2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
92 #include <asm/uaccess.h>
93 #include <asm/unistd.h>
95 #include <net/compat.h>
97 #include <net/cls_cgroup.h>
100 #include <linux/netfilter.h>
102 #include <linux/if_tun.h>
103 #include <linux/ipv6_route.h>
104 #include <linux/route.h>
105 #include <linux/sockios.h>
106 #include <linux/atalk.h>
108 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
109 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
110 unsigned long nr_segs
, loff_t pos
);
111 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
112 unsigned long nr_segs
, loff_t pos
);
113 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
115 static int sock_close(struct inode
*inode
, struct file
*file
);
116 static unsigned int sock_poll(struct file
*file
,
117 struct poll_table_struct
*wait
);
118 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
120 static long compat_sock_ioctl(struct file
*file
,
121 unsigned int cmd
, unsigned long arg
);
123 static int sock_fasync(int fd
, struct file
*filp
, int on
);
124 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
125 int offset
, size_t size
, loff_t
*ppos
, int more
);
126 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
127 struct pipe_inode_info
*pipe
, size_t len
,
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
135 static const struct file_operations socket_file_ops
= {
136 .owner
= THIS_MODULE
,
138 .aio_read
= sock_aio_read
,
139 .aio_write
= sock_aio_write
,
141 .unlocked_ioctl
= sock_ioctl
,
143 .compat_ioctl
= compat_sock_ioctl
,
146 .open
= sock_no_open
, /* special open code to disallow open via /proc */
147 .release
= sock_close
,
148 .fasync
= sock_fasync
,
149 .sendpage
= sock_sendpage
,
150 .splice_write
= generic_splice_sendpage
,
151 .splice_read
= sock_splice_read
,
155 * The protocol list. Each protocol is registered in here.
158 static DEFINE_SPINLOCK(net_family_lock
);
159 static const struct net_proto_family
*net_families
[NPROTO
] __read_mostly
;
162 * Statistics counters of the socket lists
165 static DEFINE_PER_CPU(int, sockets_in_use
);
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr
*kaddr
)
186 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
190 if (copy_from_user(kaddr
, uaddr
, ulen
))
192 return audit_sockaddr(ulen
, kaddr
);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 int move_addr_to_user(struct sockaddr
*kaddr
, int klen
, void __user
*uaddr
,
218 err
= get_user(len
, ulen
);
223 if (len
< 0 || len
> sizeof(struct sockaddr_storage
))
226 if (audit_sockaddr(klen
, kaddr
))
228 if (copy_to_user(uaddr
, kaddr
, len
))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen
, ulen
);
238 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
240 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
242 struct socket_alloc
*ei
;
244 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
247 ei
->socket
.wq
= kmalloc(sizeof(struct socket_wq
), GFP_KERNEL
);
248 if (!ei
->socket
.wq
) {
249 kmem_cache_free(sock_inode_cachep
, ei
);
252 init_waitqueue_head(&ei
->socket
.wq
->wait
);
253 ei
->socket
.wq
->fasync_list
= NULL
;
255 ei
->socket
.state
= SS_UNCONNECTED
;
256 ei
->socket
.flags
= 0;
257 ei
->socket
.ops
= NULL
;
258 ei
->socket
.sk
= NULL
;
259 ei
->socket
.file
= NULL
;
261 return &ei
->vfs_inode
;
265 static void wq_free_rcu(struct rcu_head
*head
)
267 struct socket_wq
*wq
= container_of(head
, struct socket_wq
, rcu
);
272 static void sock_destroy_inode(struct inode
*inode
)
274 struct socket_alloc
*ei
;
276 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
277 call_rcu(&ei
->socket
.wq
->rcu
, wq_free_rcu
);
278 kmem_cache_free(sock_inode_cachep
, ei
);
281 static void init_once(void *foo
)
283 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
285 inode_init_once(&ei
->vfs_inode
);
288 static int init_inodecache(void)
290 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
291 sizeof(struct socket_alloc
),
293 (SLAB_HWCACHE_ALIGN
|
294 SLAB_RECLAIM_ACCOUNT
|
297 if (sock_inode_cachep
== NULL
)
302 static const struct super_operations sockfs_ops
= {
303 .alloc_inode
= sock_alloc_inode
,
304 .destroy_inode
= sock_destroy_inode
,
305 .statfs
= simple_statfs
,
308 static int sockfs_get_sb(struct file_system_type
*fs_type
,
309 int flags
, const char *dev_name
, void *data
,
310 struct vfsmount
*mnt
)
312 return get_sb_pseudo(fs_type
, "socket:", &sockfs_ops
, SOCKFS_MAGIC
,
316 static struct vfsmount
*sock_mnt __read_mostly
;
318 static struct file_system_type sock_fs_type
= {
320 .get_sb
= sockfs_get_sb
,
321 .kill_sb
= kill_anon_super
,
325 * sockfs_dname() is called from d_path().
