be at least one bigger than
178 the AF_UNIX size (see net/unix/af_unix.c
183 * move_addr_to_kernel - copy a socket address into kernel space
184 * @uaddr: Address in user space
185 * @kaddr: Address in kernel space
186 * @ulen: Length in user space
188 * The address is copied into kernel space. If the provided address is
189 * too long an error code of -EINVAL is returned. If the copy gives
190 * invalid addresses -EFAULT is returned. On a success 0 is returned.
193 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr
*kaddr
)
195 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
199 if (copy_from_user(kaddr
, uaddr
, ulen
))
201 return audit_sockaddr(ulen
, kaddr
);
205 * move_addr_to_user - copy an address to user space
206 * @kaddr: kernel space address
207 * @klen: length of address in kernel
208 * @uaddr: user space address
209 * @ulen: pointer to user length field
211 * The value pointed to by ulen on entry is the buffer length available.
212 * This is overwritten with the buffer space used. -EINVAL is returned
213 * if an overlong buffer is specified or a negative buffer size. -EFAULT
214 * is returned if either the buffer or the length field are not
216 * After copying the data up to the limit the user specifies, the true
217 * length of the data is written over the length limit the user
218 * specified. Zero is returned for a success.
221 int move_addr_to_user(struct sockaddr
*kaddr
, int klen
, void __user
*uaddr
,
227 err
= get_user(len
, ulen
);
232 if (len
< 0 || len
> sizeof(struct sockaddr_storage
))
235 if (audit_sockaddr(klen
, kaddr
))
237 if (copy_to_user(uaddr
, kaddr
, len
))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen
, ulen
);
247 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
249 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
251 struct socket_alloc
*ei
;
253 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
256 ei
->socket
.wq
= kmalloc(sizeof(struct socket_wq
), GFP_KERNEL
);
257 if (!ei
->socket
.wq
) {
258 kmem_cache_free(sock_inode_cachep
, ei
);
261 init_waitqueue_head(&ei
->socket
.wq
->wait
);
262 ei
->socket
.wq
->fasync_list
= NULL
;
264 ei
->socket
.state
= SS_UNCONNECTED
;
265 ei
->socket
.flags
= 0;
266 ei
->socket
.ops
= NULL
;
267 ei
->socket
.sk
= NULL
;
268 ei
->socket
.file
= NULL
;
270 return &ei
->vfs_inode
;
274 static void wq_free_rcu(struct rcu_head
*head
)
276 struct socket_wq
*wq
= container_of(head
, struct socket_wq
, rcu
);
281 static void sock_destroy_inode(struct inode
*inode
)
283 struct socket_alloc
*ei
;
285 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
286 call_rcu(&ei
->socket
.wq
->rcu
, wq_free_rcu
);
287 kmem_cache_free(sock_inode_cachep
, ei
);
290 static void init_once(void *foo
)
292 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
294 inode_init_once(&ei
->vfs_inode
);
297 static int init_inodecache(void)
299 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
300 sizeof(struct socket_alloc
),
302 (SLAB_HWCACHE_ALIGN
|
303 SLAB_RECLAIM_ACCOUNT
|
306 if (sock_inode_cachep
== NULL
)
311 static const struct super_operations sockfs_ops
= {
312 .alloc_inode
= sock_alloc_inode
,
313 .destroy_inode
=sock_destroy_inode
,
314 .statfs
= simple_statfs
,
317 static int sockfs_get_sb(struct file_system_type
*fs_type
,
318 int flags
, const char *dev_name
, void *data
,
319 struct vfsmount
*mnt
)
321 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
325 static struct vfsmount
*sock_mnt __read_mostly
;
327 static struct file_system_type sock_fs_type
= {
329 .get_sb
= sockfs_get_sb
,
330 .kill_sb
= kill_anon_super
,
334 * sockfs_dname() is called from d_path().
336 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
338 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
339 dentry
->d_inode
->i_ino
);
342 static const struct dentry_operations sockfs_dentry_operations
= {
343 .d_dname
= sockfs_dname
,
347 * Obtains the first available file descriptor and sets it up for use.
349 * These functions create file structures and maps them to fd space
350 * of the current process. On success it returns file descriptor
351 * and file struct implicitly stored in sock->file.
352 * Note that another thread may close file descriptor before we return
353 * from this function. We use the fact that now we do not refer
354 * to socket after mapping. If one day we will need it, this
355 * function will increment ref. count on file by 1.
357 * In any case returned fd MAY BE not valid!
358 * This race condition is unavoidable
359 * with shared fd spaces, we cannot solve it inside kernel,
360 * but we take care of internal coherence yet.
