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>
91 #include <asm/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/sockios.h>
104 #include <linux/atalk.h>
106 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
107 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
108 unsigned long nr_segs
, loff_t pos
);
109 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
110 unsigned long nr_segs
, loff_t pos
);
111 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
113 static int sock_close(struct inode
*inode
, struct file
*file
);
114 static unsigned int sock_poll(struct file
*file
,
115 struct poll_table_struct
*wait
);
116 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
118 static long compat_sock_ioctl(struct file
*file
,
119 unsigned int cmd
, unsigned long arg
);
121 static int sock_fasync(int fd
, struct file
*filp
, int on
);
122 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
123 int offset
, size_t size
, loff_t
*ppos
, int more
);
124 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
125 struct pipe_inode_info
*pipe
, size_t len
,
129 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
130 * in the operation structures but are done directly via the socketcall() multiplexor.
133 static const struct file_operations socket_file_ops
= {
134 .owner
= THIS_MODULE
,
136 .aio_read
= sock_aio_read
,
137 .aio_write
= sock_aio_write
,
139 .unlocked_ioctl
= sock_ioctl
,
141 .compat_ioctl
= compat_sock_ioctl
,
144 .open
= sock_no_open
, /* special open code to disallow open via /proc */
145 .release
= sock_close
,
146 .fasync
= sock_fasync
,
147 .sendpage
= sock_sendpage
,
148 .splice_write
= generic_splice_sendpage
,
149 .splice_read
= sock_splice_read
,
153 * The protocol list. Each protocol is registered in here.
156 static DEFINE_SPINLOCK(net_family_lock
);
157 static const struct net_proto_family
*net_families
[NPROTO
] __read_mostly
;
160 * Statistics counters of the socket lists
163 static DEFINE_PER_CPU(int, sockets_in_use
) = 0;
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
171 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
172 16 for IP, 16 for IPX,
175 must be at least one bigger than
176 the AF_UNIX size (see net/unix/af_unix.c
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
191 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr
*kaddr
)
193 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
197 if (copy_from_user(kaddr
, uaddr
, ulen
))
199 return audit_sockaddr(ulen
, kaddr
);
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
219 int move_addr_to_user(struct sockaddr
*kaddr
, int klen
, void __user
*uaddr
,
225 err
= get_user(len
, ulen
);
230 if (len
< 0 || len
> sizeof(struct sockaddr_storage
))
233 if (audit_sockaddr(klen
, kaddr
))
235 if (copy_to_user(uaddr
, kaddr
, len
))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen
, ulen
);
245 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
247 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
249 struct socket_alloc
*ei
;
251 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
254 init_waitqueue_head(&ei
->socket
.wait
);
256 ei
->socket
.fasync_list
= NULL
;
257 ei
->socket
.state
= SS_UNCONNECTED
;
258 ei
->socket
.flags
= 0;
259 ei
->socket
.ops
= NULL
;
260 ei
->socket
.sk
= NULL
;
261 ei
->socket
.file
= NULL
;
263 return &ei
->vfs_inode
;
266 static void sock_destroy_inode(struct inode
*inode
)
268 kmem_cache_free(sock_inode_cachep
,
269 container_of(inode
, struct socket_alloc
, vfs_inode
));
272 static void init_once(void *foo
)
274 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
276 inode_init_once(&ei
->vfs_inode
);
279 static int init_inodecache(void)
281 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
282 sizeof(struct socket_alloc
),
284 (SLAB_HWCACHE_ALIGN
|
285 SLAB_RECLAIM_ACCOUNT
|
288 if (sock_inode_cachep
== NULL
)
293 static const struct super_operations sockfs_ops
= {
294 .alloc_inode
= sock_alloc_inode
,
295 .destroy_inode
=sock_destroy_inode
,
296 .statfs
= simple_statfs
,
299 static int sockfs_get_sb(struct file_system_type
*fs_type
,
300 int flags
, const char *dev_name
, void *data
,
301 struct vfsmount
*mnt
)
303 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
307 static struct vfsmount
*sock_mnt __read_mostly
;
309 static struct file_system_type sock_fs_type
= {
311 .get_sb
= sockfs_get_sb
,
312 .kill_sb
= kill_anon_super
,
315 static int sockfs_delete_dentry(struct dentry
*dentry
)
318 * At creation time, we pretended this dentry was hashed
319 * (by clearing DCACHE_UNHASHED bit in d_flags)
320 * At delete time, we restore the truth : not hashed.
321 * (so that dput() can proceed correctly)
323 dentry
->d_flags
|= DCACHE_UNHASHED
;
328 * sockfs_dname() is called from d_path().
330 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
332 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
333 dentry
->d_inode
->i_ino
);
336 static const struct dentry_operations sockfs_dentry_operations
= {
337 .d_delete
= sockfs_delete_dentry
,
338 .d_dname
= sockfs_dname
,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 static int sock_alloc_file(struct socket
*sock
, struct file
**f
, int flags
)
360 struct qstr name
= { .name
= "" };
365 fd
= get_unused_fd_flags(flags
);
366 if (unlikely(fd
< 0))
369 path
.dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &name
);
370 if (unlikely(!path
.dentry
)) {
374 path
.mnt
= mntget(sock_mnt
);
376 path
.dentry
->d_op
= &sockfs_dentry_operations
;
378 * We dont want to push this dentry into global dentry hash table.
379 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
380 * This permits a working /proc/$pid/fd/XXX on sockets
382 path
.dentry
->d_flags
&= ~DCACHE_UNHASHED
;
383 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
384 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
386 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
388 if (unlikely(!file
)) {
389 /* drop dentry, keep inode */
390 atomic_inc(&path
.dentry
->d_inode
->i_count
);
397 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
399 file
->private_data
= sock
;
405 int sock_map_fd(struct socket
*sock
, int flags
)
407 struct file
*newfile
;
408 int fd
= sock_alloc_file(sock
, &newfile
, flags
);
411 fd_install(fd
, newfile
);
416 static struct socket
*sock_from_file(struct file
*file
, int *err
)
418 if (file
->f_op
== &socket_file_ops
)
419 return file
->private_data
; /* set in sock_map_fd */
426 * sockfd_lookup - Go from a file number to its socket slot
428 * @err: pointer to an error code return
430 * The file handle passed in is locked and the socket it is bound
431 * too is returned. If an error occurs the err pointer is overwritten
432 * with a negative errno code and NULL is returned. The function checks
433 * for both invalid handles and passing a handle which is not a socket.
435 * On a success the socket object pointer is returned.