327 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
329 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
330 dentry
->d_inode
->i_ino
);
333 static const struct dentry_operations sockfs_dentry_operations
= {
334 .d_dname
= sockfs_dname
,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 static int sock_alloc_file(struct socket
*sock
, struct file
**f
, int flags
)
356 struct qstr name
= { .name
= "" };
361 fd
= get_unused_fd_flags(flags
);
362 if (unlikely(fd
< 0))
365 path
.dentry
= d_alloc(sock_mnt
->mnt_sb
->s_root
, &name
);
366 if (unlikely(!path
.dentry
)) {
370 path
.mnt
= mntget(sock_mnt
);
372 path
.dentry
->d_op
= &sockfs_dentry_operations
;
373 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
374 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
376 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
378 if (unlikely(!file
)) {
379 /* drop dentry, keep inode */
380 atomic_inc(&path
.dentry
->d_inode
->i_count
);
387 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
389 file
->private_data
= sock
;
395 int sock_map_fd(struct socket
*sock
, int flags
)
397 struct file
*newfile
;
398 int fd
= sock_alloc_file(sock
, &newfile
, flags
);
401 fd_install(fd
, newfile
);
405 EXPORT_SYMBOL(sock_map_fd
);
407 static struct socket
*sock_from_file(struct file
*file
, int *err
)
409 if (file
->f_op
== &socket_file_ops
)
410 return file
->private_data
; /* set in sock_map_fd */
417 * sockfd_lookup - Go from a file number to its socket slot
419 * @err: pointer to an error code return
421 * The file handle passed in is locked and the socket it is bound
422 * too is returned. If an error occurs the err pointer is overwritten
423 * with a negative errno code and NULL is returned. The function checks
424 * for both invalid handles and passing a handle which is not a socket.
426 * On a success the socket object pointer is returned.
429 struct socket
*sockfd_lookup(int fd
, int *err
)
440 sock
= sock_from_file(file
, err
);
445 EXPORT_SYMBOL(sockfd_lookup
);
447 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
453 file
= fget_light(fd
, fput_needed
);
455 sock
= sock_from_file(file
, err
);
458 fput_light(file
, *fput_needed
);
464 * sock_alloc - allocate a socket
466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
471 static struct socket
*sock_alloc(void)
476 inode
= new_inode(sock_mnt
->mnt_sb
);
480 sock
= SOCKET_I(inode
);
482 kmemcheck_annotate_bitfield(sock
, type
);
483 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
484 inode
->i_uid
= current_fsuid();
485 inode
->i_gid
= current_fsgid();
487 percpu_add(sockets_in_use
, 1);
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
497 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
502 const struct file_operations bad_sock_fops
= {
503 .owner
= THIS_MODULE
,
504 .open
= sock_no_open
,
508 * sock_release - close a socket
509 * @sock: socket to close
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
513 * an inode not a file.
516 void sock_release(struct socket
*sock
)
519 struct module
*owner
= sock
->ops
->owner
;
521 sock
->ops
->release(sock
);
526 if (sock
->wq
->fasync_list
)
527 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
529 percpu_sub(sockets_in_use
, 1);
531 iput(SOCK_INODE(sock
));
536 EXPORT_SYMBOL(sock_release
);
538 int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
)
541 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
542 *tx_flags
|= SKBTX_HW_TSTAMP
;
543 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
544 *tx_flags
|= SKBTX_SW_TSTAMP
;
547 EXPORT_SYMBOL(sock_tx_timestamp
);
549 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
550 struct msghdr
*msg
, size_t size
)
552 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
555 sock_update_classid(sock
->sk
);
562 err
= security_socket_sendmsg(sock
, msg
, size
);
566 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
569 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
572 struct sock_iocb siocb
;
575 init_sync_kiocb(&iocb
, NULL
);
576 iocb
.private = &siocb
;
577 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
578 if (-EIOCBQUEUED
== ret
)
579 ret
= wait_on_sync_kiocb(&iocb
);
582 EXPORT_SYMBOL(sock_sendmsg
);
584 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
585 struct kvec
*vec
, size_t num
, size_t size
)
587 mm_segment_t oldfs
= get_fs();
592 * the following is safe, since for compiler definitions of kvec and
593 * iovec are identical, yielding the same in-core layout and alignment
595 msg
->msg_iov
= (struct iovec
*)vec
;
596 msg
->msg_iovlen
= num
;
597 result
= sock_sendmsg(sock
, msg
, size
);
601 EXPORT_SYMBOL(kernel_sendmsg
);
603 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
606 *ts
= ktime_to_timespec(kt
);
614 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
616 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
619 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
620 struct timespec ts
[3];
622 struct skb_shared_hwtstamps
*shhwtstamps
=
625 /* Race occurred between timestamp enabling and packet
626 receiving. Fill in the current time for now. */
627 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
628 __net_timestamp(skb
);
630 if (need_software_tstamp
) {
631 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
633 skb_get_timestamp(skb
, &tv
);
634 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
637 skb_get_timestampns(skb
, &ts
[0]);
638 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
639 sizeof(ts
[0]), &ts
[0]);
644 memset(ts
, 0, sizeof(ts
));
645 if (skb
->tstamp
.tv64
&&
646 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
647 skb_get_timestampns(skb
, ts
+ 0);