363 static int sock_alloc_file(struct socket
*sock
, struct file
**f
, int flags
)
365 struct qstr name
= { .name
= "" };
370 fd
= get_unused_fd_flags(flags
);
371 if (unlikely(fd
< 0))
374 path
.dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &name
);
375 if (unlikely(!path
.dentry
)) {
379 path
.mnt
= mntget(sock_mnt
);
381 path
.dentry
->d_op
= &sockfs_dentry_operations
;
382 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
383 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
385 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
387 if (unlikely(!file
)) {
388 /* drop dentry, keep inode */
389 atomic_inc(&path
.dentry
->d_inode
->i_count
);
396 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
398 file
->private_data
= sock
;
404 int sock_map_fd(struct socket
*sock
, int flags
)
406 struct file
*newfile
;
407 int fd
= sock_alloc_file(sock
, &newfile
, flags
);
410 fd_install(fd
, newfile
);
415 static struct socket
*sock_from_file(struct file
*file
, int *err
)
417 if (file
->f_op
== &socket_file_ops
)
418 return file
->private_data
; /* set in sock_map_fd */
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 kmemcheck_annotate_bitfield(sock
, type
);
490 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
491 inode
->i_uid
= current_fsuid();
492 inode
->i_gid
= current_fsgid();
494 percpu_add(sockets_in_use
, 1);
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
->wq
->fasync_list
)
534 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
536 percpu_sub(sockets_in_use
, 1);
538 iput(SOCK_INODE(sock
));
544 int sock_tx_timestamp(struct msghdr
*msg
, struct sock
*sk
,
545 union skb_shared_tx
*shtx
)
548 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
550 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
554 EXPORT_SYMBOL(sock_tx_timestamp
);
556 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
557 struct msghdr
*msg
, size_t size
)
559 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
562 sock_update_classid(sock
->sk
);
569 err
= security_socket_sendmsg(sock
, msg
, size
);
573 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
576 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
579 struct sock_iocb siocb
;
582 init_sync_kiocb(&iocb
, NULL
);
583 iocb
.private = &siocb
;
584 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
585 if (-EIOCBQUEUED
== ret
)
586 ret
= wait_on_sync_kiocb(&iocb
);
590 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
591 struct kvec
*vec
, size_t num
, size_t size
)
593 mm_segment_t oldfs
= get_fs();
598 * the following is safe, since for compiler definitions of kvec and
599 * iovec are identical, yielding the same in-core layout and alignment
601 msg
->msg_iov
= (struct iovec
*)vec
;
602 msg
->msg_iovlen
= num
;
603 result
= sock_sendmsg(sock
, msg
, size
);
608 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
611 *ts
= ktime_to_timespec(kt
);
619 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
621 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
624 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
625 struct timespec ts
[3];
627 struct skb_shared_hwtstamps
*shhwtstamps
=
630 /* Race occurred between timestamp enabling and packet
631 receiving. Fill in the current time for now. */
632 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
633 __net_timestamp(skb
);
635 if (need_software_tstamp
) {
636 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
638 skb_get_timestamp(skb
, &tv
);
639 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
642 skb_get_timestampns(skb
, &ts
[0]);
643 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
644 sizeof(ts
[0]), &ts
[0]);
649 memset(ts
, 0, sizeof(ts
));
650 if (skb
->tstamp
.tv64
&&
651 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
652 skb_get_timestampns(skb
, ts
+ 0);
656 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
657 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
659 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
660 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
664 put_cmsg(msg
, SOL_SOCKET
,
665 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
668 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
670 inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
672 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
673 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
674 sizeof(__u32
), &skb
->dropcount
);
677 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
680 sock_recv_timestamp(msg
, sk
, skb
);
681 sock_recv_drops(msg
, sk
, skb
);
683 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
685 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
686 struct msghdr
*msg
, size_t size
, int flags
)
688 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
690 sock_update_classid(sock
->sk
);
698 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
701 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
702 struct msghdr
*msg
, size_t size
, int flags
)
704 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
706 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
709 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
710 size_t size
, int flags
)
713 struct sock_iocb siocb
;
716 init_sync_kiocb(&iocb
, NULL
);
717 iocb
.private = &siocb
;
718 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
719 if (-EIOCBQUEUED
== ret
)
720 ret
= wait_on_sync_kiocb(&iocb
);
724 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
725 size_t size
, int flags
)
728 struct sock_iocb siocb
;
731 init_sync_kiocb(&iocb
, NULL
);
732 iocb
.private = &siocb
;
733 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
734 if (-EIOCBQUEUED
== ret
)
735 ret
= wait_on_sync_kiocb(&iocb
);
739 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
740 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
742 mm_segment_t oldfs
= get_fs();
747 * the following is safe, since for compiler definitions of kvec and
748 * iovec are identical, yielding the same in-core layout and alignment
750 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
751 result
= sock_recvmsg(sock
, msg
, size
, flags
);
756 static void sock_aio_dtor(struct kiocb
*iocb
)
758 kfree(iocb
->private);
761 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
762 int offset
, size_t size
, loff_t
*ppos
, int more
)
767 sock
= file
->private_data
;
769 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
773 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
776 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
777 struct pipe_inode_info
*pipe
, size_t len
,
780 struct socket
*sock
= file
->private_data
;
782 if (unlikely(!sock
->ops
->splice_read
))
785 sock_update_classid(sock
->sk
);
787 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
790 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
791 struct sock_iocb
*siocb
)
793 if (!is_sync_kiocb(iocb
)) {
794 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
797 iocb
->ki_dtor
= sock_aio_dtor
;
801 iocb
->private = siocb
;
805 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
806 struct file
*file
, const struct iovec
*iov
,
807 unsigned long nr_segs
)
809 struct socket
*sock
= file
->private_data
;
813 for (i
= 0; i
< nr_segs
; i
++)
814 size
+= iov
[i
].iov_len
;
816 msg
->msg_name
= NULL
;
817 msg
->msg_namelen
= 0;
818 msg
->msg_control
= NULL
;
819 msg
->msg_controllen
= 0;
820 msg
->msg_iov
= (struct iovec
*)iov
;
821 msg
->msg_iovlen
= nr_segs
;
822 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
824 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
827 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
828 unsigned long nr_segs
, loff_t pos
)
830 struct sock_iocb siocb
, *x
;
835 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
839 x
= alloc_sock_iocb(iocb
, &siocb
);
842 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
845 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
846 struct file
*file
, const struct iovec
*iov
,
847 unsigned long nr_segs
)
849 struct socket
*sock
= file
->private_data
;
853 for (i
= 0; i
< nr_segs
; i
++)
854 size
+= iov
[i
].iov_len
;
856 msg
->msg_name
= NULL
;
857 msg
->msg_namelen
= 0;
858 msg
->msg_control
= NULL
;
859 msg
->msg_controllen
= 0;
860 msg
->msg_iov
= (struct iovec
*)iov
;
861 msg
->msg_iovlen
= nr_segs
;
862 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
863 if (sock
->type
== SOCK_SEQPACKET
)
864 msg
->msg_flags
|= MSG_EOR
;
866 return __sock_sendmsg(iocb
, sock
, msg
, size
);
869 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
870 unsigned long nr_segs
, loff_t pos
)
872 struct sock_iocb siocb
, *x
;
877 x
= alloc_sock_iocb(iocb
, &siocb
);
881 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
885 * Atomic setting of ioctl hooks to avoid race
886 * with module unload.
889 static DEFINE_MUTEX(br_ioctl_mutex
);
890 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
) = NULL
;
892 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
894 mutex_lock(&br_ioctl_mutex
);
895 br_ioctl_hook
= hook
;
896 mutex_unlock(&br_ioctl_mutex
);
899 EXPORT_SYMBOL(brioctl_set
);
901 static DEFINE_MUTEX(vlan_ioctl_mutex
);
902 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
904 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
906 mutex_lock(&vlan_ioctl_mutex
);
907 vlan_ioctl_hook
= hook
;
908 mutex_unlock(&vlan_ioctl_mutex
);
911 EXPORT_SYMBOL(vlan_ioctl_set
);
913 static DEFINE_MUTEX(dlci_ioctl_mutex
);
914 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
916 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
918 mutex_lock(&dlci_ioctl_mutex
);
919 dlci_ioctl_hook
= hook
;
920 mutex_unlock(&dlci_ioctl_mutex
);
923 EXPORT_SYMBOL(dlci_ioctl_set
);
925 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
926 unsigned int cmd
, unsigned long arg
)
929 void __user
*argp
= (void __user
*)arg
;
931 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
934 * If this ioctl is unknown try to hand it down
937 if (err
== -ENOIOCTLCMD
)
938 err
= dev_ioctl(net
, cmd
, argp
);
944 * With an ioctl, arg may well be a user mode pointer, but we don't know
945 * what to do with it - that's up to the protocol still.