438 struct socket
*sockfd_lookup(int fd
, int *err
)
449 sock
= sock_from_file(file
, err
);
455 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
461 file
= fget_light(fd
, fput_needed
);
463 sock
= sock_from_file(file
, err
);
466 fput_light(file
, *fput_needed
);
472 * sock_alloc - allocate a socket
474 * Allocate a new inode and socket object. The two are bound together
475 * and initialised. The socket is then returned. If we are out of inodes
479 static struct socket
*sock_alloc(void)
484 inode
= new_inode(sock_mnt
->mnt_sb
);
488 sock
= SOCKET_I(inode
);
490 kmemcheck_annotate_bitfield(sock
, type
);
491 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
492 inode
->i_uid
= current_fsuid();
493 inode
->i_gid
= current_fsgid();
495 percpu_add(sockets_in_use
, 1);
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
505 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
510 const struct file_operations bad_sock_fops
= {
511 .owner
= THIS_MODULE
,
512 .open
= sock_no_open
,
516 * sock_release - close a socket
517 * @sock: socket to close
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
521 * an inode not a file.
524 void sock_release(struct socket
*sock
)
527 struct module
*owner
= sock
->ops
->owner
;
529 sock
->ops
->release(sock
);
534 if (sock
->fasync_list
)
535 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
537 percpu_sub(sockets_in_use
, 1);
539 iput(SOCK_INODE(sock
));
545 int sock_tx_timestamp(struct msghdr
*msg
, struct sock
*sk
,
546 union skb_shared_tx
*shtx
)
549 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
551 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
555 EXPORT_SYMBOL(sock_tx_timestamp
);
557 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
558 struct msghdr
*msg
, size_t size
)
560 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
568 err
= security_socket_sendmsg(sock
, msg
, size
);
572 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
575 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
578 struct sock_iocb siocb
;
581 init_sync_kiocb(&iocb
, NULL
);
582 iocb
.private = &siocb
;
583 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
584 if (-EIOCBQUEUED
== ret
)
585 ret
= wait_on_sync_kiocb(&iocb
);
589 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
590 struct kvec
*vec
, size_t num
, size_t size
)
592 mm_segment_t oldfs
= get_fs();
597 * the following is safe, since for compiler definitions of kvec and
598 * iovec are identical, yielding the same in-core layout and alignment
600 msg
->msg_iov
= (struct iovec
*)vec
;
601 msg
->msg_iovlen
= num
;
602 result
= sock_sendmsg(sock
, msg
, size
);
607 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
610 *ts
= ktime_to_timespec(kt
);
618 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
620 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
623 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
624 struct timespec ts
[3];
626 struct skb_shared_hwtstamps
*shhwtstamps
=
629 /* Race occurred between timestamp enabling and packet
630 receiving. Fill in the current time for now. */
631 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
632 __net_timestamp(skb
);
634 if (need_software_tstamp
) {
635 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
637 skb_get_timestamp(skb
, &tv
);
638 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
642 skb_get_timestampns(skb
, &ts
);
643 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
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
);
696 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
699 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
700 struct msghdr
*msg
, size_t size
, int flags
)
702 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
704 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
707 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
708 size_t size
, int flags
)
711 struct sock_iocb siocb
;
714 init_sync_kiocb(&iocb
, NULL
);
715 iocb
.private = &siocb
;
716 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
717 if (-EIOCBQUEUED
== ret
)
718 ret
= wait_on_sync_kiocb(&iocb
);
722 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
723 size_t size
, int flags
)
726 struct sock_iocb siocb
;
729 init_sync_kiocb(&iocb
, NULL
);
730 iocb
.private = &siocb
;
731 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
732 if (-EIOCBQUEUED
== ret
)
733 ret
= wait_on_sync_kiocb(&iocb
);
737 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
738 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
740 mm_segment_t oldfs
= get_fs();
745 * the following is safe, since for compiler definitions of kvec and
746 * iovec are identical, yielding the same in-core layout and alignment
748 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
749 result
= sock_recvmsg(sock
, msg
, size
, flags
);
754 static void sock_aio_dtor(struct kiocb
*iocb
)
756 kfree(iocb
->private);
759 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
760 int offset
, size_t size
, loff_t
*ppos
, int more
)
765 sock
= file
->private_data
;
767 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
771 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
774 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
775 struct pipe_inode_info
*pipe
, size_t len
,
778 struct socket
*sock
= file
->private_data
;
780 if (unlikely(!sock
->ops
->splice_read
))
783 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
786 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
787 struct sock_iocb
*siocb
)
789 if (!is_sync_kiocb(iocb
)) {
790 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
793 iocb
->ki_dtor
= sock_aio_dtor
;
797 iocb
->private = siocb
;
801 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
802 struct file
*file
, const struct iovec
*iov
,
803 unsigned long nr_segs
)
805 struct socket
*sock
= file
->private_data
;
809 for (i
= 0; i
< nr_segs
; i
++)
810 size
+= iov
[i
].iov_len
;
812 msg
->msg_name
= NULL
;
813 msg
->msg_namelen
= 0;
814 msg
->msg_control
= NULL
;
815 msg
->msg_controllen
= 0;
816 msg
->msg_iov
= (struct iovec
*)iov
;
817 msg
->msg_iovlen
= nr_segs
;
818 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
820 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
823 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
824 unsigned long nr_segs
, loff_t pos
)
826 struct sock_iocb siocb
, *x
;
831 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
835 x
= alloc_sock_iocb(iocb
, &siocb
);
838 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
841 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
842 struct file
*file
, const struct iovec
*iov
,
843 unsigned long nr_segs
)
845 struct socket
*sock
= file
->private_data
;
849 for (i
= 0; i
< nr_segs
; i
++)
850 size
+= iov
[i
].iov_len
;
852 msg
->msg_name
= NULL
;
853 msg
->msg_namelen
= 0;
854 msg
->msg_control
= NULL
;
855 msg
->msg_controllen
= 0;
856 msg
->msg_iov
= (struct iovec
*)iov
;
857 msg
->msg_iovlen
= nr_segs
;
858 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
859 if (sock
->type
== SOCK_SEQPACKET
)
860 msg
->msg_flags
|= MSG_EOR
;
862 return __sock_sendmsg(iocb
, sock
, msg
, size
);
865 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
866 unsigned long nr_segs
, loff_t pos
)
868 struct sock_iocb siocb
, *x
;
873 x
= alloc_sock_iocb(iocb
, &siocb
);
877 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
881 * Atomic setting of ioctl hooks to avoid race
882 * with module unload.