651 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
652 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
654 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
655 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
659 put_cmsg(msg
, SOL_SOCKET
,
660 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
662 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
664 inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
666 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
667 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
668 sizeof(__u32
), &skb
->dropcount
);
671 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
674 sock_recv_timestamp(msg
, sk
, skb
);
675 sock_recv_drops(msg
, sk
, skb
);
677 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
679 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
680 struct msghdr
*msg
, size_t size
, int flags
)
682 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
684 sock_update_classid(sock
->sk
);
692 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
695 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
696 struct msghdr
*msg
, size_t size
, int flags
)
698 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
700 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
703 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
704 size_t size
, int flags
)
707 struct sock_iocb siocb
;
710 init_sync_kiocb(&iocb
, NULL
);
711 iocb
.private = &siocb
;
712 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
713 if (-EIOCBQUEUED
== ret
)
714 ret
= wait_on_sync_kiocb(&iocb
);
717 EXPORT_SYMBOL(sock_recvmsg
);
719 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
720 size_t size
, int flags
)
723 struct sock_iocb siocb
;
726 init_sync_kiocb(&iocb
, NULL
);
727 iocb
.private = &siocb
;
728 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
729 if (-EIOCBQUEUED
== ret
)
730 ret
= wait_on_sync_kiocb(&iocb
);
734 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
735 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
737 mm_segment_t oldfs
= get_fs();
742 * the following is safe, since for compiler definitions of kvec and
743 * iovec are identical, yielding the same in-core layout and alignment
745 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
746 result
= sock_recvmsg(sock
, msg
, size
, flags
);
750 EXPORT_SYMBOL(kernel_recvmsg
);
752 static void sock_aio_dtor(struct kiocb
*iocb
)
754 kfree(iocb
->private);
757 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
758 int offset
, size_t size
, loff_t
*ppos
, int more
)
763 sock
= file
->private_data
;
765 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
769 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
772 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
773 struct pipe_inode_info
*pipe
, size_t len
,
776 struct socket
*sock
= file
->private_data
;
778 if (unlikely(!sock
->ops
->splice_read
))
781 sock_update_classid(sock
->sk
);
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
);
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
);
894 EXPORT_SYMBOL(brioctl_set
);
896 static DEFINE_MUTEX(vlan_ioctl_mutex
);
897 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
899 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
901 mutex_lock(&vlan_ioctl_mutex
);
902 vlan_ioctl_hook
= hook
;
903 mutex_unlock(&vlan_ioctl_mutex
);
905 EXPORT_SYMBOL(vlan_ioctl_set
);
907 static DEFINE_MUTEX(dlci_ioctl_mutex
);
908 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
910 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
912 mutex_lock(&dlci_ioctl_mutex
);
913 dlci_ioctl_hook
= hook
;
914 mutex_unlock(&dlci_ioctl_mutex
);
916 EXPORT_SYMBOL(dlci_ioctl_set
);
918 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
919 unsigned int cmd
, unsigned long arg
)
922 void __user
*argp
= (void __user
*)arg
;
924 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
927 * If this ioctl is unknown try to hand it down
930 if (err
== -ENOIOCTLCMD
)
931 err
= dev_ioctl(net
, cmd
, argp
);
937 * With an ioctl, arg may well be a user mode pointer, but we don't know
938 * what to do with it - that's up to the protocol still.
941 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
945 void __user
*argp
= (void __user
*)arg
;
949 sock
= file
->private_data
;
952 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
953 err
= dev_ioctl(net
, cmd
, argp
);
955 #ifdef CONFIG_WEXT_CORE
956 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
957 err
= dev_ioctl(net
, cmd
, argp
);
964 if (get_user(pid
, (int __user
*)argp
))
966 err
= f_setown(sock
->file
, pid
, 1);
970 err
= put_user(f_getown(sock
->file
),
979 request_module("bridge");
981 mutex_lock(&br_ioctl_mutex
);
983 err
= br_ioctl_hook(net
, cmd
, argp
);
984 mutex_unlock(&br_ioctl_mutex
);
989 if (!vlan_ioctl_hook
)
990 request_module("8021q");
992 mutex_lock(&vlan_ioctl_mutex
);
994 err
= vlan_ioctl_hook(net
, argp
);
995 mutex_unlock(&vlan_ioctl_mutex
);
1000 if (!dlci_ioctl_hook
)
1001 request_module("dlci");
1003 mutex_lock(&dlci_ioctl_mutex
);
1004 if (dlci_ioctl_hook
)
1005 err
= dlci_ioctl_hook(cmd
, argp
);
1006 mutex_unlock(&dlci_ioctl_mutex
);
1009 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1015 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1018 struct socket
*sock
= NULL
;
1020 err
= security_socket_create(family
, type
, protocol
, 1);
1024 sock
= sock_alloc();
1031 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1043 EXPORT_SYMBOL(sock_create_lite
);
1045 /* No kernel lock held - perfect */
1046 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1048 struct socket
*sock
;
1051 * We can't return errors to poll, so it's either yes or no.