948 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
952 void __user
*argp
= (void __user
*)arg
;
956 sock
= file
->private_data
;
959 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
960 err
= dev_ioctl(net
, cmd
, argp
);
962 #ifdef CONFIG_WEXT_CORE
963 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
964 err
= dev_ioctl(net
, cmd
, argp
);
971 if (get_user(pid
, (int __user
*)argp
))
973 err
= f_setown(sock
->file
, pid
, 1);
977 err
= put_user(f_getown(sock
->file
),
986 request_module("bridge");
988 mutex_lock(&br_ioctl_mutex
);
990 err
= br_ioctl_hook(net
, cmd
, argp
);
991 mutex_unlock(&br_ioctl_mutex
);
996 if (!vlan_ioctl_hook
)
997 request_module("8021q");
999 mutex_lock(&vlan_ioctl_mutex
);
1000 if (vlan_ioctl_hook
)
1001 err
= vlan_ioctl_hook(net
, argp
);
1002 mutex_unlock(&vlan_ioctl_mutex
);
1007 if (!dlci_ioctl_hook
)
1008 request_module("dlci");
1010 mutex_lock(&dlci_ioctl_mutex
);
1011 if (dlci_ioctl_hook
)
1012 err
= dlci_ioctl_hook(cmd
, argp
);
1013 mutex_unlock(&dlci_ioctl_mutex
);
1016 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1022 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1025 struct socket
*sock
= NULL
;
1027 err
= security_socket_create(family
, type
, protocol
, 1);
1031 sock
= sock_alloc();
1038 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1051 /* No kernel lock held - perfect */
1052 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1054 struct socket
*sock
;
1057 * We can't return errors to poll, so it's either yes or no.
1059 sock
= file
->private_data
;
1060 return sock
->ops
->poll(file
, sock
, wait
);
1063 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1065 struct socket
*sock
= file
->private_data
;
1067 return sock
->ops
->mmap(file
, sock
, vma
);
1070 static int sock_close(struct inode
*inode
, struct file
*filp
)
1073 * It was possible the inode is NULL we were
1074 * closing an unfinished socket.
1078 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1081 sock_release(SOCKET_I(inode
));
1086 * Update the socket async list
1088 * Fasync_list locking strategy.
1090 * 1. fasync_list is modified only under process context socket lock
1091 * i.e. under semaphore.
1092 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1093 * or under socket lock
1096 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1098 struct socket
*sock
= filp
->private_data
;
1099 struct sock
*sk
= sock
->sk
;
1106 fasync_helper(fd
, filp
, on
, &sock
->wq
->fasync_list
);
1108 if (!sock
->wq
->fasync_list
)
1109 sock_reset_flag(sk
, SOCK_FASYNC
);
1111 sock_set_flag(sk
, SOCK_FASYNC
);
1117 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1119 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1121 struct socket_wq
*wq
;
1126 wq
= rcu_dereference(sock
->wq
);
1127 if (!wq
|| !wq
->fasync_list
) {
1132 case SOCK_WAKE_WAITD
:
1133 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1136 case SOCK_WAKE_SPACE
:
1137 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1142 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1145 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1151 static int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1152 struct socket
**res
, int kern
)
1155 struct socket
*sock
;
1156 const struct net_proto_family
*pf
;
1159 * Check protocol is in range
1161 if (family
< 0 || family
>= NPROTO
)
1162 return -EAFNOSUPPORT
;
1163 if (type
< 0 || type
>= SOCK_MAX
)
1168 This uglymoron is moved from INET layer to here to avoid
1169 deadlock in module load.
1171 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1175 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1181 err
= security_socket_create(family
, type
, protocol
, kern
);
1186 * Allocate the socket and allow the family to set things up. if
1187 * the protocol is 0, the family is instructed to select an appropriate
1190 sock
= sock_alloc();
1192 if (net_ratelimit())
1193 printk(KERN_WARNING
"socket: no more sockets\n");
1194 return -ENFILE
; /* Not exactly a match, but its the
1195 closest posix thing */
1200 #ifdef CONFIG_MODULES
1201 /* Attempt to load a protocol module if the find failed.
1203 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1204 * requested real, full-featured networking support upon configuration.
1205 * Otherwise module support will break!
1207 if (net_families
[family
] == NULL
)
1208 request_module("net-pf-%d", family
);
1212 pf
= rcu_dereference(net_families
[family
]);
1213 err
= -EAFNOSUPPORT
;
1218 * We will call the ->create function, that possibly is in a loadable
1219 * module, so we have to bump that loadable module refcnt first.
1221 if (!try_module_get(pf
->owner
))
1224 /* Now protected by module ref count */
1227 err
= pf
->create(net
, sock
, protocol
, kern
);
1229 goto out_module_put
;
1232 * Now to bump the refcnt of the [loadable] module that owns this
1233 * socket at sock_release time we decrement its refcnt.
1235 if (!try_module_get(sock
->ops
->owner
))
1236 goto out_module_busy
;
1239 * Now that we're done with the ->create function, the [loadable]
1240 * module can have its refcnt decremented
1242 module_put(pf
->owner
);
1243 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1245 goto out_sock_release
;
1251 err
= -EAFNOSUPPORT
;
1254 module_put(pf
->owner
);
1261 goto out_sock_release
;
1264 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1266 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1269 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1271 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1274 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1277 struct socket
*sock
;
1280 /* Check the SOCK_* constants for consistency. */
1281 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1282 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1283 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1284 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1286 flags
= type
& ~SOCK_TYPE_MASK
;
1287 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1289 type
&= SOCK_TYPE_MASK
;
1291 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1292 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1294 retval
= sock_create(family
, type
, protocol
, &sock
);
1298 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1303 /* It may be already another descriptor 8) Not kernel problem. */
1312 * Create a pair of connected sockets.
