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_fd(struct file
**filep
, int flags
)
362 fd
= get_unused_fd_flags(flags
);
363 if (likely(fd
>= 0)) {
364 struct file
*file
= get_empty_filp();
367 if (unlikely(!file
)) {
376 static int sock_attach_fd(struct socket
*sock
, struct file
*file
, int flags
)
378 struct dentry
*dentry
;
379 struct qstr name
= { .name
= "" };
381 dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &name
);
382 if (unlikely(!dentry
))
385 dentry
->d_op
= &sockfs_dentry_operations
;
387 * We dont want to push this dentry into global dentry hash table.
388 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
389 * This permits a working /proc/$pid/fd/XXX on sockets
391 dentry
->d_flags
&= ~DCACHE_UNHASHED
;
392 d_instantiate(dentry
, SOCK_INODE(sock
));
395 init_file(file
, sock_mnt
, dentry
, FMODE_READ
| FMODE_WRITE
,
397 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
398 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
400 file
->private_data
= sock
;
405 int sock_map_fd(struct socket
*sock
, int flags
)
407 struct file
*newfile
;
408 int fd
= sock_alloc_fd(&newfile
, flags
);
410 if (likely(fd
>= 0)) {
411 int err
= sock_attach_fd(sock
, newfile
, flags
);
413 if (unlikely(err
< 0)) {
418 fd_install(fd
, newfile
);
423 static struct socket
*sock_from_file(struct file
*file
, int *err
)
425 if (file
->f_op
== &socket_file_ops
)
426 return file
->private_data
; /* set in sock_map_fd */
433 * sockfd_lookup - Go from a file number to its socket slot
435 * @err: pointer to an error code return
437 * The file handle passed in is locked and the socket it is bound
438 * too is returned. If an error occurs the err pointer is overwritten
439 * with a negative errno code and NULL is returned. The function checks
440 * for both invalid handles and passing a handle which is not a socket.
442 * On a success the socket object pointer is returned.
445 struct socket
*sockfd_lookup(int fd
, int *err
)
456 sock
= sock_from_file(file
, err
);
462 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
468 file
= fget_light(fd
, fput_needed
);
470 sock
= sock_from_file(file
, err
);
473 fput_light(file
, *fput_needed
);
479 * sock_alloc - allocate a socket
481 * Allocate a new inode and socket object. The two are bound together
482 * and initialised. The socket is then returned. If we are out of inodes
486 static struct socket
*sock_alloc(void)
491 inode
= new_inode(sock_mnt
->mnt_sb
);
495 sock
= SOCKET_I(inode
);
497 kmemcheck_annotate_bitfield(sock
, type
);
498 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
499 inode
->i_uid
= current_fsuid();
500 inode
->i_gid
= current_fsgid();
502 percpu_add(sockets_in_use
, 1);
507 * In theory you can't get an open on this inode, but /proc provides
508 * a back door. Remember to keep it shut otherwise you'll let the
509 * creepy crawlies in.
512 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
517 const struct file_operations bad_sock_fops
= {
518 .owner
= THIS_MODULE
,
519 .open
= sock_no_open
,
523 * sock_release - close a socket
524 * @sock: socket to close
526 * The socket is released from the protocol stack if it has a release
527 * callback, and the inode is then released if the socket is bound to
528 * an inode not a file.
531 void sock_release(struct socket
*sock
)
534 struct module
*owner
= sock
->ops
->owner
;
536 sock
->ops
->release(sock
);
541 if (sock
->fasync_list
)
542 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
544 percpu_sub(sockets_in_use
, 1);
546 iput(SOCK_INODE(sock
));
552 int sock_tx_timestamp(struct msghdr
*msg
, struct sock
*sk
,
553 union skb_shared_tx
*shtx
)
556 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
558 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
562 EXPORT_SYMBOL(sock_tx_timestamp
);
564 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
565 struct msghdr
*msg
, size_t size
)
567 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
575 err
= security_socket_sendmsg(sock
, msg
, size
);
579 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
582 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
585 struct sock_iocb siocb
;
588 init_sync_kiocb(&iocb
, NULL
);
589 iocb
.private = &siocb
;
590 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
591 if (-EIOCBQUEUED
== ret
)
592 ret
= wait_on_sync_kiocb(&iocb
);
596 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
597 struct kvec
*vec
, size_t num
, size_t size
)
599 mm_segment_t oldfs
= get_fs();
604 * the following is safe, since for compiler definitions of kvec and
605 * iovec are identical, yielding the same in-core layout and alignment
607 msg
->msg_iov
= (struct iovec
*)vec
;
608 msg
->msg_iovlen
= num
;
609 result
= sock_sendmsg(sock
, msg
, size
);
614 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
617 *ts
= ktime_to_timespec(kt
);
625 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
627 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
630 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
631 struct timespec ts
[3];
633 struct skb_shared_hwtstamps
*shhwtstamps
=
636 /* Race occurred between timestamp enabling and packet
637 receiving. Fill in the current time for now. */
638 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
639 __net_timestamp(skb
);
641 if (need_software_tstamp
) {
642 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
644 skb_get_timestamp(skb
, &tv
);
645 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
649 skb_get_timestampns(skb
, &ts
);
650 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
656 memset(ts
, 0, sizeof(ts
));
657 if (skb
->tstamp
.tv64
&&
658 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
659 skb_get_timestampns(skb
, ts
+ 0);
663 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
664 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
666 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
667 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
671 put_cmsg(msg
, SOL_SOCKET
,
672 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
675 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
677 inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
679 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
680 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
681 sizeof(__u32
), &skb
->dropcount
);
684 void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
687 sock_recv_timestamp(msg
, sk
, skb
);
688 sock_recv_drops(msg
, sk
, skb
);
690 EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops
);
692 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
693 struct msghdr
*msg
, size_t size
, int flags
)
695 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
703 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
706 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
707 struct msghdr
*msg
, size_t size
, int flags
)
709 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
711 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