1053 sock
= file
->private_data
;
1054 return sock
->ops
->poll(file
, sock
, wait
);
1057 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1059 struct socket
*sock
= file
->private_data
;
1061 return sock
->ops
->mmap(file
, sock
, vma
);
1064 static int sock_close(struct inode
*inode
, struct file
*filp
)
1067 * It was possible the inode is NULL we were
1068 * closing an unfinished socket.
1072 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1075 sock_release(SOCKET_I(inode
));
1080 * Update the socket async list
1082 * Fasync_list locking strategy.
1084 * 1. fasync_list is modified only under process context socket lock
1085 * i.e. under semaphore.
1086 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1087 * or under socket lock
1090 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1092 struct socket
*sock
= filp
->private_data
;
1093 struct sock
*sk
= sock
->sk
;
1100 fasync_helper(fd
, filp
, on
, &sock
->wq
->fasync_list
);
1102 if (!sock
->wq
->fasync_list
)
1103 sock_reset_flag(sk
, SOCK_FASYNC
);
1105 sock_set_flag(sk
, SOCK_FASYNC
);
1111 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1113 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1115 struct socket_wq
*wq
;
1120 wq
= rcu_dereference(sock
->wq
);
1121 if (!wq
|| !wq
->fasync_list
) {
1126 case SOCK_WAKE_WAITD
:
1127 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1130 case SOCK_WAKE_SPACE
:
1131 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1136 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1139 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1144 EXPORT_SYMBOL(sock_wake_async
);
1146 static int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1147 struct socket
**res
, int kern
)
1150 struct socket
*sock
;
1151 const struct net_proto_family
*pf
;
1154 * Check protocol is in range
1156 if (family
< 0 || family
>= NPROTO
)
1157 return -EAFNOSUPPORT
;
1158 if (type
< 0 || type
>= SOCK_MAX
)
1163 This uglymoron is moved from INET layer to here to avoid
1164 deadlock in module load.
1166 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1170 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1176 err
= security_socket_create(family
, type
, protocol
, kern
);
1181 * Allocate the socket and allow the family to set things up. if
1182 * the protocol is 0, the family is instructed to select an appropriate
1185 sock
= sock_alloc();
1187 if (net_ratelimit())
1188 printk(KERN_WARNING
"socket: no more sockets\n");
1189 return -ENFILE
; /* Not exactly a match, but its the
1190 closest posix thing */
1195 #ifdef CONFIG_MODULES
1196 /* Attempt to load a protocol module if the find failed.
1198 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1199 * requested real, full-featured networking support upon configuration.
1200 * Otherwise module support will break!
1202 if (net_families
[family
] == NULL
)
1203 request_module("net-pf-%d", family
);
1207 pf
= rcu_dereference(net_families
[family
]);
1208 err
= -EAFNOSUPPORT
;
1213 * We will call the ->create function, that possibly is in a loadable
1214 * module, so we have to bump that loadable module refcnt first.
1216 if (!try_module_get(pf
->owner
))
1219 /* Now protected by module ref count */
1222 err
= pf
->create(net
, sock
, protocol
, kern
);
1224 goto out_module_put
;
1227 * Now to bump the refcnt of the [loadable] module that owns this
1228 * socket at sock_release time we decrement its refcnt.
1230 if (!try_module_get(sock
->ops
->owner
))
1231 goto out_module_busy
;
1234 * Now that we're done with the ->create function, the [loadable]
1235 * module can have its refcnt decremented
1237 module_put(pf
->owner
);
1238 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1240 goto out_sock_release
;
1246 err
= -EAFNOSUPPORT
;
1249 module_put(pf
->owner
);
1256 goto out_sock_release
;
1259 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1261 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1263 EXPORT_SYMBOL(sock_create
);
1265 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1267 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1269 EXPORT_SYMBOL(sock_create_kern
);
1271 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1274 struct socket
*sock
;
1277 /* Check the SOCK_* constants for consistency. */
1278 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1279 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1280 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1281 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1283 flags
= type
& ~SOCK_TYPE_MASK
;
1284 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1286 type
&= SOCK_TYPE_MASK
;
1288 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1289 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1291 retval
= sock_create(family
, type
, protocol
, &sock
);
1295 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1300 /* It may be already another descriptor 8) Not kernel problem. */
1309 * Create a pair of connected sockets.
1312 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1313 int __user
*, usockvec
)
1315 struct socket
*sock1
, *sock2
;
1317 struct file
*newfile1
, *newfile2
;
1320 flags
= type
& ~SOCK_TYPE_MASK
;
1321 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1323 type
&= SOCK_TYPE_MASK
;
1325 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1326 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1329 * Obtain the first socket and check if the underlying protocol
1330 * supports the socketpair call.