885 static DEFINE_MUTEX(br_ioctl_mutex
);
886 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
) = NULL
;
888 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
890 mutex_lock(&br_ioctl_mutex
);
891 br_ioctl_hook
= hook
;
892 mutex_unlock(&br_ioctl_mutex
);
895 EXPORT_SYMBOL(brioctl_set
);
897 static DEFINE_MUTEX(vlan_ioctl_mutex
);
898 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
900 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
902 mutex_lock(&vlan_ioctl_mutex
);
903 vlan_ioctl_hook
= hook
;
904 mutex_unlock(&vlan_ioctl_mutex
);
907 EXPORT_SYMBOL(vlan_ioctl_set
);
909 static DEFINE_MUTEX(dlci_ioctl_mutex
);
910 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
912 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
914 mutex_lock(&dlci_ioctl_mutex
);
915 dlci_ioctl_hook
= hook
;
916 mutex_unlock(&dlci_ioctl_mutex
);
919 EXPORT_SYMBOL(dlci_ioctl_set
);
921 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
922 unsigned int cmd
, unsigned long arg
)
925 void __user
*argp
= (void __user
*)arg
;
927 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
930 * If this ioctl is unknown try to hand it down
933 if (err
== -ENOIOCTLCMD
)
934 err
= dev_ioctl(net
, cmd
, argp
);
940 * With an ioctl, arg may well be a user mode pointer, but we don't know
941 * what to do with it - that's up to the protocol still.
944 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
948 void __user
*argp
= (void __user
*)arg
;
952 sock
= file
->private_data
;
955 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
956 err
= dev_ioctl(net
, cmd
, argp
);
958 #ifdef CONFIG_WEXT_CORE
959 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
960 err
= dev_ioctl(net
, cmd
, argp
);
967 if (get_user(pid
, (int __user
*)argp
))
969 err
= f_setown(sock
->file
, pid
, 1);
973 err
= put_user(f_getown(sock
->file
),
982 request_module("bridge");
984 mutex_lock(&br_ioctl_mutex
);
986 err
= br_ioctl_hook(net
, cmd
, argp
);
987 mutex_unlock(&br_ioctl_mutex
);
992 if (!vlan_ioctl_hook
)
993 request_module("8021q");
995 mutex_lock(&vlan_ioctl_mutex
);
997 err
= vlan_ioctl_hook(net
, argp
);
998 mutex_unlock(&vlan_ioctl_mutex
);
1003 if (!dlci_ioctl_hook
)
1004 request_module("dlci");
1006 mutex_lock(&dlci_ioctl_mutex
);
1007 if (dlci_ioctl_hook
)
1008 err
= dlci_ioctl_hook(cmd
, argp
);
1009 mutex_unlock(&dlci_ioctl_mutex
);
1012 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1018 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1021 struct socket
*sock
= NULL
;
1023 err
= security_socket_create(family
, type
, protocol
, 1);
1027 sock
= sock_alloc();
1034 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1047 /* No kernel lock held - perfect */
1048 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1050 struct socket
*sock
;
1053 * We can't return errors to poll, so it's either yes or no.
1055 sock
= file
->private_data
;
1056 return sock
->ops
->poll(file
, sock
, wait
);
1059 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1061 struct socket
*sock
= file
->private_data
;
1063 return sock
->ops
->mmap(file
, sock
, vma
);
1066 static int sock_close(struct inode
*inode
, struct file
*filp
)
1069 * It was possible the inode is NULL we were
1070 * closing an unfinished socket.
1074 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1077 sock_release(SOCKET_I(inode
));
1082 * Update the socket async list
1084 * Fasync_list locking strategy.
1086 * 1. fasync_list is modified only under process context socket lock
1087 * i.e. under semaphore.
1088 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1089 * or under socket lock.
1090 * 3. fasync_list can be used from softirq context, so that
1091 * modification under socket lock have to be enhanced with
1092 * write_lock_bh(&sk->sk_callback_lock).
1096 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1098 struct fasync_struct
*fa
, *fna
= NULL
, **prev
;
1099 struct socket
*sock
;
1103 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1108 sock
= filp
->private_data
;
1118 spin_lock(&filp
->f_lock
);
1120 filp
->f_flags
|= FASYNC
;
1122 filp
->f_flags
&= ~FASYNC
;
1123 spin_unlock(&filp
->f_lock
);
1125 prev
= &(sock
->fasync_list
);
1127 for (fa
= *prev
; fa
!= NULL
; prev
= &fa
->fa_next
, fa
= *prev
)
1128 if (fa
->fa_file
== filp
)
1133 write_lock_bh(&sk
->sk_callback_lock
);
1135 write_unlock_bh(&sk
->sk_callback_lock
);
1140 fna
->fa_file
= filp
;
1142 fna
->magic
= FASYNC_MAGIC
;
1143 fna
->fa_next
= sock
->fasync_list
;
1144 write_lock_bh(&sk
->sk_callback_lock
);
1145 sock
->fasync_list
= fna
;
1146 sock_set_flag(sk
, SOCK_FASYNC
);
1147 write_unlock_bh(&sk
->sk_callback_lock
);
1150 write_lock_bh(&sk
->sk_callback_lock
);
1151 *prev
= fa
->fa_next
;
1152 if (!sock
->fasync_list
)
1153 sock_reset_flag(sk
, SOCK_FASYNC
);
1154 write_unlock_bh(&sk
->sk_callback_lock
);
1160 release_sock(sock
->sk
);
1164 /* This function may be called only under socket lock or callback_lock */
1166 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1168 if (!sock
|| !sock
->fasync_list
)
1171 case SOCK_WAKE_WAITD
:
1172 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1175 case SOCK_WAKE_SPACE
:
1176 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1181 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1184 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1189 static int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1190 struct socket
**res
, int kern
)
1193 struct socket
*sock
;
1194 const struct net_proto_family
*pf
;
1197 * Check protocol is in range
1199 if (family
< 0 || family
>= NPROTO
)
1200 return -EAFNOSUPPORT
;
1201 if (type
< 0 || type
>= SOCK_MAX
)
1206 This uglymoron is moved from INET layer to here to avoid
1207 deadlock in module load.
1209 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1213 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1219 err
= security_socket_create(family
, type
, protocol
, kern
);
1224 * Allocate the socket and allow the family to set things up. if
1225 * the protocol is 0, the family is instructed to select an appropriate
1228 sock
= sock_alloc();
1230 if (net_ratelimit())
1231 printk(KERN_WARNING
"socket: no more sockets\n");
1232 return -ENFILE
; /* Not exactly a match, but its the
1233 closest posix thing */
1238 #ifdef CONFIG_MODULES
1239 /* Attempt to load a protocol module if the find failed.
1241 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1242 * requested real, full-featured networking support upon configuration.
1243 * Otherwise module support will break!
1245 if (net_families
[family
] == NULL
)
1246 request_module("net-pf-%d", family
);
1250 pf
= rcu_dereference(net_families
[family
]);
1251 err
= -EAFNOSUPPORT
;
1256 * We will call the ->create function, that possibly is in a loadable
1257 * module, so we have to bump that loadable module refcnt first.