714 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
715 size_t size
, int flags
)
718 struct sock_iocb siocb
;
721 init_sync_kiocb(&iocb
, NULL
);
722 iocb
.private = &siocb
;
723 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
724 if (-EIOCBQUEUED
== ret
)
725 ret
= wait_on_sync_kiocb(&iocb
);
729 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
730 size_t size
, int flags
)
733 struct sock_iocb siocb
;
736 init_sync_kiocb(&iocb
, NULL
);
737 iocb
.private = &siocb
;
738 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
739 if (-EIOCBQUEUED
== ret
)
740 ret
= wait_on_sync_kiocb(&iocb
);
744 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
745 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
747 mm_segment_t oldfs
= get_fs();
752 * the following is safe, since for compiler definitions of kvec and
753 * iovec are identical, yielding the same in-core layout and alignment
755 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
756 result
= sock_recvmsg(sock
, msg
, size
, flags
);
761 static void sock_aio_dtor(struct kiocb
*iocb
)
763 kfree(iocb
->private);
766 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
767 int offset
, size_t size
, loff_t
*ppos
, int more
)
772 sock
= file
->private_data
;
774 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
778 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
781 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
782 struct pipe_inode_info
*pipe
, size_t len
,
785 struct socket
*sock
= file
->private_data
;
787 if (unlikely(!sock
->ops
->splice_read
))
790 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
793 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
794 struct sock_iocb
*siocb
)
796 if (!is_sync_kiocb(iocb
)) {
797 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
800 iocb
->ki_dtor
= sock_aio_dtor
;
804 iocb
->private = siocb
;
808 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
809 struct file
*file
, const struct iovec
*iov
,
810 unsigned long nr_segs
)
812 struct socket
*sock
= file
->private_data
;
816 for (i
= 0; i
< nr_segs
; i
++)
817 size
+= iov
[i
].iov_len
;
819 msg
->msg_name
= NULL
;
820 msg
->msg_namelen
= 0;
821 msg
->msg_control
= NULL
;
822 msg
->msg_controllen
= 0;
823 msg
->msg_iov
= (struct iovec
*)iov
;
824 msg
->msg_iovlen
= nr_segs
;
825 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
827 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
830 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
831 unsigned long nr_segs
, loff_t pos
)
833 struct sock_iocb siocb
, *x
;
838 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
842 x
= alloc_sock_iocb(iocb
, &siocb
);
845 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
848 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
849 struct file
*file
, const struct iovec
*iov
,
850 unsigned long nr_segs
)
852 struct socket
*sock
= file
->private_data
;
856 for (i
= 0; i
< nr_segs
; i
++)
857 size
+= iov
[i
].iov_len
;
859 msg
->msg_name
= NULL
;
860 msg
->msg_namelen
= 0;
861 msg
->msg_control
= NULL
;
862 msg
->msg_controllen
= 0;
863 msg
->msg_iov
= (struct iovec
*)iov
;
864 msg
->msg_iovlen
= nr_segs
;
865 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
866 if (sock
->type
== SOCK_SEQPACKET
)
867 msg
->msg_flags
|= MSG_EOR
;
869 return __sock_sendmsg(iocb
, sock
, msg
, size
);
872 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
873 unsigned long nr_segs
, loff_t pos
)
875 struct sock_iocb siocb
, *x
;
880 x
= alloc_sock_iocb(iocb
, &siocb
);
884 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
888 * Atomic setting of ioctl hooks to avoid race
889 * with module unload.
892 static DEFINE_MUTEX(br_ioctl_mutex
);
893 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
) = NULL
;
895 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
897 mutex_lock(&br_ioctl_mutex
);
898 br_ioctl_hook
= hook
;
899 mutex_unlock(&br_ioctl_mutex
);
902 EXPORT_SYMBOL(brioctl_set
);
904 static DEFINE_MUTEX(vlan_ioctl_mutex
);
905 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
907 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
909 mutex_lock(&vlan_ioctl_mutex
);
910 vlan_ioctl_hook
= hook
;
911 mutex_unlock(&vlan_ioctl_mutex
);
914 EXPORT_SYMBOL(vlan_ioctl_set
);
916 static DEFINE_MUTEX(dlci_ioctl_mutex
);
917 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
919 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
921 mutex_lock(&dlci_ioctl_mutex
);
922 dlci_ioctl_hook
= hook
;
923 mutex_unlock(&dlci_ioctl_mutex
);
926 EXPORT_SYMBOL(dlci_ioctl_set
);
928 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
929 unsigned int cmd
, unsigned long arg
)
932 void __user
*argp
= (void __user
*)arg
;
934 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
937 * If this ioctl is unknown try to hand it down
940 if (err
== -ENOIOCTLCMD
)
941 err
= dev_ioctl(net
, cmd
, argp
);
947 * With an ioctl, arg may well be a user mode pointer, but we don't know
948 * what to do with it - that's up to the protocol still.
951 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
955 void __user
*argp
= (void __user
*)arg
;
959 sock
= file
->private_data
;
962 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
963 err
= dev_ioctl(net
, cmd
, argp
);
965 #ifdef CONFIG_WEXT_CORE
966 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
967 err
= dev_ioctl(net
, cmd
, argp
);
974 if (get_user(pid
, (int __user
*)argp
))
976 err
= f_setown(sock
->file
, pid
, 1);
980 err
= put_user(f_getown(sock
->file
),
989 request_module("bridge");
991 mutex_lock(&br_ioctl_mutex
);
993 err
= br_ioctl_hook(net
, cmd
, argp
);
994 mutex_unlock(&br_ioctl_mutex
);
999 if (!vlan_ioctl_hook
)
1000 request_module("8021q");
1002 mutex_lock(&vlan_ioctl_mutex
);
1003 if (vlan_ioctl_hook
)
1004 err
= vlan_ioctl_hook(net
, argp
);
1005 mutex_unlock(&vlan_ioctl_mutex
);
1010 if (!dlci_ioctl_hook
)
1011 request_module("dlci");
1013 mutex_lock(&dlci_ioctl_mutex
);
1014 if (dlci_ioctl_hook
)
1015 err
= dlci_ioctl_hook(cmd
, argp
);
1016 mutex_unlock(&dlci_ioctl_mutex
);
1019 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1025 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1028 struct socket
*sock
= NULL
;
1030 err
= security_socket_create(family
, type
, protocol
, 1);
1034 sock
= sock_alloc();
1041 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1054 /* No kernel lock held - perfect */
1055 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1057 struct socket
*sock
;
1060 * We can't return errors to poll, so it's either yes or no.
1062 sock
= file
->private_data
;
1063 return sock
->ops
->poll(file
, sock
, wait
);
1066 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1068 struct socket
*sock
= file
->private_data
;
1070 return sock
->ops
->mmap(file
, sock
, vma
);
1073 static int sock_close(struct inode
*inode
, struct file
*filp
)
1076 * It was possible the inode is NULL we were
1077 * closing an unfinished socket.