1333 err
= sock_create(family
, type
, protocol
, &sock1
);
1337 err
= sock_create(family
, type
, protocol
, &sock2
);
1341 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1343 goto out_release_both
;
1345 fd1
= sock_alloc_file(sock1
, &newfile1
, flags
);
1346 if (unlikely(fd1
< 0)) {
1348 goto out_release_both
;
1351 fd2
= sock_alloc_file(sock2
, &newfile2
, flags
);
1352 if (unlikely(fd2
< 0)) {
1356 sock_release(sock2
);
1360 audit_fd_pair(fd1
, fd2
);
1361 fd_install(fd1
, newfile1
);
1362 fd_install(fd2
, newfile2
);
1363 /* fd1 and fd2 may be already another descriptors.
1364 * Not kernel problem.
1367 err
= put_user(fd1
, &usockvec
[0]);
1369 err
= put_user(fd2
, &usockvec
[1]);
1378 sock_release(sock2
);
1380 sock_release(sock1
);
1386 * Bind a name to a socket. Nothing much to do here since it's
1387 * the protocol's responsibility to handle the local address.
1389 * We move the socket address to kernel space before we call
1390 * the protocol layer (having also checked the address is ok).
1393 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1395 struct socket
*sock
;
1396 struct sockaddr_storage address
;
1397 int err
, fput_needed
;
1399 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1401 err
= move_addr_to_kernel(umyaddr
, addrlen
, (struct sockaddr
*)&address
);
1403 err
= security_socket_bind(sock
,
1404 (struct sockaddr
*)&address
,
1407 err
= sock
->ops
->bind(sock
,
1411 fput_light(sock
->file
, fput_needed
);
1417 * Perform a listen. Basically, we allow the protocol to do anything
1418 * necessary for a listen, and if that works, we mark the socket as
1419 * ready for listening.
1422 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1424 struct socket
*sock
;
1425 int err
, fput_needed
;
1428 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1430 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1431 if ((unsigned)backlog
> somaxconn
)
1432 backlog
= somaxconn
;
1434 err
= security_socket_listen(sock
, backlog
);
1436 err
= sock
->ops
->listen(sock
, backlog
);
1438 fput_light(sock
->file
, fput_needed
);
1444 * For accept, we attempt to create a new socket, set up the link
1445 * with the client, wake up the client, then return the new
1446 * connected fd. We collect the address of the connector in kernel
1447 * space and move it to user at the very end. This is unclean because
1448 * we open the socket then return an error.
1450 * 1003.1g adds the ability to recvmsg() to query connection pending
1451 * status to recvmsg. We need to add that support in a way thats
1452 * clean when we restucture accept also.
1455 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1456 int __user
*, upeer_addrlen
, int, flags
)
1458 struct socket
*sock
, *newsock
;
1459 struct file
*newfile
;
1460 int err
, len
, newfd
, fput_needed
;
1461 struct sockaddr_storage address
;
1463 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1466 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1467 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1469 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1474 newsock
= sock_alloc();
1478 newsock
->type
= sock
->type
;
1479 newsock
->ops
= sock
->ops
;
1482 * We don't need try_module_get here, as the listening socket (sock)
1483 * has the protocol module (sock->ops->owner) held.
1485 __module_get(newsock
->ops
->owner
);
1487 newfd
= sock_alloc_file(newsock
, &newfile
, flags
);
1488 if (unlikely(newfd
< 0)) {
1490 sock_release(newsock
);
1494 err
= security_socket_accept(sock
, newsock
);
1498 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1502 if (upeer_sockaddr
) {
1503 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1505 err
= -ECONNABORTED
;
1508 err
= move_addr_to_user((struct sockaddr
*)&address
,
1509 len
, upeer_sockaddr
, upeer_addrlen
);
1514 /* File flags are not inherited via accept() unlike another OSes. */
1516 fd_install(newfd
, newfile
);
1520 fput_light(sock
->file
, fput_needed
);
1525 put_unused_fd(newfd
);
1529 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1530 int __user
*, upeer_addrlen
)
1532 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1536 * Attempt to connect to a socket with the server address. The address
1537 * is in user space so we verify it is OK and move it to kernel space.
1539 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1542 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1543 * other SEQPACKET protocols that take time to connect() as it doesn't
1544 * include the -EINPROGRESS status for such sockets.
1547 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1550 struct socket
*sock
;
1551 struct sockaddr_storage address
;
1552 int err
, fput_needed
;
1554 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1557 err
= move_addr_to_kernel(uservaddr
, addrlen
, (struct sockaddr
*)&address
);
1562 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1566 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1567 sock
->file
->f_flags
);
1569 fput_light(sock
->file
, fput_needed
);
1575 * Get the local address ('name') of a socket object. Move the obtained
1576 * name to user space.
1579 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1580 int __user
*, usockaddr_len
)
1582 struct socket
*sock
;
1583 struct sockaddr_storage address
;
1584 int len
, err
, fput_needed
;
1586 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1590 err
= security_socket_getsockname(sock
);
1594 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1597 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
, usockaddr_len
);
1600 fput_light(sock
->file
, fput_needed
);
1606 * Get the remote address ('name') of a socket object. Move the obtained
1607 * name to user space.