1259 if (!try_module_get(pf
->owner
))
1262 /* Now protected by module ref count */
1265 err
= pf
->create(net
, sock
, protocol
, kern
);
1267 goto out_module_put
;
1270 * Now to bump the refcnt of the [loadable] module that owns this
1271 * socket at sock_release time we decrement its refcnt.
1273 if (!try_module_get(sock
->ops
->owner
))
1274 goto out_module_busy
;
1277 * Now that we're done with the ->create function, the [loadable]
1278 * module can have its refcnt decremented
1280 module_put(pf
->owner
);
1281 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1283 goto out_sock_release
;
1289 err
= -EAFNOSUPPORT
;
1292 module_put(pf
->owner
);
1299 goto out_sock_release
;
1302 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1304 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1307 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1309 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1312 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1315 struct socket
*sock
;
1318 /* Check the SOCK_* constants for consistency. */
1319 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1320 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1321 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1322 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1324 flags
= type
& ~SOCK_TYPE_MASK
;
1325 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1327 type
&= SOCK_TYPE_MASK
;
1329 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1330 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1332 retval
= sock_create(family
, type
, protocol
, &sock
);
1336 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1341 /* It may be already another descriptor 8) Not kernel problem. */
1350 * Create a pair of connected sockets.
1353 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1354 int __user
*, usockvec
)
1356 struct socket
*sock1
, *sock2
;
1358 struct file
*newfile1
, *newfile2
;
1361 flags
= type
& ~SOCK_TYPE_MASK
;
1362 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1364 type
&= SOCK_TYPE_MASK
;
1366 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1367 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1370 * Obtain the first socket and check if the underlying protocol
1371 * supports the socketpair call.
1374 err
= sock_create(family
, type
, protocol
, &sock1
);
1378 err
= sock_create(family
, type
, protocol
, &sock2
);
1382 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1384 goto out_release_both
;
1386 fd1
= sock_alloc_file(sock1
, &newfile1
, flags
);
1387 if (unlikely(fd1
< 0)) {
1389 goto out_release_both
;
1392 fd2
= sock_alloc_file(sock2
, &newfile2
, flags
);
1393 if (unlikely(fd2
< 0)) {
1397 sock_release(sock2
);
1401 audit_fd_pair(fd1
, fd2
);
1402 fd_install(fd1
, newfile1
);
1403 fd_install(fd2
, newfile2
);
1404 /* fd1 and fd2 may be already another descriptors.
1405 * Not kernel problem.
1408 err
= put_user(fd1
, &usockvec
[0]);
1410 err
= put_user(fd2
, &usockvec
[1]);
1419 sock_release(sock2
);
1421 sock_release(sock1
);
1427 * Bind a name to a socket. Nothing much to do here since it's
1428 * the protocol's responsibility to handle the local address.
1430 * We move the socket address to kernel space before we call
1431 * the protocol layer (having also checked the address is ok).
1434 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1436 struct socket
*sock
;
1437 struct sockaddr_storage address
;
1438 int err
, fput_needed
;
1440 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1442 err
= move_addr_to_kernel(umyaddr
, addrlen
, (struct sockaddr
*)&address
);
1444 err
= security_socket_bind(sock
,
1445 (struct sockaddr
*)&address
,
1448 err
= sock
->ops
->bind(sock
,
1452 fput_light(sock
->file
, fput_needed
);
1458 * Perform a listen. Basically, we allow the protocol to do anything
1459 * necessary for a listen, and if that works, we mark the socket as
1460 * ready for listening.
1463 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1465 struct socket
*sock
;
1466 int err
, fput_needed
;
1469 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1471 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1472 if ((unsigned)backlog
> somaxconn
)
1473 backlog
= somaxconn
;
1475 err
= security_socket_listen(sock
, backlog
);
1477 err
= sock
->ops
->listen(sock
, backlog
);
1479 fput_light(sock
->file
, fput_needed
);
1485 * For accept, we attempt to create a new socket, set up the link
1486 * with the client, wake up the client, then return the new
1487 * connected fd. We collect the address of the connector in kernel
1488 * space and move it to user at the very end. This is unclean because
1489 * we open the socket then return an error.
1491 * 1003.1g adds the ability to recvmsg() to query connection pending
1492 * status to recvmsg. We need to add that support in a way thats
1493 * clean when we restucture accept also.
1496 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1497 int __user
*, upeer_addrlen
, int, flags
)
1499 struct socket
*sock
, *newsock
;
1500 struct file
*newfile
;
1501 int err
, len
, newfd
, fput_needed
;
1502 struct sockaddr_storage address
;
1504 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1507 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1508 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1510 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1515 if (!(newsock
= sock_alloc()))
1518 newsock
->type
= sock
->type
;
1519 newsock
->ops
= sock
->ops
;
1522 * We don't need try_module_get here, as the listening socket (sock)
1523 * has the protocol module (sock->ops->owner) held.
1525 __module_get(newsock
->ops
->owner
);
1527 newfd
= sock_alloc_file(newsock
, &newfile
, flags
);
1528 if (unlikely(newfd
< 0)) {
1530 sock_release(newsock
);
1534 err
= security_socket_accept(sock
, newsock
);
1538 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1542 if (upeer_sockaddr
) {
1543 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1545 err
= -ECONNABORTED
;
1548 err
= move_addr_to_user((struct sockaddr
*)&address
,
1549 len
, upeer_sockaddr
, upeer_addrlen
);
1554 /* File flags are not inherited via accept() unlike another OSes. */
1556 fd_install(newfd
, newfile
);
1560 fput_light(sock
->file
, fput_needed
);
1565 put_unused_fd(newfd
);
1569 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1570 int __user
*, upeer_addrlen
)
1572 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1576 * Attempt to connect to a socket with the server address. The address
1577 * is in user space so we verify it is OK and move it to kernel space.
1579 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1582 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1583 * other SEQPACKET protocols that take time to connect() as it doesn't
1584 * include the -EINPROGRESS status for such sockets.
1587 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1590 struct socket
*sock
;
1591 struct sockaddr_storage address
;
1592 int err
, fput_needed
;
1594 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1597 err
= move_addr_to_kernel(uservaddr
, addrlen
, (struct sockaddr
*)&address
);
1602 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1606 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1607 sock
->file
->f_flags
);
1609 fput_light(sock
->file
, fput_needed
);
1615 * Get the local address ('name') of a socket object. Move the obtained
1616 * name to user space.