1081 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1084 sock_release(SOCKET_I(inode
));
1089 * Update the socket async list
1091 * Fasync_list locking strategy.
1093 * 1. fasync_list is modified only under process context socket lock
1094 * i.e. under semaphore.
1095 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1096 * or under socket lock.
1097 * 3. fasync_list can be used from softirq context, so that
1098 * modification under socket lock have to be enhanced with
1099 * write_lock_bh(&sk->sk_callback_lock).
1103 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1105 struct fasync_struct
*fa
, *fna
= NULL
, **prev
;
1106 struct socket
*sock
;
1110 fna
= kmalloc(sizeof(struct fasync_struct
), GFP_KERNEL
);
1115 sock
= filp
->private_data
;
1125 spin_lock(&filp
->f_lock
);
1127 filp
->f_flags
|= FASYNC
;
1129 filp
->f_flags
&= ~FASYNC
;
1130 spin_unlock(&filp
->f_lock
);
1132 prev
= &(sock
->fasync_list
);
1134 for (fa
= *prev
; fa
!= NULL
; prev
= &fa
->fa_next
, fa
= *prev
)
1135 if (fa
->fa_file
== filp
)
1140 write_lock_bh(&sk
->sk_callback_lock
);
1142 write_unlock_bh(&sk
->sk_callback_lock
);
1147 fna
->fa_file
= filp
;
1149 fna
->magic
= FASYNC_MAGIC
;
1150 fna
->fa_next
= sock
->fasync_list
;
1151 write_lock_bh(&sk
->sk_callback_lock
);
1152 sock
->fasync_list
= fna
;
1153 sock_set_flag(sk
, SOCK_FASYNC
);
1154 write_unlock_bh(&sk
->sk_callback_lock
);
1157 write_lock_bh(&sk
->sk_callback_lock
);
1158 *prev
= fa
->fa_next
;
1159 if (!sock
->fasync_list
)
1160 sock_reset_flag(sk
, SOCK_FASYNC
);
1161 write_unlock_bh(&sk
->sk_callback_lock
);
1167 release_sock(sock
->sk
);
1171 /* This function may be called only under socket lock or callback_lock */
1173 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1175 if (!sock
|| !sock
->fasync_list
)
1178 case SOCK_WAKE_WAITD
:
1179 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1182 case SOCK_WAKE_SPACE
:
1183 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1188 __kill_fasync(sock
->fasync_list
, SIGIO
, band
);
1191 __kill_fasync(sock
->fasync_list
, SIGURG
, band
);
1196 static int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1197 struct socket
**res
, int kern
)
1200 struct socket
*sock
;
1201 const struct net_proto_family
*pf
;
1204 * Check protocol is in range
1206 if (family
< 0 || family
>= NPROTO
)
1207 return -EAFNOSUPPORT
;
1208 if (type
< 0 || type
>= SOCK_MAX
)
1213 This uglymoron is moved from INET layer to here to avoid
1214 deadlock in module load.
1216 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1220 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1226 err
= security_socket_create(family
, type
, protocol
, kern
);
1231 * Allocate the socket and allow the family to set things up. if
1232 * the protocol is 0, the family is instructed to select an appropriate
1235 sock
= sock_alloc();
1237 if (net_ratelimit())
1238 printk(KERN_WARNING
"socket: no more sockets\n");
1239 return -ENFILE
; /* Not exactly a match, but its the
1240 closest posix thing */
1245 #ifdef CONFIG_MODULES
1246 /* Attempt to load a protocol module if the find failed.
1248 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1249 * requested real, full-featured networking support upon configuration.
1250 * Otherwise module support will break!
1252 if (net_families
[family
] == NULL
)
1253 request_module("net-pf-%d", family
);
1257 pf
= rcu_dereference(net_families
[family
]);
1258 err
= -EAFNOSUPPORT
;
1263 * We will call the ->create function, that possibly is in a loadable
1264 * module, so we have to bump that loadable module refcnt first.
1266 if (!try_module_get(pf
->owner
))
1269 /* Now protected by module ref count */
1272 err
= pf
->create(net
, sock
, protocol
, kern
);
1274 goto out_module_put
;
1277 * Now to bump the refcnt of the [loadable] module that owns this
1278 * socket at sock_release time we decrement its refcnt.
1280 if (!try_module_get(sock
->ops
->owner
))
1281 goto out_module_busy
;
1284 * Now that we're done with the ->create function, the [loadable]
1285 * module can have its refcnt decremented
1287 module_put(pf
->owner
);
1288 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1290 goto out_sock_release
;
1296 err
= -EAFNOSUPPORT
;
1299 module_put(pf
->owner
);
1306 goto out_sock_release
;
1309 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1311 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1314 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1316 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1319 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1322 struct socket
*sock
;
1325 /* Check the SOCK_* constants for consistency. */
1326 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1327 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1328 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1329 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1331 flags
= type
& ~SOCK_TYPE_MASK
;
1332 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1334 type
&= SOCK_TYPE_MASK
;
1336 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1337 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1339 retval
= sock_create(family
, type
, protocol
, &sock
);
1343 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1348 /* It may be already another descriptor 8) Not kernel problem. */
1357 * Create a pair of connected sockets.
1360 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1361 int __user
*, usockvec
)
1363 struct socket
*sock1
, *sock2
;
1365 struct file
*newfile1
, *newfile2
;
1368 flags
= type
& ~SOCK_TYPE_MASK
;
1369 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1371 type
&= SOCK_TYPE_MASK
;
1373 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1374 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1377 * Obtain the first socket and check if the underlying protocol
1378 * supports the socketpair call.