1315 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1316 int __user
*, usockvec
)
1318 struct socket
*sock1
, *sock2
;
1320 struct file
*newfile1
, *newfile2
;
1323 flags
= type
& ~SOCK_TYPE_MASK
;
1324 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1326 type
&= SOCK_TYPE_MASK
;
1328 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1329 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1332 * Obtain the first socket and check if the underlying protocol
1333 * supports the socketpair call.
1336 err
= sock_create(family
, type
, protocol
, &sock1
);
1340 err
= sock_create(family
, type
, protocol
, &sock2
);
1344 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1346 goto out_release_both
;
1348 fd1
= sock_alloc_file(sock1
, &newfile1
, flags
);
1349 if (unlikely(fd1
< 0)) {
1351 goto out_release_both
;
1354 fd2
= sock_alloc_file(sock2
, &newfile2
, flags
);
1355 if (unlikely(fd2
< 0)) {
1359 sock_release(sock2
);
1363 audit_fd_pair(fd1
, fd2
);
1364 fd_install(fd1
, newfile1
);
1365 fd_install(fd2
, newfile2
);
1366 /* fd1 and fd2 may be already another descriptors.
1367 * Not kernel problem.
1370 err
= put_user(fd1
, &usockvec
[0]);
1372 err
= put_user(fd2
, &usockvec
[1]);
1381 sock_release(sock2
);
1383 sock_release(sock1
);
1389 * Bind a name to a socket. Nothing much to do here since it's
1390 * the protocol's responsibility to handle the local address.
1392 * We move the socket address to kernel space before we call
1393 * the protocol layer (having also checked the address is ok).
1396 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1398 struct socket
*sock
;
1399 struct sockaddr_storage address
;
1400 int err
, fput_needed
;
1402 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1404 err
= move_addr_to_kernel(umyaddr
, addrlen
, (struct sockaddr
*)&address
);
1406 err
= security_socket_bind(sock
,
1407 (struct sockaddr
*)&address
,
1410 err
= sock
->ops
->bind(sock
,
1414 fput_light(sock
->file
, fput_needed
);
1420 * Perform a listen. Basically, we allow the protocol to do anything
1421 * necessary for a listen, and if that works, we mark the socket as
1422 * ready for listening.
1425 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1427 struct socket
*sock
;
1428 int err
, fput_needed
;
1431 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1433 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1434 if ((unsigned)backlog
> somaxconn
)
1435 backlog
= somaxconn
;
1437 err
= security_socket_listen(sock
, backlog
);
1439 err
= sock
->ops
->listen(sock
, backlog
);
1441 fput_light(sock
->file
, fput_needed
);
1447 * For accept, we attempt to create a new socket, set up the link
1448 * with the client, wake up the client, then return the new
1449 * connected fd. We collect the address of the connector in kernel
1450 * space and move it to user at the very end. This is unclean because
1451 * we open the socket then return an error.
1453 * 1003.1g adds the ability to recvmsg() to query connection pending
1454 * status to recvmsg. We need to add that support in a way thats
1455 * clean when we restucture accept also.
1458 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1459 int __user
*, upeer_addrlen
, int, flags
)
1461 struct socket
*sock
, *newsock
;
1462 struct file
*newfile
;
1463 int err
, len
, newfd
, fput_needed
;
1464 struct sockaddr_storage address
;
1466 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1469 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1470 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1472 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1477 if (!(newsock
= sock_alloc()))
1480 newsock
->type
= sock
->type
;
1481 newsock
->ops
= sock
->ops
;
1484 * We don't need try_module_get here, as the listening socket (sock)
1485 * has the protocol module (sock->ops->owner) held.
1487 __module_get(newsock
->ops
->owner
);
1489 newfd
= sock_alloc_file(newsock
, &newfile
, flags
);
1490 if (unlikely(newfd
< 0)) {
1492 sock_release(newsock
);
1496 err
= security_socket_accept(sock
, newsock
);
1500 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1504 if (upeer_sockaddr
) {
1505 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1507 err
= -ECONNABORTED
;
1510 err
= move_addr_to_user((struct sockaddr
*)&address
,
1511 len
, upeer_sockaddr
, upeer_addrlen
);
1516 /* File flags are not inherited via accept() unlike another OSes. */
1518 fd_install(newfd
, newfile
);
1522 fput_light(sock
->file
, fput_needed
);
1527 put_unused_fd(newfd
);
1531 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1532 int __user
*, upeer_addrlen
)
1534 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1538 * Attempt to connect to a socket with the server address. The address
1539 * is in user space so we verify it is OK and move it to kernel space.
1541 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1544 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1545 * other SEQPACKET protocols that take time to connect() as it doesn't
1546 * include the -EINPROGRESS status for such sockets.
1549 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1552 struct socket
*sock
;
1553 struct sockaddr_storage address
;
1554 int err
, fput_needed
;
1556 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1559 err
= move_addr_to_kernel(uservaddr
, addrlen
, (struct sockaddr
*)&address
);
1564 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1568 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1569 sock
->file
->f_flags
);
1571 fput_light(sock
->file
, fput_needed
);
1577 * Get the local address ('name') of a socket object. Move the obtained
1578 * name to user space.
1581 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1582 int __user
*, usockaddr_len
)
1584 struct socket
*sock
;
1585 struct sockaddr_storage address
;
1586 int len
, err
, fput_needed
;
1588 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1592 err
= security_socket_getsockname(sock
);
1596 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1599 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
, usockaddr_len
);
1602 fput_light(sock
->file
, fput_needed
);
1608 * Get the remote address ('name') of a socket object. Move the obtained
1609 * name to user space.
1612 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1613 int __user
*, usockaddr_len
)
1615 struct socket
*sock
;
1616 struct sockaddr_storage address
;
1617 int len
, err
, fput_needed
;
1619 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1621 err
= security_socket_getpeername(sock
);
1623 fput_light(sock
->file
, fput_needed
);
1628 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1631 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
,
1633 fput_light(sock
->file
, fput_needed
);
1639 * Send a datagram to a given address. We move the address into kernel
1640 * space and check the user space data area is readable before invoking
1644 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1645 unsigned, flags
, struct sockaddr __user
*, addr
,
1648 struct socket
*sock
;
1649 struct sockaddr_storage address
;
1655 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1659 iov
.iov_base
= buff
;
1661 msg
.msg_name
= NULL
;
1664 msg
.msg_control
= NULL
;
1665 msg
.msg_controllen
= 0;
1666 msg
.msg_namelen
= 0;
1668 err
= move_addr_to_kernel(addr
, addr_len
, (struct sockaddr
*)&address
);
1671 msg
.msg_name
= (struct sockaddr
*)&address
;
1672 msg
.msg_namelen
= addr_len
;
1674 if (sock
->file
->f_flags
& O_NONBLOCK
)
1675 flags
|= MSG_DONTWAIT
;
1676 msg
.msg_flags
= flags
;
1677 err
= sock_sendmsg(sock
, &msg
, len
);
1680 fput_light(sock
->file
, fput_needed
);
1686 * Send a datagram down a socket.