1610 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1611 int __user
*, usockaddr_len
)
1613 struct socket
*sock
;
1614 struct sockaddr_storage address
;
1615 int len
, err
, fput_needed
;
1617 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1619 err
= security_socket_getpeername(sock
);
1621 fput_light(sock
->file
, fput_needed
);
1626 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1629 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
,
1631 fput_light(sock
->file
, fput_needed
);
1637 * Send a datagram to a given address. We move the address into kernel
1638 * space and check the user space data area is readable before invoking
1642 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1643 unsigned, flags
, struct sockaddr __user
*, addr
,
1646 struct socket
*sock
;
1647 struct sockaddr_storage address
;
1653 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1657 iov
.iov_base
= buff
;
1659 msg
.msg_name
= NULL
;
1662 msg
.msg_control
= NULL
;
1663 msg
.msg_controllen
= 0;
1664 msg
.msg_namelen
= 0;
1666 err
= move_addr_to_kernel(addr
, addr_len
, (struct sockaddr
*)&address
);
1669 msg
.msg_name
= (struct sockaddr
*)&address
;
1670 msg
.msg_namelen
= addr_len
;
1672 if (sock
->file
->f_flags
& O_NONBLOCK
)
1673 flags
|= MSG_DONTWAIT
;
1674 msg
.msg_flags
= flags
;
1675 err
= sock_sendmsg(sock
, &msg
, len
);
1678 fput_light(sock
->file
, fput_needed
);
1684 * Send a datagram down a socket.
1687 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1690 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1694 * Receive a frame from the socket and optionally record the address of the
1695 * sender. We verify the buffers are writable and if needed move the
1696 * sender address from kernel to user space.
1699 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1700 unsigned, flags
, struct sockaddr __user
*, addr
,
1701 int __user
*, addr_len
)
1703 struct socket
*sock
;
1706 struct sockaddr_storage address
;
1710 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1714 msg
.msg_control
= NULL
;
1715 msg
.msg_controllen
= 0;
1719 iov
.iov_base
= ubuf
;
1720 msg
.msg_name
= (struct sockaddr
*)&address
;
1721 msg
.msg_namelen
= sizeof(address
);
1722 if (sock
->file
->f_flags
& O_NONBLOCK
)
1723 flags
|= MSG_DONTWAIT
;
1724 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1726 if (err
>= 0 && addr
!= NULL
) {
1727 err2
= move_addr_to_user((struct sockaddr
*)&address
,
1728 msg
.msg_namelen
, addr
, addr_len
);
1733 fput_light(sock
->file
, fput_needed
);
1739 * Receive a datagram from a socket.
1742 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1745 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1749 * Set a socket option. Because we don't know the option lengths we have
1750 * to pass the user mode parameter for the protocols to sort out.
1753 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1754 char __user
*, optval
, int, optlen
)
1756 int err
, fput_needed
;
1757 struct socket
*sock
;
1762 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1764 err
= security_socket_setsockopt(sock
, level
, optname
);
1768 if (level
== SOL_SOCKET
)
1770 sock_setsockopt(sock
, level
, optname
, optval
,
1774 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1777 fput_light(sock
->file
, fput_needed
);
1783 * Get a socket option. Because we don't know the option lengths we have
1784 * to pass a user mode parameter for the protocols to sort out.
1787 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1788 char __user
*, optval
, int __user
*, optlen
)
1790 int err
, fput_needed
;
1791 struct socket
*sock
;
1793 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1795 err
= security_socket_getsockopt(sock
, level
, optname
);
1799 if (level
== SOL_SOCKET
)
1801 sock_getsockopt(sock
, level
, optname
, optval
,
1805 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1808 fput_light(sock
->file
, fput_needed
);
1814 * Shutdown a socket.
1817 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1819 int err
, fput_needed
;
1820 struct socket
*sock
;
1822 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1824 err
= security_socket_shutdown(sock
, how
);
1826 err
= sock
->ops
->shutdown(sock
, how
);
1827 fput_light(sock
->file
, fput_needed
);
1832 /* A couple of helpful macros for getting the address of the 32/64 bit
1833 * fields which are the same type (int / unsigned) on our platforms.