1619 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1620 int __user
*, usockaddr_len
)
1622 struct socket
*sock
;
1623 struct sockaddr_storage address
;
1624 int len
, err
, fput_needed
;
1626 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1630 err
= security_socket_getsockname(sock
);
1634 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1637 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
, usockaddr_len
);
1640 fput_light(sock
->file
, fput_needed
);
1646 * Get the remote address ('name') of a socket object. Move the obtained
1647 * name to user space.
1650 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1651 int __user
*, usockaddr_len
)
1653 struct socket
*sock
;
1654 struct sockaddr_storage address
;
1655 int len
, err
, fput_needed
;
1657 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1659 err
= security_socket_getpeername(sock
);
1661 fput_light(sock
->file
, fput_needed
);
1666 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1669 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
,
1671 fput_light(sock
->file
, fput_needed
);
1677 * Send a datagram to a given address. We move the address into kernel
1678 * space and check the user space data area is readable before invoking
1682 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1683 unsigned, flags
, struct sockaddr __user
*, addr
,
1686 struct socket
*sock
;
1687 struct sockaddr_storage address
;
1693 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1697 iov
.iov_base
= buff
;
1699 msg
.msg_name
= NULL
;
1702 msg
.msg_control
= NULL
;
1703 msg
.msg_controllen
= 0;
1704 msg
.msg_namelen
= 0;
1706 err
= move_addr_to_kernel(addr
, addr_len
, (struct sockaddr
*)&address
);
1709 msg
.msg_name
= (struct sockaddr
*)&address
;
1710 msg
.msg_namelen
= addr_len
;
1712 if (sock
->file
->f_flags
& O_NONBLOCK
)
1713 flags
|= MSG_DONTWAIT
;
1714 msg
.msg_flags
= flags
;
1715 err
= sock_sendmsg(sock
, &msg
, len
);
1718 fput_light(sock
->file
, fput_needed
);
1724 * Send a datagram down a socket.
1727 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1730 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1734 * Receive a frame from the socket and optionally record the address of the
1735 * sender. We verify the buffers are writable and if needed move the
1736 * sender address from kernel to user space.
1739 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1740 unsigned, flags
, struct sockaddr __user
*, addr
,
1741 int __user
*, addr_len
)
1743 struct socket
*sock
;
1746 struct sockaddr_storage address
;
1750 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1754 msg
.msg_control
= NULL
;
1755 msg
.msg_controllen
= 0;
1759 iov
.iov_base
= ubuf
;
1760 msg
.msg_name
= (struct sockaddr
*)&address
;
1761 msg
.msg_namelen
= sizeof(address
);
1762 if (sock
->file
->f_flags
& O_NONBLOCK
)
1763 flags
|= MSG_DONTWAIT
;
1764 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1766 if (err
>= 0 && addr
!= NULL
) {
1767 err2
= move_addr_to_user((struct sockaddr
*)&address
,
1768 msg
.msg_namelen
, addr
, addr_len
);
1773 fput_light(sock
->file
, fput_needed
);
1779 * Receive a datagram from a socket.
1782 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1785 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1789 * Set a socket option. Because we don't know the option lengths we have
1790 * to pass the user mode parameter for the protocols to sort out.
1793 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1794 char __user
*, optval
, int, optlen
)
1796 int err
, fput_needed
;
1797 struct socket
*sock
;
1802 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1804 err
= security_socket_setsockopt(sock
, level
, optname
);
1808 if (level
== SOL_SOCKET
)
1810 sock_setsockopt(sock
, level
, optname
, optval
,
1814 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1817 fput_light(sock
->file
, fput_needed
);
1823 * Get a socket option. Because we don't know the option lengths we have
1824 * to pass a user mode parameter for the protocols to sort out.
1827 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1828 char __user
*, optval
, int __user
*, optlen
)
1830 int err
, fput_needed
;
1831 struct socket
*sock
;
1833 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1835 err
= security_socket_getsockopt(sock
, level
, optname
);
1839 if (level
== SOL_SOCKET
)
1841 sock_getsockopt(sock
, level
, optname
, optval
,
1845 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1848 fput_light(sock
->file
, fput_needed
);
1854 * Shutdown a socket.
1857 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1859 int err
, fput_needed
;
1860 struct socket
*sock
;
1862 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1864 err
= security_socket_shutdown(sock
, how
);
1866 err
= sock
->ops
->shutdown(sock
, how
);
1867 fput_light(sock
->file
, fput_needed
);
1872 /* A couple of helpful macros for getting the address of the 32/64 bit
1873 * fields which are the same type (int / unsigned) on our platforms.
1875 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1876 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1877 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1880 * BSD sendmsg interface
1883 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned, flags
)
1885 struct compat_msghdr __user
*msg_compat
=
1886 (struct compat_msghdr __user
*)msg
;
1887 struct socket
*sock
;
1888 struct sockaddr_storage address
;
1889 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1890 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1891 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1892 /* 20 is size of ipv6_pktinfo */
1893 unsigned char *ctl_buf
= ctl
;
1894 struct msghdr msg_sys
;
1895 int err
, ctl_len
, iov_size
, total_len
;
1899 if (MSG_CMSG_COMPAT
& flags
) {
1900 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1903 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1906 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1910 /* do not move before msg_sys is valid */
1912 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1915 /* Check whether to allocate the iovec area */
1917 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1918 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1919 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1924 /* This will also move the address data into kernel space */
1925 if (MSG_CMSG_COMPAT
& flags
) {
1926 err
= verify_compat_iovec(&msg_sys
, iov
,
1927 (struct sockaddr
*)&address
,
1930 err
= verify_iovec(&msg_sys
, iov
,
1931 (struct sockaddr
*)&address
,
1939 if (msg_sys
.msg_controllen
> INT_MAX
)
1941 ctl_len
= msg_sys
.msg_controllen
;
1942 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1944 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1948 ctl_buf
= msg_sys
.msg_control
;
1949 ctl_len
= msg_sys
.msg_controllen
;
1950 } else if (ctl_len
) {
1951 if (ctl_len
> sizeof(ctl
)) {
1952 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1953 if (ctl_buf
== NULL
)
1958 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1959 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1960 * checking falls down on this.