1381 err
= sock_create(family
, type
, protocol
, &sock1
);
1385 err
= sock_create(family
, type
, protocol
, &sock2
);
1389 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1391 goto out_release_both
;
1393 fd1
= sock_alloc_fd(&newfile1
, flags
& O_CLOEXEC
);
1394 if (unlikely(fd1
< 0)) {
1396 goto out_release_both
;
1399 err
= sock_attach_fd(sock1
, newfile1
, flags
& O_NONBLOCK
);
1400 if (unlikely(err
< 0)) {
1403 goto out_release_both
;
1406 fd2
= sock_alloc_fd(&newfile2
, flags
& O_CLOEXEC
);
1407 if (unlikely(fd2
< 0)) {
1411 sock_release(sock2
);
1415 err
= sock_attach_fd(sock2
, newfile2
, flags
& O_NONBLOCK
);
1416 if (unlikely(err
< 0)) {
1421 sock_release(sock2
);
1425 audit_fd_pair(fd1
, fd2
);
1426 fd_install(fd1
, newfile1
);
1427 fd_install(fd2
, newfile2
);
1428 /* fd1 and fd2 may be already another descriptors.
1429 * Not kernel problem.
1432 err
= put_user(fd1
, &usockvec
[0]);
1434 err
= put_user(fd2
, &usockvec
[1]);
1443 sock_release(sock2
);
1445 sock_release(sock1
);
1451 * Bind a name to a socket. Nothing much to do here since it's
1452 * the protocol's responsibility to handle the local address.
1454 * We move the socket address to kernel space before we call
1455 * the protocol layer (having also checked the address is ok).
1458 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1460 struct socket
*sock
;
1461 struct sockaddr_storage address
;
1462 int err
, fput_needed
;
1464 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1466 err
= move_addr_to_kernel(umyaddr
, addrlen
, (struct sockaddr
*)&address
);
1468 err
= security_socket_bind(sock
,
1469 (struct sockaddr
*)&address
,
1472 err
= sock
->ops
->bind(sock
,
1476 fput_light(sock
->file
, fput_needed
);
1482 * Perform a listen. Basically, we allow the protocol to do anything
1483 * necessary for a listen, and if that works, we mark the socket as
1484 * ready for listening.
1487 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1489 struct socket
*sock
;
1490 int err
, fput_needed
;
1493 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1495 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1496 if ((unsigned)backlog
> somaxconn
)
1497 backlog
= somaxconn
;
1499 err
= security_socket_listen(sock
, backlog
);
1501 err
= sock
->ops
->listen(sock
, backlog
);
1503 fput_light(sock
->file
, fput_needed
);
1509 * For accept, we attempt to create a new socket, set up the link
1510 * with the client, wake up the client, then return the new
1511 * connected fd. We collect the address of the connector in kernel
1512 * space and move it to user at the very end. This is unclean because
1513 * we open the socket then return an error.
1515 * 1003.1g adds the ability to recvmsg() to query connection pending
1516 * status to recvmsg. We need to add that support in a way thats
1517 * clean when we restucture accept also.
1520 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1521 int __user
*, upeer_addrlen
, int, flags
)
1523 struct socket
*sock
, *newsock
;
1524 struct file
*newfile
;
1525 int err
, len
, newfd
, fput_needed
;
1526 struct sockaddr_storage address
;
1528 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1531 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1532 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1534 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1539 if (!(newsock
= sock_alloc()))
1542 newsock
->type
= sock
->type
;
1543 newsock
->ops
= sock
->ops
;
1546 * We don't need try_module_get here, as the listening socket (sock)
1547 * has the protocol module (sock->ops->owner) held.
1549 __module_get(newsock
->ops
->owner
);
1551 newfd
= sock_alloc_fd(&newfile
, flags
& O_CLOEXEC
);
1552 if (unlikely(newfd
< 0)) {
1554 sock_release(newsock
);
1558 err
= sock_attach_fd(newsock
, newfile
, flags
& O_NONBLOCK
);
1562 err
= security_socket_accept(sock
, newsock
);
1566 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1570 if (upeer_sockaddr
) {
1571 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1573 err
= -ECONNABORTED
;
1576 err
= move_addr_to_user((struct sockaddr
*)&address
,
1577 len
, upeer_sockaddr
, upeer_addrlen
);
1582 /* File flags are not inherited via accept() unlike another OSes. */
1584 fd_install(newfd
, newfile
);
1588 fput_light(sock
->file
, fput_needed
);
1592 sock_release(newsock
);
1594 put_unused_fd(newfd
);
1598 put_unused_fd(newfd
);
1602 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1603 int __user
*, upeer_addrlen
)
1605 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1609 * Attempt to connect to a socket with the server address. The address
1610 * is in user space so we verify it is OK and move it to kernel space.
1612 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1615 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1616 * other SEQPACKET protocols that take time to connect() as it doesn't
1617 * include the -EINPROGRESS status for such sockets.
1620 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1623 struct socket
*sock
;
1624 struct sockaddr_storage address
;
1625 int err
, fput_needed
;
1627 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1630 err
= move_addr_to_kernel(uservaddr
, addrlen
, (struct sockaddr
*)&address
);
1635 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1639 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1640 sock
->file
->f_flags
);
1642 fput_light(sock
->file
, fput_needed
);
1648 * Get the local address ('name') of a socket object. Move the obtained
1649 * name to user space.
1652 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1653 int __user
*, usockaddr_len
)
1655 struct socket
*sock
;
1656 struct sockaddr_storage address
;
1657 int len
, err
, fput_needed
;
1659 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1663 err
= security_socket_getsockname(sock
);
1667 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1670 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
, usockaddr_len
);
1673 fput_light(sock
->file
, fput_needed
);
1679 * Get the remote address ('name') of a socket object. Move the obtained
1680 * name to user space.
1683 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1684 int __user
*, usockaddr_len
)
1686 struct socket
*sock
;
1687 struct sockaddr_storage address
;
1688 int len
, err
, fput_needed
;
1690 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1692 err
= security_socket_getpeername(sock
);
1694 fput_light(sock
->file
, fput_needed
);
1699 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1702 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
,
1704 fput_light(sock
->file
, fput_needed
);
1710 * Send a datagram to a given address. We move the address into kernel
1711 * space and check the user space data area is readable before invoking
1715 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1716 unsigned, flags
, struct sockaddr __user
*, addr
,
1719 struct socket
*sock
;
1720 struct sockaddr_storage address
;
1726 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1730 iov
.iov_base
= buff
;
1732 msg
.msg_name
= NULL
;
1735 msg
.msg_control
= NULL
;
1736 msg
.msg_controllen
= 0;
1737 msg
.msg_namelen
= 0;
1739 err
= move_addr_to_kernel(addr
, addr_len
, (struct sockaddr
*)&address
);
1742 msg
.msg_name
= (struct sockaddr
*)&address
;
1743 msg
.msg_namelen
= addr_len
;
1745 if (sock
->file
->f_flags
& O_NONBLOCK
)
1746 flags
|= MSG_DONTWAIT
;
1747 msg
.msg_flags
= flags
;
1748 err
= sock_sendmsg(sock
, &msg
, len
);
1751 fput_light(sock
->file
, fput_needed
);
1757 * Send a datagram down a socket.