1689 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1692 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1696 * Receive a frame from the socket and optionally record the address of the
1697 * sender. We verify the buffers are writable and if needed move the
1698 * sender address from kernel to user space.
1701 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1702 unsigned, flags
, struct sockaddr __user
*, addr
,
1703 int __user
*, addr_len
)
1705 struct socket
*sock
;
1708 struct sockaddr_storage address
;
1712 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1716 msg
.msg_control
= NULL
;
1717 msg
.msg_controllen
= 0;
1721 iov
.iov_base
= ubuf
;
1722 msg
.msg_name
= (struct sockaddr
*)&address
;
1723 msg
.msg_namelen
= sizeof(address
);
1724 if (sock
->file
->f_flags
& O_NONBLOCK
)
1725 flags
|= MSG_DONTWAIT
;
1726 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1728 if (err
>= 0 && addr
!= NULL
) {
1729 err2
= move_addr_to_user((struct sockaddr
*)&address
,
1730 msg
.msg_namelen
, addr
, addr_len
);
1735 fput_light(sock
->file
, fput_needed
);
1741 * Receive a datagram from a socket.
1744 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1747 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1751 * Set a socket option. Because we don't know the option lengths we have
1752 * to pass the user mode parameter for the protocols to sort out.
1755 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1756 char __user
*, optval
, int, optlen
)
1758 int err
, fput_needed
;
1759 struct socket
*sock
;
1764 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1766 err
= security_socket_setsockopt(sock
, level
, optname
);
1770 if (level
== SOL_SOCKET
)
1772 sock_setsockopt(sock
, level
, optname
, optval
,
1776 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1779 fput_light(sock
->file
, fput_needed
);
1785 * Get a socket option. Because we don't know the option lengths we have
1786 * to pass a user mode parameter for the protocols to sort out.
1789 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1790 char __user
*, optval
, int __user
*, optlen
)
1792 int err
, fput_needed
;
1793 struct socket
*sock
;
1795 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1797 err
= security_socket_getsockopt(sock
, level
, optname
);
1801 if (level
== SOL_SOCKET
)
1803 sock_getsockopt(sock
, level
, optname
, optval
,
1807 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1810 fput_light(sock
->file
, fput_needed
);
1816 * Shutdown a socket.
1819 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1821 int err
, fput_needed
;
1822 struct socket
*sock
;
1824 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1826 err
= security_socket_shutdown(sock
, how
);
1828 err
= sock
->ops
->shutdown(sock
, how
);
1829 fput_light(sock
->file
, fput_needed
);
1834 /* A couple of helpful macros for getting the address of the 32/64 bit
1835 * fields which are the same type (int / unsigned) on our platforms.
1837 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1838 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1839 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1842 * BSD sendmsg interface
1845 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned, flags
)
1847 struct compat_msghdr __user
*msg_compat
=
1848 (struct compat_msghdr __user
*)msg
;
1849 struct socket
*sock
;
1850 struct sockaddr_storage address
;
1851 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1852 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1853 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1854 /* 20 is size of ipv6_pktinfo */
1855 unsigned char *ctl_buf
= ctl
;
1856 struct msghdr msg_sys
;
1857 int err
, ctl_len
, iov_size
, total_len
;
1861 if (MSG_CMSG_COMPAT
& flags
) {
1862 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1865 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1868 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1872 /* do not move before msg_sys is valid */
1874 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1877 /* Check whether to allocate the iovec area */
1879 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1880 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1881 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1886 /* This will also move the address data into kernel space */
1887 if (MSG_CMSG_COMPAT
& flags
) {
1888 err
= verify_compat_iovec(&msg_sys
, iov
,
1889 (struct sockaddr
*)&address
,
1892 err
= verify_iovec(&msg_sys
, iov
,
1893 (struct sockaddr
*)&address
,
1901 if (msg_sys
.msg_controllen
> INT_MAX
)
1903 ctl_len
= msg_sys
.msg_controllen
;
1904 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1906 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1910 ctl_buf
= msg_sys
.msg_control
;
1911 ctl_len
= msg_sys
.msg_controllen
;
1912 } else if (ctl_len
) {
1913 if (ctl_len
> sizeof(ctl
)) {
1914 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1915 if (ctl_buf
== NULL
)
1920 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1921 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1922 * checking falls down on this.
1924 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1927 msg_sys
.msg_control
= ctl_buf
;
1929 msg_sys
.msg_flags
= flags
;
1931 if (sock
->file
->f_flags
& O_NONBLOCK
)
1932 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1933 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1937 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1939 if (iov
!= iovstack
)
1940 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1942 fput_light(sock
->file
, fput_needed
);
1947 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1948 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
1950 struct compat_msghdr __user
*msg_compat
=
1951 (struct compat_msghdr __user
*)msg
;
1952 struct iovec iovstack
[UIO_FASTIOV
];
1953 struct iovec
*iov
= iovstack
;
1954 unsigned long cmsg_ptr
;
1955 int err
, iov_size
, total_len
, len
;
1957 /* kernel mode address */
1958 struct sockaddr_storage addr
;
1960 /* user mode address pointers */
1961 struct sockaddr __user
*uaddr
;
1962 int __user
*uaddr_len
;
1964 if (MSG_CMSG_COMPAT
& flags
) {
1965 if (get_compat_msghdr(msg_sys
, msg_compat
))
1968 else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
1972 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
1975 /* Check whether to allocate the iovec area */
1977 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
1978 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
1979 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1985 * Save the user-mode address (verify_iovec will change the
1986 * kernel msghdr to use the kernel address space)
1989 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
1990 uaddr_len
= COMPAT_NAMELEN(msg
);
1991 if (MSG_CMSG_COMPAT
& flags
) {
1992 err
= verify_compat_iovec(msg_sys
, iov
,
1993 (struct sockaddr
*)&addr
,
1996 err
= verify_iovec(msg_sys
, iov
,
1997 (struct sockaddr
*)&addr
,
2003 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2004 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2006 if (sock
->file
->f_flags
& O_NONBLOCK
)
2007 flags
|= MSG_DONTWAIT
;
2008 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2014 if (uaddr
!= NULL
) {
2015 err
= move_addr_to_user((struct sockaddr
*)&addr
,
2016 msg_sys
->msg_namelen
, uaddr
,
2021 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2025 if (MSG_CMSG_COMPAT
& flags
)
2026 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2027 &msg_compat
->msg_controllen
);
2029 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2030 &msg
->msg_controllen
);
2036 if (iov
!= iovstack
)
2037 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2043 * BSD recvmsg interface
2046 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2047 unsigned int, flags
)
2049 int fput_needed
, err
;
2050 struct msghdr msg_sys
;
2051 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2056 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2058 fput_light(sock
->file
, fput_needed
);
2064 * Linux recvmmsg interface
2067 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2068 unsigned int flags
, struct timespec
*timeout
)
2070 int fput_needed
, err
, datagrams
;
2071 struct socket
*sock
;
2072 struct mmsghdr __user
*entry
;
2073 struct compat_mmsghdr __user
*compat_entry
;
2074 struct msghdr msg_sys
;
2075 struct timespec end_time
;
2078 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2084 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2088 err
= sock_error(sock
->sk
);
2093 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2095 while (datagrams
< vlen
) {
2097 * No need to ask LSM for more than the first datagram.