1962 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1965 msg_sys
.msg_control
= ctl_buf
;
1967 msg_sys
.msg_flags
= flags
;
1969 if (sock
->file
->f_flags
& O_NONBLOCK
)
1970 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1971 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1975 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1977 if (iov
!= iovstack
)
1978 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1980 fput_light(sock
->file
, fput_needed
);
1985 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1986 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
1988 struct compat_msghdr __user
*msg_compat
=
1989 (struct compat_msghdr __user
*)msg
;
1990 struct iovec iovstack
[UIO_FASTIOV
];
1991 struct iovec
*iov
= iovstack
;
1992 unsigned long cmsg_ptr
;
1993 int err
, iov_size
, total_len
, len
;
1995 /* kernel mode address */
1996 struct sockaddr_storage addr
;
1998 /* user mode address pointers */
1999 struct sockaddr __user
*uaddr
;
2000 int __user
*uaddr_len
;
2002 if (MSG_CMSG_COMPAT
& flags
) {
2003 if (get_compat_msghdr(msg_sys
, msg_compat
))
2006 else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2010 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2013 /* Check whether to allocate the iovec area */
2015 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
2016 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2017 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
2023 * Save the user-mode address (verify_iovec will change the
2024 * kernel msghdr to use the kernel address space)
2027 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2028 uaddr_len
= COMPAT_NAMELEN(msg
);
2029 if (MSG_CMSG_COMPAT
& flags
) {
2030 err
= verify_compat_iovec(msg_sys
, iov
,
2031 (struct sockaddr
*)&addr
,
2034 err
= verify_iovec(msg_sys
, iov
,
2035 (struct sockaddr
*)&addr
,
2041 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2042 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2044 if (sock
->file
->f_flags
& O_NONBLOCK
)
2045 flags
|= MSG_DONTWAIT
;
2046 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2052 if (uaddr
!= NULL
) {
2053 err
= move_addr_to_user((struct sockaddr
*)&addr
,
2054 msg_sys
->msg_namelen
, uaddr
,
2059 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2063 if (MSG_CMSG_COMPAT
& flags
)
2064 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2065 &msg_compat
->msg_controllen
);
2067 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2068 &msg
->msg_controllen
);
2074 if (iov
!= iovstack
)
2075 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2081 * BSD recvmsg interface
2084 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2085 unsigned int, flags
)
2087 int fput_needed
, err
;
2088 struct msghdr msg_sys
;
2089 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2094 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2096 fput_light(sock
->file
, fput_needed
);
2102 * Linux recvmmsg interface
2105 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2106 unsigned int flags
, struct timespec
*timeout
)
2108 int fput_needed
, err
, datagrams
;
2109 struct socket
*sock
;
2110 struct mmsghdr __user
*entry
;
2111 struct compat_mmsghdr __user
*compat_entry
;
2112 struct msghdr msg_sys
;
2113 struct timespec end_time
;
2116 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2122 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2126 err
= sock_error(sock
->sk
);
2131 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2133 while (datagrams
< vlen
) {
2135 * No need to ask LSM for more than the first datagram.
2137 if (MSG_CMSG_COMPAT
& flags
) {
2138 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2139 &msg_sys
, flags
, datagrams
);
2142 err
= __put_user(err
, &compat_entry
->msg_len
);
2145 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2146 &msg_sys
, flags
, datagrams
);
2149 err
= put_user(err
, &entry
->msg_len
);
2158 ktime_get_ts(timeout
);
2159 *timeout
= timespec_sub(end_time
, *timeout
);
2160 if (timeout
->tv_sec
< 0) {
2161 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2165 /* Timeout, return less than vlen datagrams */
2166 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2170 /* Out of band data, return right away */
2171 if (msg_sys
.msg_flags
& MSG_OOB
)
2176 fput_light(sock
->file
, fput_needed
);
2181 if (datagrams
!= 0) {
2183 * We may return less entries than requested (vlen) if the
2184 * sock is non block and there aren't enough datagrams...
2186 if (err
!= -EAGAIN
) {
2188 * ... or if recvmsg returns an error after we
2189 * received some datagrams, where we record the
2190 * error to return on the next call or if the
2191 * app asks about it using getsockopt(SO_ERROR).
2193 sock
->sk
->sk_err
= -err
;
2202 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2203 unsigned int, vlen
, unsigned int, flags
,
2204 struct timespec __user
*, timeout
)
2207 struct timespec timeout_sys
;
2210 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2212 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2215 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2217 if (datagrams
> 0 &&
2218 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2219 datagrams
= -EFAULT
;
2224 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2225 /* Argument list sizes for sys_socketcall */
2226 #define AL(x) ((x) * sizeof(unsigned long))
2227 static const unsigned char nargs
[20] = {
2228 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2229 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2230 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2237 * System call vectors.
2239 * Argument checking cleaned up. Saved 20% in size.
2240 * This function doesn't need to set the kernel lock because
2241 * it is set by the callees.
2244 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2247 unsigned long a0
, a1
;
2251 if (call
< 1 || call
> SYS_RECVMMSG
)
2255 if (len
> sizeof(a
))
2258 /* copy_from_user should be SMP safe. */
2259 if (copy_from_user(a
, args
, len
))
2262 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2269 err
= sys_socket(a0
, a1
, a
[2]);
2272 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2275 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2278 err
= sys_listen(a0
, a1
);
2281 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2282 (int __user
*)a
[2], 0);
2284 case SYS_GETSOCKNAME
:
2286 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2287 (int __user
*)a
[2]);
2289 case SYS_GETPEERNAME
:
2291 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2292 (int __user
*)a
[2]);
2294 case SYS_SOCKETPAIR
:
2295 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2298 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2301 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2302 (struct sockaddr __user
*)a
[4], a
[5]);
2305 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2308 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2309 (struct sockaddr __user
*)a
[4],
2310 (int __user
*)a
[5]);
2313 err
= sys_shutdown(a0
, a1
);
2315 case SYS_SETSOCKOPT
:
2316 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2318 case SYS_GETSOCKOPT
:
2320 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2321 (int __user
*)a
[4]);
2324 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2327 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2330 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2331 (struct timespec __user
*)a
[4]);
2334 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2335 (int __user
*)a
[2], a
[3]);
2344 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2347 * sock_register - add a socket protocol handler
2348 * @ops: description of protocol
2350 * This function is called by a protocol handler that wants to
2351 * advertise its address family, and have it linked into the
2352 * socket interface. The value ops->family coresponds to the
2353 * socket system call protocol family.
2355 int sock_register(const struct net_proto_family
*ops
)
2359 if (ops
->family
>= NPROTO
) {
2360 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2365 spin_lock(&net_family_lock
);
2366 if (net_families
[ops
->family
])
2369 net_families
[ops
->family
] = ops
;
2372 spin_unlock(&net_family_lock
);
2374 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2379 * sock_unregister - remove a protocol handler
2380 * @family: protocol family to remove
2382 * This function is called by a protocol handler that wants to
2383 * remove its address family, and have it unlinked from the
2384 * new socket creation.
2386 * If protocol handler is a module, then it can use module reference
2387 * counts to protect against new references. If protocol handler is not
2388 * a module then it needs to provide its own protection in
2389 * the ops->create routine.