1835 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1836 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1837 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1840 * BSD sendmsg interface
1843 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned, flags
)
1845 struct compat_msghdr __user
*msg_compat
=
1846 (struct compat_msghdr __user
*)msg
;
1847 struct socket
*sock
;
1848 struct sockaddr_storage address
;
1849 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1850 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1851 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1852 /* 20 is size of ipv6_pktinfo */
1853 unsigned char *ctl_buf
= ctl
;
1854 struct msghdr msg_sys
;
1855 int err
, ctl_len
, iov_size
, total_len
;
1859 if (MSG_CMSG_COMPAT
& flags
) {
1860 if (get_compat_msghdr(&msg_sys
, msg_compat
))
1862 } else if (copy_from_user(&msg_sys
, msg
, sizeof(struct msghdr
)))
1865 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1869 /* do not move before msg_sys is valid */
1871 if (msg_sys
.msg_iovlen
> UIO_MAXIOV
)
1874 /* Check whether to allocate the iovec area */
1876 iov_size
= msg_sys
.msg_iovlen
* sizeof(struct iovec
);
1877 if (msg_sys
.msg_iovlen
> UIO_FASTIOV
) {
1878 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1883 /* This will also move the address data into kernel space */
1884 if (MSG_CMSG_COMPAT
& flags
) {
1885 err
= verify_compat_iovec(&msg_sys
, iov
,
1886 (struct sockaddr
*)&address
,
1889 err
= verify_iovec(&msg_sys
, iov
,
1890 (struct sockaddr
*)&address
,
1898 if (msg_sys
.msg_controllen
> INT_MAX
)
1900 ctl_len
= msg_sys
.msg_controllen
;
1901 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1903 cmsghdr_from_user_compat_to_kern(&msg_sys
, sock
->sk
, ctl
,
1907 ctl_buf
= msg_sys
.msg_control
;
1908 ctl_len
= msg_sys
.msg_controllen
;
1909 } else if (ctl_len
) {
1910 if (ctl_len
> sizeof(ctl
)) {
1911 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1912 if (ctl_buf
== NULL
)
1917 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1918 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1919 * checking falls down on this.
1921 if (copy_from_user(ctl_buf
, (void __user
*)msg_sys
.msg_control
,
1924 msg_sys
.msg_control
= ctl_buf
;
1926 msg_sys
.msg_flags
= flags
;
1928 if (sock
->file
->f_flags
& O_NONBLOCK
)
1929 msg_sys
.msg_flags
|= MSG_DONTWAIT
;
1930 err
= sock_sendmsg(sock
, &msg_sys
, total_len
);
1934 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1936 if (iov
!= iovstack
)
1937 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1939 fput_light(sock
->file
, fput_needed
);
1944 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1945 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
1947 struct compat_msghdr __user
*msg_compat
=
1948 (struct compat_msghdr __user
*)msg
;
1949 struct iovec iovstack
[UIO_FASTIOV
];
1950 struct iovec
*iov
= iovstack
;
1951 unsigned long cmsg_ptr
;
1952 int err
, iov_size
, total_len
, len
;
1954 /* kernel mode address */
1955 struct sockaddr_storage addr
;
1957 /* user mode address pointers */
1958 struct sockaddr __user
*uaddr
;
1959 int __user
*uaddr_len
;
1961 if (MSG_CMSG_COMPAT
& flags
) {
1962 if (get_compat_msghdr(msg_sys
, msg_compat
))
1964 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
1968 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
1971 /* Check whether to allocate the iovec area */
1973 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
1974 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
1975 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1981 * Save the user-mode address (verify_iovec will change the
1982 * kernel msghdr to use the kernel address space)
1985 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
1986 uaddr_len
= COMPAT_NAMELEN(msg
);
1987 if (MSG_CMSG_COMPAT
& flags
) {
1988 err
= verify_compat_iovec(msg_sys
, iov
,
1989 (struct sockaddr
*)&addr
,
1992 err
= verify_iovec(msg_sys
, iov
,
1993 (struct sockaddr
*)&addr
,
1999 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2000 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2002 if (sock
->file
->f_flags
& O_NONBLOCK
)
2003 flags
|= MSG_DONTWAIT
;
2004 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2010 if (uaddr
!= NULL
) {
2011 err
= move_addr_to_user((struct sockaddr
*)&addr
,
2012 msg_sys
->msg_namelen
, uaddr
,
2017 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2021 if (MSG_CMSG_COMPAT
& flags
)
2022 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2023 &msg_compat
->msg_controllen
);
2025 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2026 &msg
->msg_controllen
);
2032 if (iov
!= iovstack
)
2033 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2039 * BSD recvmsg interface
2042 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2043 unsigned int, flags
)
2045 int fput_needed
, err
;
2046 struct msghdr msg_sys
;
2047 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2052 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2054 fput_light(sock
->file
, fput_needed
);
2060 * Linux recvmmsg interface
2063 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2064 unsigned int flags
, struct timespec
*timeout
)
2066 int fput_needed
, err
, datagrams
;
2067 struct socket
*sock
;
2068 struct mmsghdr __user
*entry
;
2069 struct compat_mmsghdr __user
*compat_entry
;
2070 struct msghdr msg_sys
;
2071 struct timespec end_time
;
2074 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2080 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2084 err
= sock_error(sock
->sk
);
2089 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2091 while (datagrams
< vlen
) {
2093 * No need to ask LSM for more than the first datagram.