2099 if (MSG_CMSG_COMPAT
& flags
) {
2100 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2101 &msg_sys
, flags
, datagrams
);
2104 err
= __put_user(err
, &compat_entry
->msg_len
);
2107 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2108 &msg_sys
, flags
, datagrams
);
2111 err
= put_user(err
, &entry
->msg_len
);
2119 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2120 if (flags
& MSG_WAITFORONE
)
2121 flags
|= MSG_DONTWAIT
;
2124 ktime_get_ts(timeout
);
2125 *timeout
= timespec_sub(end_time
, *timeout
);
2126 if (timeout
->tv_sec
< 0) {
2127 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2131 /* Timeout, return less than vlen datagrams */
2132 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2136 /* Out of band data, return right away */
2137 if (msg_sys
.msg_flags
& MSG_OOB
)
2142 fput_light(sock
->file
, fput_needed
);
2147 if (datagrams
!= 0) {
2149 * We may return less entries than requested (vlen) if the
2150 * sock is non block and there aren't enough datagrams...
2152 if (err
!= -EAGAIN
) {
2154 * ... or if recvmsg returns an error after we
2155 * received some datagrams, where we record the
2156 * error to return on the next call or if the
2157 * app asks about it using getsockopt(SO_ERROR).
2159 sock
->sk
->sk_err
= -err
;
2168 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2169 unsigned int, vlen
, unsigned int, flags
,
2170 struct timespec __user
*, timeout
)
2173 struct timespec timeout_sys
;
2176 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2178 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2181 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2183 if (datagrams
> 0 &&
2184 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2185 datagrams
= -EFAULT
;
2190 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2191 /* Argument list sizes for sys_socketcall */
2192 #define AL(x) ((x) * sizeof(unsigned long))
2193 static const unsigned char nargs
[20] = {
2194 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2195 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2196 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2203 * System call vectors.
2205 * Argument checking cleaned up. Saved 20% in size.
2206 * This function doesn't need to set the kernel lock because
2207 * it is set by the callees.
2210 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2213 unsigned long a0
, a1
;
2217 if (call
< 1 || call
> SYS_RECVMMSG
)
2221 if (len
> sizeof(a
))
2224 /* copy_from_user should be SMP safe. */
2225 if (copy_from_user(a
, args
, len
))
2228 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2235 err
= sys_socket(a0
, a1
, a
[2]);
2238 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2241 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2244 err
= sys_listen(a0
, a1
);
2247 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2248 (int __user
*)a
[2], 0);
2250 case SYS_GETSOCKNAME
:
2252 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2253 (int __user
*)a
[2]);
2255 case SYS_GETPEERNAME
:
2257 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2258 (int __user
*)a
[2]);
2260 case SYS_SOCKETPAIR
:
2261 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2264 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2267 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2268 (struct sockaddr __user
*)a
[4], a
[5]);
2271 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2274 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2275 (struct sockaddr __user
*)a
[4],
2276 (int __user
*)a
[5]);
2279 err
= sys_shutdown(a0
, a1
);
2281 case SYS_SETSOCKOPT
:
2282 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2284 case SYS_GETSOCKOPT
:
2286 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2287 (int __user
*)a
[4]);
2290 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2293 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2296 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2297 (struct timespec __user
*)a
[4]);
2300 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2301 (int __user
*)a
[2], a
[3]);
2310 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2313 * sock_register - add a socket protocol handler
2314 * @ops: description of protocol
2316 * This function is called by a protocol handler that wants to
2317 * advertise its address family, and have it linked into the
2318 * socket interface. The value ops->family coresponds to the
2319 * socket system call protocol family.
2321 int sock_register(const struct net_proto_family
*ops
)
2325 if (ops
->family
>= NPROTO
) {
2326 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2331 spin_lock(&net_family_lock
);
2332 if (net_families
[ops
->family
])
2335 net_families
[ops
->family
] = ops
;
2338 spin_unlock(&net_family_lock
);
2340 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2345 * sock_unregister - remove a protocol handler
2346 * @family: protocol family to remove
2348 * This function is called by a protocol handler that wants to
2349 * remove its address family, and have it unlinked from the
2350 * new socket creation.
2352 * If protocol handler is a module, then it can use module reference
2353 * counts to protect against new references. If protocol handler is not
2354 * a module then it needs to provide its own protection in
2355 * the ops->create routine.
2357 void sock_unregister(int family
)
2359 BUG_ON(family
< 0 || family
>= NPROTO
);
2361 spin_lock(&net_family_lock
);
2362 net_families
[family
] = NULL
;
2363 spin_unlock(&net_family_lock
);
2367 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2370 static int __init
sock_init(void)
2373 * Initialize sock SLAB cache.
2379 * Initialize skbuff SLAB cache
2384 * Initialize the protocols module.
2388 register_filesystem(&sock_fs_type
);
2389 sock_mnt
= kern_mount(&sock_fs_type
);
2391 /* The real protocol initialization is performed in later initcalls.