2391 void sock_unregister(int family
)
2393 BUG_ON(family
< 0 || family
>= NPROTO
);
2395 spin_lock(&net_family_lock
);
2396 net_families
[family
] = NULL
;
2397 spin_unlock(&net_family_lock
);
2401 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2404 static int __init
sock_init(void)
2407 * Initialize sock SLAB cache.
2413 * Initialize skbuff SLAB cache
2418 * Initialize the protocols module.
2422 register_filesystem(&sock_fs_type
);
2423 sock_mnt
= kern_mount(&sock_fs_type
);
2425 /* The real protocol initialization is performed in later initcalls.
2428 #ifdef CONFIG_NETFILTER
2435 core_initcall(sock_init
); /* early initcall */
2437 #ifdef CONFIG_PROC_FS
2438 void socket_seq_show(struct seq_file
*seq
)
2443 for_each_possible_cpu(cpu
)
2444 counter
+= per_cpu(sockets_in_use
, cpu
);
2446 /* It can be negative, by the way. 8) */
2450 seq_printf(seq
, "sockets: used %d\n", counter
);
2452 #endif /* CONFIG_PROC_FS */
2454 #ifdef CONFIG_COMPAT
2455 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2456 unsigned int cmd
, struct compat_timeval __user
*up
)
2458 mm_segment_t old_fs
= get_fs();
2463 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2466 err
= put_user(ktv
.tv_sec
, &up
->tv_sec
);
2467 err
|= __put_user(ktv
.tv_usec
, &up
->tv_usec
);
2472 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2473 unsigned int cmd
, struct compat_timespec __user
*up
)
2475 mm_segment_t old_fs
= get_fs();
2476 struct timespec kts
;
2480 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2483 err
= put_user(kts
.tv_sec
, &up
->tv_sec
);
2484 err
|= __put_user(kts
.tv_nsec
, &up
->tv_nsec
);
2489 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2491 struct ifreq __user
*uifr
;
2494 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2495 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2498 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2502 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2508 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2510 struct compat_ifconf ifc32
;
2512 struct ifconf __user
*uifc
;
2513 struct compat_ifreq __user
*ifr32
;
2514 struct ifreq __user
*ifr
;
2518 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2521 if (ifc32
.ifcbuf
== 0) {
2525 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2527 size_t len
=((ifc32
.ifc_len
/ sizeof (struct compat_ifreq
)) + 1) *
2528 sizeof (struct ifreq
);
2529 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2531 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2532 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2533 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof (struct compat_ifreq
)) {
2534 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2540 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2543 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2547 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2551 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2553 i
+ sizeof (struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2554 i
+= sizeof (struct compat_ifreq
), j
+= sizeof (struct ifreq
)) {
2555 if (copy_in_user(ifr32
, ifr
, sizeof (struct compat_ifreq
)))
2561 if (ifc32
.ifcbuf
== 0) {
2562 /* Translate from 64-bit structure multiple to
2566 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2571 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2577 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2579 struct ifreq __user
*ifr
;
2583 ifr
= compat_alloc_user_space(sizeof(*ifr
));
2585 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2588 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2591 datap
= compat_ptr(data
);
2592 if (put_user(datap
, &ifr
->ifr_ifru
.ifru_data
))
2595 return dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2598 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2601 compat_uptr_t uptr32
;
2602 struct ifreq __user
*uifr
;
2604 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2605 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2608 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2611 uptr
= compat_ptr(uptr32
);
2613 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2616 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2619 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2620 struct compat_ifreq __user
*ifr32
)
2623 struct ifreq __user
*uifr
;
2624 mm_segment_t old_fs
;
2630 case SIOCBONDENSLAVE
:
2631 case SIOCBONDRELEASE
:
2632 case SIOCBONDSETHWADDR
:
2633 case SIOCBONDCHANGEACTIVE
:
2634 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2639 err
= dev_ioctl(net
, cmd
, &kifr
);
2643 case SIOCBONDSLAVEINFOQUERY
:
2644 case SIOCBONDINFOQUERY
:
2645 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2646 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2649 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2652 datap
= compat_ptr(data
);
2653 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2656 return dev_ioctl(net
, cmd
, uifr
);
2662 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2663 struct compat_ifreq __user
*u_ifreq32
)
2665 struct ifreq __user
*u_ifreq64
;
2666 char tmp_buf
[IFNAMSIZ
];
2667 void __user
*data64
;
2670 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2673 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2675 data64
= compat_ptr(data32
);
2677 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2679 /* Don't check these user accesses, just let that get trapped
2680 * in the ioctl handler instead.
2682 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2685 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2688 return dev_ioctl(net
, cmd
, u_ifreq64
);
2691 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2692 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2694 struct ifreq __user
*uifr
;
2697 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2698 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2701 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2712 case SIOCGIFBRDADDR
:
2713 case SIOCGIFDSTADDR
:
2714 case SIOCGIFNETMASK
:
2719 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2727 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2728 struct compat_ifreq __user
*uifr32
)
2731 struct compat_ifmap __user
*uifmap32
;
2732 mm_segment_t old_fs
;
2735 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2736 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2737 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2738 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2739 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2740 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2741 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2742 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2748 err
= dev_ioctl(net
, cmd
, (void __user
*)&ifr
);
2751 if (cmd
== SIOCGIFMAP
&& !err
) {
2752 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2753 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2754 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2755 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2756 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2757 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2758 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2765 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2768 compat_uptr_t uptr32
;
2769 struct ifreq __user
*uifr
;
2771 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2772 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2775 if (get_user(uptr32
, &uifr32
->ifr_data