1760 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1763 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1767 * Receive a frame from the socket and optionally record the address of the
1768 * sender. We verify the buffers are writable and if needed move the
1769 * sender address from kernel to user space.
1772 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1773 unsigned, flags
, struct sockaddr __user
*, addr
,
1774 int __user
*, addr_len
)
1776 struct socket
*sock
;
1779 struct sockaddr_storage address
;
1783 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1787 msg
.msg_control
= NULL
;
1788 msg
.msg_controllen
= 0;
1792 iov
.iov_base
= ubuf
;
1793 msg
.msg_name
= (struct sockaddr
*)&address
;
1794 msg
.msg_namelen
= sizeof(address
);
1795 if (sock
->file
->f_flags
& O_NONBLOCK
)
1796 flags
|= MSG_DONTWAIT
;
1797 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1799 if (err
>= 0 && addr
!= NULL
) {
1800 err2
= move_addr_to_user((struct sockaddr
*)&address
,
1801 msg
.msg_namelen
, addr
, addr_len
);
1806 fput_light(sock
->file
, fput_needed
);
1812 * Receive a datagram from a socket.
1815 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1818 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1822 * Set a socket option. Because we don't know the option lengths we have
1823 * to pass the user mode parameter for the protocols to sort out.
1826 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1827 char __user
*, optval
, int, optlen
)
1829 int err
, fput_needed
;
1830 struct socket
*sock
;
1835 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1837 err
= security_socket_setsockopt(sock
, level
, optname
);
1841 if (level
== SOL_SOCKET
)
1843 sock_setsockopt(sock
, level
, optname
, optval
,
1847 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1850 fput_light(sock
->file
, fput_needed
);
1856 * Get a socket option. Because we don't know the option lengths we have
1857 * to pass a user mode parameter for the protocols to sort out.
1860 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1861 char __user
*, optval
, int __user
*, optlen
)
1863 int err
, fput_needed
;
1864 struct socket
*sock
;
1866 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1868 err
= security_socket_getsockopt(sock
, level
, optname
);
1872 if (level
== SOL_SOCKET
)
1874 sock_getsockopt(sock
, level
, optname
, optval
,
1878 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1881 fput_light(sock
->file
, fput_needed
);
1887 * Shutdown a socket.
1890 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1892 int err
, fput_needed
;
1893 struct socket
*sock
;
1895 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1897 err
= security_socket_shutdown(sock
, how
);
1899 err
= sock
->ops
->shutdown(sock
, how
);
1900 fput_light(sock
->file
, fput_needed
);
1905 /* A couple of helpful macros for getting the address of the 32/64 bit
1906 * fields which are the same type (int / unsigned) on our platforms.
1908 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1909 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1910 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1913 * BSD sendmsg interface
1916 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned, flags
)
1918 struct compat_msghdr __user
*msg_compat
=
1919 (struct compat_msghdr __user
*)msg
;
1920 struct socket
*sock
;
1921 struct sockaddr_storage address
;
1922 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1923 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1924 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1925 /* 20 is size of ipv6_pktinfo */
1926 unsigned char *ctl_buf
= ctl
;
1927 struct msghdr msg_sys
;
1928 int err
, ctl_len
, iov_size
, total_len
;
1932 if (MSG_CMSG_COMPAT
& flags
) {
1933 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1936 else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1939 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1943 /* do not move before msg_sys is valid */
1945 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1948 /* Check whether to allocate the iovec area */
1950 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1951 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1952 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1957 /* This will also move the address data into kernel space */
1958 if (MSG_CMSG_COMPAT
& flags
) {
1959 err
= verify_compat_iovec(&msg_sys
, iov
,
1960 (struct sockaddr
*)&address
,
1963 err
= verify_iovec(&msg_sys
, iov
,
1964 (struct sockaddr
*)&address
,
1972 if (msg_sys
.msg_controllen
> INT_MAX
)
1974 ctl_len
= msg_sys
.msg_controllen
;
1975 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1977 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1981 ctl_buf
= msg_sys
.msg_control
;
1982 ctl_len
= msg_sys
.msg_controllen
;
1983 } else if (ctl_len
) {
1984 if (ctl_len
> sizeof(ctl
)) {
1985 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1986 if (ctl_buf
== NULL
)
1991 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1992 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1993 * checking falls down on this.
1995 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1998 msg_sys
.msg_control
= ctl_buf
;
2000 msg_sys
.msg_flags
= flags
;
2002 if (sock
->file
->f_flags
& O_NONBLOCK
)
2003 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
2004 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
2008 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2010 if (iov
!= iovstack
)
2011 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2013 fput_light(sock
->file
, fput_needed
);
2018 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
2019 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
2021 struct compat_msghdr __user
*msg_compat
=
2022 (struct compat_msghdr __user
*)msg
;
2023 struct iovec iovstack
[UIO_FASTIOV
];
2024 struct iovec
*iov
= iovstack
;
2025 unsigned long cmsg_ptr
;
2026 int err
, iov_size
, total_len
, len
;
2028 /* kernel mode address */
2029 struct sockaddr_storage addr
;
2031 /* user mode address pointers */
2032 struct sockaddr __user
*uaddr
;
2033 int __user
*uaddr_len
;
2035 if (MSG_CMSG_COMPAT
& flags
) {
2036 if (get_compat_msghdr(msg_sys
, msg_compat
))
2039 else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2043 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2046 /* Check whether to allocate the iovec area */
2048 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
2049 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2050 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
2056 * Save the user-mode address (verify_iovec will change the
2057 * kernel msghdr to use the kernel address space)
2060 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2061 uaddr_len
= COMPAT_NAMELEN(msg
);
2062 if (MSG_CMSG_COMPAT
& flags
) {
2063 err
= verify_compat_iovec(msg_sys
, iov
,
2064 (struct sockaddr
*)&addr
,
2067 err
= verify_iovec(msg_sys
, iov
,
2068 (struct sockaddr
*)&addr
,
2074 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2075 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2077 if (sock
->file
->f_flags
& O_NONBLOCK
)
2078 flags
|= MSG_DONTWAIT
;
2079 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2085 if (uaddr
!= NULL
) {
2086 err
= move_addr_to_user((struct sockaddr
*)&addr
,
2087 msg_sys
->msg_namelen
, uaddr
,
2092 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2096 if (MSG_CMSG_COMPAT
& flags
)
2097 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2098 &msg_compat
->msg_controllen
);
2100 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2101 &msg
->msg_controllen
);
2107 if (iov
!= iovstack
)
2108 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2114 * BSD recvmsg interface
2117 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2118 unsigned int, flags
)
2120 int fput_needed
, err
;
2121 struct msghdr msg_sys
;
2122 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2127 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2129 fput_light(sock
->file
, fput_needed
);
2135 * Linux recvmmsg interface
2138 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2139 unsigned int flags
, struct timespec
*timeout
)
2141 int fput_needed
, err
, datagrams
;
2142 struct socket
*sock
;
2143 struct mmsghdr __user
*entry
;
2144 struct compat_mmsghdr __user
*compat_entry
;
2145 struct msghdr msg_sys
;
2146 struct timespec end_time
;
2149 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2155 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2159 err
= sock_error(sock
->sk
);
2164 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2166 while (datagrams
< vlen
) {
2168 * No need to ask LSM for more than the first datagram.