2394 #ifdef CONFIG_NETFILTER
2401 core_initcall(sock_init
); /* early initcall */
2403 #ifdef CONFIG_PROC_FS
2404 void socket_seq_show(struct seq_file
*seq
)
2409 for_each_possible_cpu(cpu
)
2410 counter
+= per_cpu(sockets_in_use
, cpu
);
2412 /* It can be negative, by the way. 8) */
2416 seq_printf(seq
, "sockets: used %d\n", counter
);
2418 #endif /* CONFIG_PROC_FS */
2420 #ifdef CONFIG_COMPAT
2421 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2422 unsigned int cmd
, struct compat_timeval __user
*up
)
2424 mm_segment_t old_fs
= get_fs();
2429 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2432 err
= put_user(ktv
.tv_sec
, &up
->tv_sec
);
2433 err
|= __put_user(ktv
.tv_usec
, &up
->tv_usec
);
2438 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2439 unsigned int cmd
, struct compat_timespec __user
*up
)
2441 mm_segment_t old_fs
= get_fs();
2442 struct timespec kts
;
2446 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2449 err
= put_user(kts
.tv_sec
, &up
->tv_sec
);
2450 err
|= __put_user(kts
.tv_nsec
, &up
->tv_nsec
);
2455 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2457 struct ifreq __user
*uifr
;
2460 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2461 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2464 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2468 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2474 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2476 struct compat_ifconf ifc32
;
2478 struct ifconf __user
*uifc
;
2479 struct compat_ifreq __user
*ifr32
;
2480 struct ifreq __user
*ifr
;
2484 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2487 if (ifc32
.ifcbuf
== 0) {
2491 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2493 size_t len
=((ifc32
.ifc_len
/ sizeof (struct compat_ifreq
)) + 1) *
2494 sizeof (struct ifreq
);
2495 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2497 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2498 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2499 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof (struct compat_ifreq
)) {
2500 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2506 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2509 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2513 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2517 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2519 i
+ sizeof (struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2520 i
+= sizeof (struct compat_ifreq
), j
+= sizeof (struct ifreq
)) {
2521 if (copy_in_user(ifr32
, ifr
, sizeof (struct compat_ifreq
)))
2527 if (ifc32
.ifcbuf
== 0) {
2528 /* Translate from 64-bit structure multiple to
2532 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2537 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2543 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2545 struct ifreq __user
*ifr
;
2549 ifr
= compat_alloc_user_space(sizeof(*ifr
));
2551 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2554 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2557 datap
= compat_ptr(data
);
2558 if (put_user(datap
, &ifr
->ifr_ifru
.ifru_data
))
2561 return dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2564 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2567 compat_uptr_t uptr32
;
2568 struct ifreq __user
*uifr
;
2570 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2571 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2574 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2577 uptr
= compat_ptr(uptr32
);
2579 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2582 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2585 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2586 struct compat_ifreq __user
*ifr32
)
2589 struct ifreq __user
*uifr
;
2590 mm_segment_t old_fs
;
2596 case SIOCBONDENSLAVE
:
2597 case SIOCBONDRELEASE
:
2598 case SIOCBONDSETHWADDR
:
2599 case SIOCBONDCHANGEACTIVE
:
2600 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2605 err
= dev_ioctl(net
, cmd
, &kifr
);
2609 case SIOCBONDSLAVEINFOQUERY
:
2610 case SIOCBONDINFOQUERY
:
2611 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2612 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2615 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2618 datap
= compat_ptr(data
);
2619 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2622 return dev_ioctl(net
, cmd
, uifr
);
2628 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2629 struct compat_ifreq __user
*u_ifreq32
)
2631 struct ifreq __user
*u_ifreq64
;
2632 char tmp_buf
[IFNAMSIZ
];
2633 void __user
*data64
;
2636 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2639 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2641 data64
= compat_ptr(data32
);
2643 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2645 /* Don't check these user accesses, just let that get trapped
2646 * in the ioctl handler instead.
2648 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2651 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2654 return dev_ioctl(net
, cmd
, u_ifreq64
);
2657 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2658 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2660 struct ifreq __user
*uifr
;
2663 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2664 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2667 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2678 case SIOCGIFBRDADDR
:
2679 case SIOCGIFDSTADDR
:
2680 case SIOCGIFNETMASK
:
2685 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2693 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2694 struct compat_ifreq __user
*uifr32
)
2697 struct compat_ifmap __user
*uifmap32
;
2698 mm_segment_t old_fs
;
2701 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2702 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2703 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2704 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2705 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2706 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2707 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2708 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2714 err
= dev_ioctl(net
, cmd
, (void __user
*)&ifr
);
2717 if (cmd
== SIOCGIFMAP
&& !err
) {
2718 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2719 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2720 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2721 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2722 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2723 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2724 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2731 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2734 compat_uptr_t uptr32
;
2735 struct ifreq __user
*uifr
;
2737 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2738 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2741 if (get_user(uptr32
, &uifr32
->ifr_data
))
2744 uptr
= compat_ptr(uptr32
);
2746 if (put_user(uptr
, &uifr
->ifr_data
))
2749 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
2754 struct sockaddr rt_dst
; /* target address */
2755 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2756 struct sockaddr rt_genmask
; /* target network mask (IP) */