))
2778 uptr
= compat_ptr(uptr32
);
2780 if (put_user(uptr
, &uifr
->ifr_data
))
2783 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
2788 struct sockaddr rt_dst
; /* target address */
2789 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2790 struct sockaddr rt_genmask
; /* target network mask (IP) */
2791 unsigned short rt_flags
;
2794 unsigned char rt_tos
;
2795 unsigned char rt_class
;
2797 short rt_metric
; /* +1 for binary compatibility! */
2798 /* char * */ u32 rt_dev
; /* forcing the device at add */
2799 u32 rt_mtu
; /* per route MTU/Window */
2800 u32 rt_window
; /* Window clamping */
2801 unsigned short rt_irtt
; /* Initial RTT */
2804 struct in6_rtmsg32
{
2805 struct in6_addr rtmsg_dst
;
2806 struct in6_addr rtmsg_src
;
2807 struct in6_addr rtmsg_gateway
;
2817 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
2818 unsigned int cmd
, void __user
*argp
)
2822 struct in6_rtmsg r6
;
2826 mm_segment_t old_fs
= get_fs();
2828 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
2829 struct in6_rtmsg32 __user
*ur6
= argp
;
2830 ret
= copy_from_user (&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
2831 3 * sizeof(struct in6_addr
));
2832 ret
|= __get_user (r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
2833 ret
|= __get_user (r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
2834 ret
|= __get_user (r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
2835 ret
|= __get_user (r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
2836 ret
|= __get_user (r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
2837 ret
|= __get_user (r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
2838 ret
|= __get_user (r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
2842 struct rtentry32 __user
*ur4
= argp
;
2843 ret
= copy_from_user (&r4
.rt_dst
, &(ur4
->rt_dst
),
2844 3 * sizeof(struct sockaddr
));
2845 ret
|= __get_user (r4
.rt_flags
, &(ur4
->rt_flags
));
2846 ret
|= __get_user (r4
.rt_metric
, &(ur4
->rt_metric
));
2847 ret
|= __get_user (r4
.rt_mtu
, &(ur4
->rt_mtu
));
2848 ret
|= __get_user (r4
.rt_window
, &(ur4
->rt_window
));
2849 ret
|= __get_user (r4
.rt_irtt
, &(ur4
->rt_irtt
));
2850 ret
|= __get_user (rtdev
, &(ur4
->rt_dev
));
2852 ret
|= copy_from_user (devname
, compat_ptr(rtdev
), 15);
2853 r4
.rt_dev
= devname
; devname
[15] = 0;
2866 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
2873 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2874 * for some operations; this forces use of the newer bridge-utils that
2875 * use compatiable ioctls
2877 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
2881 if (get_user(tmp
, argp
))
2883 if (tmp
== BRCTL_GET_VERSION
)
2884 return BRCTL_VERSION
+ 1;
2888 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
2889 unsigned int cmd
, unsigned long arg
)
2891 void __user
*argp
= compat_ptr(arg
);
2892 struct sock
*sk
= sock
->sk
;
2893 struct net
*net
= sock_net(sk
);
2895 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
2896 return siocdevprivate_ioctl(net
, cmd
, argp
);
2901 return old_bridge_ioctl(argp
);
2903 return dev_ifname32(net
, argp
);
2905 return dev_ifconf(net
, argp
);
2907 return ethtool_ioctl(net
, argp
);
2909 return compat_siocwandev(net
, argp
);
2912 return compat_sioc_ifmap(net
, cmd
, argp
);
2913 case SIOCBONDENSLAVE
:
2914 case SIOCBONDRELEASE
:
2915 case SIOCBONDSETHWADDR
:
2916 case SIOCBONDSLAVEINFOQUERY
:
2917 case SIOCBONDINFOQUERY
:
2918 case SIOCBONDCHANGEACTIVE
:
2919 return bond_ioctl(net
, cmd
, argp
);
2922 return routing_ioctl(net
, sock
, cmd
, argp
);
2924 return do_siocgstamp(net
, sock
, cmd
, argp
);
2926 return do_siocgstampns(net
, sock
, cmd
, argp
);
2928 return compat_siocshwtstamp(net
, argp
);
2940 return sock_ioctl(file
, cmd
, arg
);
2957 case SIOCSIFHWBROADCAST
:
2959 case SIOCGIFBRDADDR
:
2960 case SIOCSIFBRDADDR
:
2961 case SIOCGIFDSTADDR
:
2962 case SIOCSIFDSTADDR
:
2963 case SIOCGIFNETMASK
:
2964 case SIOCSIFNETMASK
:
2975 return dev_ifsioc(net
, sock
, cmd
, argp
);
2981 return sock_do_ioctl(net
, sock
, cmd
, arg
);
2984 /* Prevent warning from compat_sys_ioctl, these always
2985 * result in -EINVAL in the native case anyway. */
2998 return -ENOIOCTLCMD
;
3001 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
3004 struct socket
*sock
= file
->private_data
;
3005 int ret
= -ENOIOCTLCMD
;
3012 if (sock
->ops
->compat_ioctl
)
3013 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3015 if (ret
== -ENOIOCTLCMD
&&
3016 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3017 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3019 if (ret
== -ENOIOCTLCMD
)
3020 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3026 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3028 return sock
->ops
->bind(sock
, addr
, addrlen
);
3031 int kernel_listen(struct socket
*sock
, int backlog
)
3033 return sock
->ops
->listen(sock
, backlog
);
3036 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3038 struct sock
*sk
= sock
->sk
;
3041 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3046 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3048 sock_release(*newsock
);
3053 (*newsock
)->ops
= sock
->ops
;
3054 __module_get((*newsock
)->ops
->owner
);
3060 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3063 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3066 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3069 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3072 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3075 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3078 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3079 char *optval
, int *optlen
)
3081 mm_segment_t oldfs
= get_fs();
3085 if (level
== SOL_SOCKET
)
3086 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
3088 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
3094 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3095 char *optval
, unsigned int optlen
)
3097 mm_segment_t oldfs
= get_fs();
3101 if (level
== SOL_SOCKET
)
3102 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
3104 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
3110 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3111 size_t size
, int flags
)
3113 if (sock
->ops
->sendpage
)
3114 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3116 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3119 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3121 mm_segment_t oldfs
= get_fs();
3125 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3131 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3133 return sock
->ops
->shutdown(sock
, how
);
3136 EXPORT_SYMBOL(sock_create
);
3137 EXPORT_SYMBOL(sock_create_kern
);
3138 EXPORT_SYMBOL(sock_create_lite
);
3139 EXPORT_SYMBOL(sock_map_fd
);
3140 EXPORT_SYMBOL(sock_recvmsg
);
3141 EXPORT_SYMBOL(sock_register
);
3142 EXPORT_SYMBOL(sock_release
);
3143 EXPORT_SYMBOL(sock_sendmsg
);
3144 EXPORT_SYMBOL(sock_unregister
);
3145 EXPORT_SYMBOL(sock_wake_async
);
3146 EXPORT_SYMBOL(sockfd_lookup
);
3147 EXPORT_SYMBOL(kernel_sendmsg
);
3148 EXPORT_SYMBOL(kernel_recvmsg
);
3149 EXPORT_SYMBOL(kernel_bind
);
3150 EXPORT_SYMBOL(kernel_listen
);
3151 EXPORT_SYMBOL(kernel_accept
);
3152 EXPORT_SYMBOL(kernel_connect
);
3153 EXPORT_SYMBOL(kernel_getsockname
);
3154 EXPORT_SYMBOL(kernel_getpeername
);
3155 EXPORT_SYMBOL(kernel_getsockopt
);
3156 EXPORT_SYMBOL(kernel_setsockopt
);
3157 EXPORT_SYMBOL(kernel_sendpage
);
3158 EXPORT_SYMBOL(kernel_sock_ioctl
);
3159 EXPORT_SYMBOL(kernel_sock_shutdown
);