2170 if (MSG_CMSG_COMPAT
& flags
) {
2171 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2172 &msg_sys
, flags
, datagrams
);
2175 err
= __put_user(err
, &compat_entry
->msg_len
);
2178 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2179 &msg_sys
, flags
, datagrams
);
2182 err
= put_user(err
, &entry
->msg_len
);
2191 ktime_get_ts(timeout
);
2192 *timeout
= timespec_sub(end_time
, *timeout
);
2193 if (timeout
->tv_sec
< 0) {
2194 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2198 /* Timeout, return less than vlen datagrams */
2199 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2203 /* Out of band data, return right away */
2204 if (msg_sys
.msg_flags
& MSG_OOB
)
2209 fput_light(sock
->file
, fput_needed
);
2214 if (datagrams
!= 0) {
2216 * We may return less entries than requested (vlen) if the
2217 * sock is non block and there aren't enough datagrams...
2219 if (err
!= -EAGAIN
) {
2221 * ... or if recvmsg returns an error after we
2222 * received some datagrams, where we record the
2223 * error to return on the next call or if the
2224 * app asks about it using getsockopt(SO_ERROR).
2226 sock
->sk
->sk_err
= -err
;
2235 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2236 unsigned int, vlen
, unsigned int, flags
,
2237 struct timespec __user
*, timeout
)
2240 struct timespec timeout_sys
;
2243 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2245 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2248 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2250 if (datagrams
> 0 &&
2251 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2252 datagrams
= -EFAULT
;
2257 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2258 /* Argument list sizes for sys_socketcall */
2259 #define AL(x) ((x) * sizeof(unsigned long))
2260 static const unsigned char nargs
[20] = {
2261 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2262 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2263 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2270 * System call vectors.
2272 * Argument checking cleaned up. Saved 20% in size.
2273 * This function doesn't need to set the kernel lock because
2274 * it is set by the callees.
2277 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2280 unsigned long a0
, a1
;
2284 if (call
< 1 || call
> SYS_RECVMMSG
)
2288 if (len
> sizeof(a
))
2291 /* copy_from_user should be SMP safe. */
2292 if (copy_from_user(a
, args
, len
))
2295 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2302 err
= sys_socket(a0
, a1
, a
[2]);
2305 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2308 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2311 err
= sys_listen(a0
, a1
);
2314 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2315 (int __user
*)a
[2], 0);
2317 case SYS_GETSOCKNAME
:
2319 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2320 (int __user
*)a
[2]);
2322 case SYS_GETPEERNAME
:
2324 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2325 (int __user
*)a
[2]);
2327 case SYS_SOCKETPAIR
:
2328 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2331 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2334 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2335 (struct sockaddr __user
*)a
[4], a
[5]);
2338 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2341 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2342 (struct sockaddr __user
*)a
[4],
2343 (int __user
*)a
[5]);
2346 err
= sys_shutdown(a0
, a1
);
2348 case SYS_SETSOCKOPT
:
2349 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2351 case SYS_GETSOCKOPT
:
2353 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2354 (int __user
*)a
[4]);
2357 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2360 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2363 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2364 (struct timespec __user
*)a
[4]);
2367 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2368 (int __user
*)a
[2], a
[3]);
2377 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2380 * sock_register - add a socket protocol handler
2381 * @ops: description of protocol
2383 * This function is called by a protocol handler that wants to
2384 * advertise its address family, and have it linked into the
2385 * socket interface. The value ops->family coresponds to the
2386 * socket system call protocol family.
2388 int sock_register(const struct net_proto_family
*ops
)
2392 if (ops
->family
>= NPROTO
) {
2393 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2398 spin_lock(&net_family_lock
);
2399 if (net_families
[ops
->family
])
2402 net_families
[ops
->family
] = ops
;
2405 spin_unlock(&net_family_lock
);
2407 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2412 * sock_unregister - remove a protocol handler
2413 * @family: protocol family to remove
2415 * This function is called by a protocol handler that wants to
2416 * remove its address family, and have it unlinked from the
2417 * new socket creation.
2419 * If protocol handler is a module, then it can use module reference
2420 * counts to protect against new references. If protocol handler is not
2421 * a module then it needs to provide its own protection in
2422 * the ops->create routine.
2424 void sock_unregister(int family
)
2426 BUG_ON(family
< 0 || family
>= NPROTO
);
2428 spin_lock(&net_family_lock
);
2429 net_families
[family
] = NULL
;
2430 spin_unlock(&net_family_lock
);
2434 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2437 static int __init
sock_init(void)
2440 * Initialize sock SLAB cache.
2446 * Initialize skbuff SLAB cache
2451 * Initialize the protocols module.
2455 register_filesystem(&sock_fs_type
);
2456 sock_mnt
= kern_mount(&sock_fs_type
);
2458 /* The real protocol initialization is performed in later initcalls.