2757 unsigned short rt_flags
;
2760 unsigned char rt_tos
;
2761 unsigned char rt_class
;
2763 short rt_metric
; /* +1 for binary compatibility! */
2764 /* char * */ u32 rt_dev
; /* forcing the device at add */
2765 u32 rt_mtu
; /* per route MTU/Window */
2766 u32 rt_window
; /* Window clamping */
2767 unsigned short rt_irtt
; /* Initial RTT */
2770 struct in6_rtmsg32
{
2771 struct in6_addr rtmsg_dst
;
2772 struct in6_addr rtmsg_src
;
2773 struct in6_addr rtmsg_gateway
;
2783 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
2784 unsigned int cmd
, void __user
*argp
)
2788 struct in6_rtmsg r6
;
2792 mm_segment_t old_fs
= get_fs();
2794 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
2795 struct in6_rtmsg32 __user
*ur6
= argp
;
2796 ret
= copy_from_user (&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
2797 3 * sizeof(struct in6_addr
));
2798 ret
|= __get_user (r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
2799 ret
|= __get_user (r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
2800 ret
|= __get_user (r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
2801 ret
|= __get_user (r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
2802 ret
|= __get_user (r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
2803 ret
|= __get_user (r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
2804 ret
|= __get_user (r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
2808 struct rtentry32 __user
*ur4
= argp
;
2809 ret
= copy_from_user (&r4
.rt_dst
, &(ur4
->rt_dst
),
2810 3 * sizeof(struct sockaddr
));
2811 ret
|= __get_user (r4
.rt_flags
, &(ur4
->rt_flags
));
2812 ret
|= __get_user (r4
.rt_metric
, &(ur4
->rt_metric
));
2813 ret
|= __get_user (r4
.rt_mtu
, &(ur4
->rt_mtu
));
2814 ret
|= __get_user (r4
.rt_window
, &(ur4
->rt_window
));
2815 ret
|= __get_user (r4
.rt_irtt
, &(ur4
->rt_irtt
));
2816 ret
|= __get_user (rtdev
, &(ur4
->rt_dev
));
2818 ret
|= copy_from_user (devname
, compat_ptr(rtdev
), 15);
2819 r4
.rt_dev
= devname
; devname
[15] = 0;
2832 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
2839 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2840 * for some operations; this forces use of the newer bridge-utils that
2841 * use compatiable ioctls
2843 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
2847 if (get_user(tmp
, argp
))
2849 if (tmp
== BRCTL_GET_VERSION
)
2850 return BRCTL_VERSION
+ 1;
2854 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
2855 unsigned int cmd
, unsigned long arg
)
2857 void __user
*argp
= compat_ptr(arg
);
2858 struct sock
*sk
= sock
->sk
;
2859 struct net
*net
= sock_net(sk
);
2861 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
2862 return siocdevprivate_ioctl(net
, cmd
, argp
);
2867 return old_bridge_ioctl(argp
);
2869 return dev_ifname32(net
, argp
);
2871 return dev_ifconf(net
, argp
);
2873 return ethtool_ioctl(net
, argp
);
2875 return compat_siocwandev(net
, argp
);
2878 return compat_sioc_ifmap(net
, cmd
, argp
);
2879 case SIOCBONDENSLAVE
:
2880 case SIOCBONDRELEASE
:
2881 case SIOCBONDSETHWADDR
:
2882 case SIOCBONDSLAVEINFOQUERY
:
2883 case SIOCBONDINFOQUERY
:
2884 case SIOCBONDCHANGEACTIVE
:
2885 return bond_ioctl(net
, cmd
, argp
);
2888 return routing_ioctl(net
, sock
, cmd
, argp
);
2890 return do_siocgstamp(net
, sock
, cmd
, argp
);
2892 return do_siocgstampns(net
, sock
, cmd
, argp
);
2894 return compat_siocshwtstamp(net
, argp
);
2906 return sock_ioctl(file
, cmd
, arg
);
2923 case SIOCSIFHWBROADCAST
:
2925 case SIOCGIFBRDADDR
:
2926 case SIOCSIFBRDADDR
:
2927 case SIOCGIFDSTADDR
:
2928 case SIOCSIFDSTADDR
:
2929 case SIOCGIFNETMASK
:
2930 case SIOCSIFNETMASK
:
2941 return dev_ifsioc(net
, sock
, cmd
, argp
);
2947 return sock_do_ioctl(net
, sock
, cmd
, arg
);
2950 /* Prevent warning from compat_sys_ioctl, these always
2951 * result in -EINVAL in the native case anyway. */
2964 return -ENOIOCTLCMD
;
2967 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2970 struct socket
*sock
= file
->private_data
;
2971 int ret
= -ENOIOCTLCMD
;
2978 if (sock
->ops
->compat_ioctl
)
2979 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2981 if (ret
== -ENOIOCTLCMD
&&
2982 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
2983 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
2985 if (ret
== -ENOIOCTLCMD
)
2986 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
2992 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
2994 return sock
->ops
->bind(sock
, addr
, addrlen
);
2997 int kernel_listen(struct socket
*sock
, int backlog
)
2999 return sock
->ops
->listen(sock
, backlog
);
3002 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3004 struct sock
*sk
= sock
->sk
;
3007 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3012 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3014 sock_release(*newsock
);
3019 (*newsock
)->ops
= sock
->ops
;
3020 __module_get((*newsock
)->ops
->owner
);
3026 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3029 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3032 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3035 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3038 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3041 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3044 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3045 char *optval
, int *optlen
)
3047 mm_segment_t oldfs
= get_fs();
3051 if (level
== SOL_SOCKET
)
3052 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
3054 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
3060 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3061 char *optval
, unsigned int optlen
)
3063 mm_segment_t oldfs
= get_fs();
3067 if (level
== SOL_SOCKET
)
3068 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
3070 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
3076 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3077 size_t size
, int flags
)
3079 sock_update_classid(sock
->sk
);
3081 if (sock
->ops
->sendpage
)
3082 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3084 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3087 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3089 mm_segment_t oldfs
= get_fs();
3093 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3099 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3101 return sock
->ops
->shutdown(sock
, how
);
3104 EXPORT_SYMBOL(sock_create
);
3105 EXPORT_SYMBOL(sock_create_kern
);
3106 EXPORT_SYMBOL(sock_create_lite
);
3107 EXPORT_SYMBOL(sock_map_fd
);
3108 EXPORT_SYMBOL(sock_recvmsg
);
3109 EXPORT_SYMBOL(sock_register
);
3110 EXPORT_SYMBOL(sock_release
);
3111 EXPORT_SYMBOL(sock_sendmsg
);
3112 EXPORT_SYMBOL(sock_unregister
);
3113 EXPORT_SYMBOL(sock_wake_async
);
3114 EXPORT_SYMBOL(sockfd_lookup
);
3115 EXPORT_SYMBOL(kernel_sendmsg
);
3116 EXPORT_SYMBOL(kernel_recvmsg
);
3117 EXPORT_SYMBOL(kernel_bind
);
3118 EXPORT_SYMBOL(kernel_listen
);
3119 EXPORT_SYMBOL(kernel_accept
);
3120 EXPORT_SYMBOL(kernel_connect
);
3121 EXPORT_SYMBOL(kernel_getsockname
);
3122 EXPORT_SYMBOL(kernel_getpeername
);
3123 EXPORT_SYMBOL(kernel_getsockopt
);
3124 EXPORT_SYMBOL(kernel_setsockopt
);
3125 EXPORT_SYMBOL(kernel_sendpage
);
3126 EXPORT_SYMBOL(kernel_sock_ioctl
);
3127 EXPORT_SYMBOL(kernel_sock_shutdown
);