2461 #ifdef CONFIG_NETFILTER
2468 core_initcall(sock_init
); /* early initcall */
2470 #ifdef CONFIG_PROC_FS
2471 void socket_seq_show(struct seq_file
*seq
)
2476 for_each_possible_cpu(cpu
)
2477 counter
+= per_cpu(sockets_in_use
, cpu
);
2479 /* It can be negative, by the way. 8) */
2483 seq_printf(seq
, "sockets: used %d\n", counter
);
2485 #endif /* CONFIG_PROC_FS */
2487 #ifdef CONFIG_COMPAT
2488 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2489 unsigned int cmd
, struct compat_timeval __user
*up
)
2491 mm_segment_t old_fs
= get_fs();
2496 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2499 err
= put_user(ktv
.tv_sec
, &up
->tv_sec
);
2500 err
|= __put_user(ktv
.tv_usec
, &up
->tv_usec
);
2505 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2506 unsigned int cmd
, struct compat_timespec __user
*up
)
2508 mm_segment_t old_fs
= get_fs();
2509 struct timespec kts
;
2513 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2516 err
= put_user(kts
.tv_sec
, &up
->tv_sec
);
2517 err
|= __put_user(kts
.tv_nsec
, &up
->tv_nsec
);
2522 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2524 struct ifreq __user
*uifr
;
2527 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2528 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2531 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2535 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2541 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2543 struct compat_ifconf ifc32
;
2545 struct ifconf __user
*uifc
;
2546 struct compat_ifreq __user
*ifr32
;
2547 struct ifreq __user
*ifr
;
2551 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2554 if (ifc32
.ifcbuf
== 0) {
2558 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2560 size_t len
=((ifc32
.ifc_len
/ sizeof (struct compat_ifreq
)) + 1) *
2561 sizeof (struct ifreq
);
2562 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2564 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2565 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2566 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof (struct compat_ifreq
)) {
2567 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2573 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2576 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2580 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2584 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2586 i
+ sizeof (struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2587 i
+= sizeof (struct compat_ifreq
), j
+= sizeof (struct ifreq
)) {
2588 if (copy_in_user(ifr32
, ifr
, sizeof (struct compat_ifreq
)))
2594 if (ifc32
.ifcbuf
== 0) {
2595 /* Translate from 64-bit structure multiple to
2599 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2604 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2610 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2612 struct ifreq __user
*ifr
;
2616 ifr
= compat_alloc_user_space(sizeof(*ifr
));
2618 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2621 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2624 datap
= compat_ptr(data
);
2625 if (put_user(datap
, &ifr
->ifr_ifru
.ifru_data
))
2628 return dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2631 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2634 compat_uptr_t uptr32
;
2635 struct ifreq __user
*uifr
;
2637 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2638 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2641 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2644 uptr
= compat_ptr(uptr32
);
2646 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2649 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2652 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2653 struct compat_ifreq __user
*ifr32
)
2656 struct ifreq __user
*uifr
;
2657 mm_segment_t old_fs
;
2663 case SIOCBONDENSLAVE
:
2664 case SIOCBONDRELEASE
:
2665 case SIOCBONDSETHWADDR
:
2666 case SIOCBONDCHANGEACTIVE
:
2667 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2672 err
= dev_ioctl(net
, cmd
, &kifr
);
2676 case SIOCBONDSLAVEINFOQUERY
:
2677 case SIOCBONDINFOQUERY
:
2678 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2679 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2682 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2685 datap
= compat_ptr(data
);
2686 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2689 return dev_ioctl(net
, cmd
, uifr
);
2695 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2696 struct compat_ifreq __user
*u_ifreq32
)
2698 struct ifreq __user
*u_ifreq64
;
2699 char tmp_buf
[IFNAMSIZ
];
2700 void __user
*data64
;
2703 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2706 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2708 data64
= compat_ptr(data32
);
2710 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2712 /* Don't check these user accesses, just let that get trapped
2713 * in the ioctl handler instead.
2715 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2718 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2721 return dev_ioctl(net
, cmd
, u_ifreq64
);
2724 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2725 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2727 struct ifreq __user
*uifr
;
2730 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2731 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2734 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2745 case SIOCGIFBRDADDR
:
2746 case SIOCGIFDSTADDR
:
2747 case SIOCGIFNETMASK
:
2752 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2760 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2761 struct compat_ifreq __user
*uifr32
)
2764 struct compat_ifmap __user
*uifmap32
;
2765 mm_segment_t old_fs
;
2768 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2769 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2770 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2771 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2772 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2773 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2774 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2775 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2781 err
= dev_ioctl(net
, cmd
, (void __user
*)&ifr
);
2784 if (cmd
== SIOCGIFMAP
&& !err
) {
2785 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2786 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2787 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2788 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2789 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2790 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2791 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2798 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2801 compat_uptr_t uptr32
;
2802 struct ifreq __user
*uifr
;
2804 uifr
= compat_alloc_user_space(sizeof (*uifr
));
2805 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2808 if (get_user(uptr32
, &uifr32
->ifr_data
))
2811 uptr
= compat_ptr(uptr32
);
2813 if (put_user(uptr
, &uifr
->ifr_data
))
2816 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
2821 struct sockaddr rt_dst
; /* target address */
2822 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2823 struct sockaddr rt_genmask
; /* target network mask (IP) */
2824 unsigned short rt_flags
;
2827 unsigned char rt_tos
;
2828 unsigned char rt_class
;
2830 short rt_metric
; /* +1 for binary compatibility! */
2831 /* char * */ u32 rt_dev
; /* forcing the device at add */
2832 u32 rt_mtu
; /* per route MTU/Window */
2833 u32 rt_window
; /* Window clamping */
2834 unsigned short rt_irtt
; /* Initial RTT */
2837 struct in6_rtmsg32
{
2838 struct in6_addr rtmsg_dst
;
2839 struct in6_addr rtmsg_src
;
2840 struct in6_addr rtmsg_gateway
;
2850 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
2851 unsigned int cmd
, void __user
*argp
)
2855 struct in6_rtmsg r6
;
2859 mm_segment_t old_fs
= get_fs();
2861 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
2862 struct in6_rtmsg32 __user
*ur6
= argp
;
2863 ret
= copy_from_user (&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
2864 3 * sizeof(struct in6_addr
));
2865 ret
|= __get_user (r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
2866 ret
|= __get_user (r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
2867 ret
|= __get_user (r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
2868 ret
|= __get_user (r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
2869 ret
|= __get_user (r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
2870 ret
|= __get_user (r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
2871 ret
|= __get_user (r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
2875 struct rtentry32 __user
*ur4
= argp
;
2876 ret
= copy_from_user (&r4
.rt_dst
, &(ur4
->rt_dst
),
2877 3 * sizeof(struct sockaddr
));
2878 ret
|= __get_user (r4
.rt_flags
, &(ur4
->rt_flags
));
2879 ret
|= __get_user (r4
.rt_metric
, &(ur4
->rt_metric
));
2880 ret
|= __get_user (r4
.rt_mtu
, &(ur4
->rt_mtu
));
2881 ret
|= __get_user (r4
.rt_window
, &(ur4
->rt_window
));
2882 ret
|= __get_user (r4
.rt_irtt
, &(ur4
->rt_irtt
));
2883 ret
|= __get_user (rtdev
, &(ur4
->rt_dev
));
2885 ret
|= copy_from_user (devname
, compat_ptr(rtdev
), 15);
2886 r4
.rt_dev
= devname
; devname
[15] = 0;
2899 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
2906 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2907 * for some operations; this forces use of the newer bridge-utils that
2908 * use compatiable ioctls
2910 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
2914 if (get_user(tmp
, argp
))
2916 if (tmp
== BRCTL_GET_VERSION
)
2917 return BRCTL_VERSION
+ 1;
2921 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
2922 unsigned int cmd
, unsigned long arg
)
2924 void __user
*argp
= compat_ptr(arg
);
2925 struct sock
*sk
= sock
->sk
;
2926 struct net
*net
= sock_net(sk
);
2928 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
2929 return siocdevprivate_ioctl(net
, cmd
, argp
);
2934 return old_bridge_ioctl(argp
);
2936 return dev_ifname32(net
, argp
);
2938 return dev_ifconf(net
, argp
);
2940 return ethtool_ioctl(net
, argp
);
2942 return compat_siocwandev(net
, argp
);
2945 return compat_sioc_ifmap(net
, cmd
, argp
);
2946 case SIOCBONDENSLAVE
:
2947 case SIOCBONDRELEASE
:
2948 case SIOCBONDSETHWADDR
:
2949 case SIOCBONDSLAVEINFOQUERY
:
2950 case SIOCBONDINFOQUERY
:
2951 case SIOCBONDCHANGEACTIVE
:
2952 return bond_ioctl(net
, cmd
, argp
);
2955 return routing_ioctl(net
, sock
, cmd
, argp
);
2957 return do_siocgstamp(net
, sock
, cmd
, argp
);
2959 return do_siocgstampns(net
, sock
, cmd
, argp
);
2961 return compat_siocshwtstamp(net
, argp
);
2973 return sock_ioctl(file
, cmd
, arg
);
2990 case SIOCSIFHWBROADCAST
:
2992 case SIOCGIFBRDADDR
:
2993 case SIOCSIFBRDADDR
:
2994 case SIOCGIFDSTADDR
:
2995 case SIOCSIFDSTADDR
:
2996 case SIOCGIFNETMASK
:
2997 case SIOCSIFNETMASK
:
3008 return dev_ifsioc(net
, sock
, cmd
, argp
);
3014 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3017 /* Prevent warning from compat_sys_ioctl, these always
3018 * result in -EINVAL in the native case anyway. */
3031 return -ENOIOCTLCMD
;
3034 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
3037 struct socket
*sock
= file
->private_data
;
3038 int ret
= -ENOIOCTLCMD
;
3045 if (sock
->ops
->compat_ioctl
)
3046 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3048 if (ret
== -ENOIOCTLCMD
&&
3049 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3050 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3052 if (ret
== -ENOIOCTLCMD
)
3053 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3059 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3061 return sock
->ops
->bind(sock
, addr
, addrlen
);
3064 int kernel_listen(struct socket
*sock
, int backlog
)
3066 return sock
->ops
->listen(sock
, backlog
);
3069 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3071 struct sock
*sk
= sock
->sk
;
3074 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3079 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3081 sock_release(*newsock
);
3086 (*newsock
)->ops
= sock
->ops
;
3087 __module_get((*newsock
)->ops
->owner
);
3093 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3096 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3099 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3102 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3105 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3108 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3111 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3112 char *optval
, int *optlen
)
3114 mm_segment_t oldfs
= get_fs();
3118 if (level
== SOL_SOCKET
)
3119 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
3121 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
3127 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3128 char *optval
, unsigned int optlen
)
3130 mm_segment_t oldfs
= get_fs();
3134 if (level
== SOL_SOCKET
)
3135 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
3137 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
3143 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3144 size_t size
, int flags
)
3146 if (sock
->ops
->sendpage
)
3147 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3149 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3152 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3154 mm_segment_t oldfs
= get_fs();
3158 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3164 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3166 return sock
->ops
->shutdown(sock
, how
);
3169 EXPORT_SYMBOL(sock_create
);
3170 EXPORT_SYMBOL(sock_create_kern
);
3171 EXPORT_SYMBOL(sock_create_lite
);
3172 EXPORT_SYMBOL(sock_map_fd
);
3173 EXPORT_SYMBOL(sock_recvmsg
);
3174 EXPORT_SYMBOL(sock_register
);
3175 EXPORT_SYMBOL(sock_release
);
3176 EXPORT_SYMBOL(sock_sendmsg
);
3177 EXPORT_SYMBOL(sock_unregister
);
3178 EXPORT_SYMBOL(sock_wake_async
);
3179 EXPORT_SYMBOL(sockfd_lookup
);
3180 EXPORT_SYMBOL(kernel_sendmsg
);
3181 EXPORT_SYMBOL(kernel_recvmsg
);
3182 EXPORT_SYMBOL(kernel_bind
);
3183 EXPORT_SYMBOL(kernel_listen
);
3184 EXPORT_SYMBOL(kernel_accept
);
3185 EXPORT_SYMBOL(kernel_connect
);
3186 EXPORT_SYMBOL(kernel_getsockname
);
3187 EXPORT_SYMBOL(kernel_getpeername
);
3188 EXPORT_SYMBOL(kernel_getsockopt
);
3189 EXPORT_SYMBOL(kernel_setsockopt
);
3190 EXPORT_SYMBOL(kernel_sendpage
);
3191 EXPORT_SYMBOL(kernel_sock_ioctl
);
3192 EXPORT_SYMBOL(kernel_sock_shutdown
);