2095 if (MSG_CMSG_COMPAT
& flags
) {
2096 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2097 &msg_sys
, flags
, datagrams
);
2100 err
= __put_user(err
, &compat_entry
->msg_len
);
2103 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2104 &msg_sys
, flags
, datagrams
);
2107 err
= put_user(err
, &entry
->msg_len
);
2115 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2116 if (flags
& MSG_WAITFORONE
)
2117 flags
|= MSG_DONTWAIT
;
2120 ktime_get_ts(timeout
);
2121 *timeout
= timespec_sub(end_time
, *timeout
);
2122 if (timeout
->tv_sec
< 0) {
2123 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2127 /* Timeout, return less than vlen datagrams */
2128 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2132 /* Out of band data, return right away */
2133 if (msg_sys
.msg_flags
& MSG_OOB
)
2138 fput_light(sock
->file
, fput_needed
);
2143 if (datagrams
!= 0) {
2145 * We may return less entries than requested (vlen) if the
2146 * sock is non block and there aren't enough datagrams...
2148 if (err
!= -EAGAIN
) {
2150 * ... or if recvmsg returns an error after we
2151 * received some datagrams, where we record the
2152 * error to return on the next call or if the
2153 * app asks about it using getsockopt(SO_ERROR).
2155 sock
->sk
->sk_err
= -err
;
2164 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2165 unsigned int, vlen
, unsigned int, flags
,
2166 struct timespec __user
*, timeout
)
2169 struct timespec timeout_sys
;
2172 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2174 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2177 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2179 if (datagrams
> 0 &&
2180 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2181 datagrams
= -EFAULT
;
2186 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2187 /* Argument list sizes for sys_socketcall */
2188 #define AL(x) ((x) * sizeof(unsigned long))
2189 static const unsigned char nargs
[20] = {
2190 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2191 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2192 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2199 * System call vectors.
2201 * Argument checking cleaned up. Saved 20% in size.
2202 * This function doesn't need to set the kernel lock because
2203 * it is set by the callees.
2206 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2209 unsigned long a0
, a1
;
2213 if (call
< 1 || call
> SYS_RECVMMSG
)
2217 if (len
> sizeof(a
))
2220 /* copy_from_user should be SMP safe. */
2221 if (copy_from_user(a
, args
, len
))
2224 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2231 err
= sys_socket(a0
, a1
, a
[2]);
2234 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2237 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2240 err
= sys_listen(a0
, a1
);
2243 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2244 (int __user
*)a
[2], 0);
2246 case SYS_GETSOCKNAME
:
2248 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2249 (int __user
*)a
[2]);
2251 case SYS_GETPEERNAME
:
2253 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2254 (int __user
*)a
[2]);
2256 case SYS_SOCKETPAIR
:
2257 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2260 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2263 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2264 (struct sockaddr __user
*)a
[4], a
[5]);
2267 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2270 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2271 (struct sockaddr __user
*)a
[4],
2272 (int __user
*)a
[5]);
2275 err
= sys_shutdown(a0
, a1
);
2277 case SYS_SETSOCKOPT
:
2278 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2280 case SYS_GETSOCKOPT
:
2282 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2283 (int __user
*)a
[4]);
2286 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2289 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2292 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2293 (struct timespec __user
*)a
[4]);
2296 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2297 (int __user
*)a
[2], a
[3]);
2306 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2309 * sock_register - add a socket protocol handler
2310 * @ops: description of protocol
2312 * This function is called by a protocol handler that wants to
2313 * advertise its address family, and have it linked into the
2314 * socket interface. The value ops->family coresponds to the
2315 * socket system call protocol family.
2317 int sock_register(const struct net_proto_family
*ops
)
2321 if (ops
->family
>= NPROTO
) {
2322 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2327 spin_lock(&net_family_lock
);
2328 if (net_families
[ops
->family
])
2331 net_families
[ops
->family
] = ops
;
2334 spin_unlock(&net_family_lock
);
2336 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2339 EXPORT_SYMBOL(sock_register
);
2342 * sock_unregister - remove a protocol handler
2343 * @family: protocol family to remove
2345 * This function is called by a protocol handler that wants to
2346 * remove its address family, and have it unlinked from the
2347 * new socket creation.
2349 * If protocol handler is a module, then it can use module reference
2350 * counts to protect against new references. If protocol handler is not
2351 * a module then it needs to provide its own protection in
2352 * the ops->create routine.
2354 void sock_unregister(int family
)
2356 BUG_ON(family
< 0 || family
>= NPROTO
);
2358 spin_lock(&net_family_lock
);
2359 net_families
[family
] = NULL
;
2360 spin_unlock(&net_family_lock
);
2364 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2366 EXPORT_SYMBOL(sock_unregister
);
2368 static int __init
sock_init(void)
2371 * Initialize sock SLAB cache.
2377 * Initialize skbuff SLAB cache
2382 * Initialize the protocols module.
2386 register_filesystem(&sock_fs_type
);
2387 sock_mnt
= kern_mount(&sock_fs_type
);
2389 /* The real protocol initialization is performed in later initcalls.
2392 #ifdef CONFIG_NETFILTER
2396 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2397 skb_timestamping_init();
2403 core_initcall(sock_init
); /* early initcall */
2405 #ifdef CONFIG_PROC_FS
2406 void socket_seq_show(struct seq_file
*seq
)
2411 for_each_possible_cpu(cpu
)
2412 counter
+= per_cpu(sockets_in_use
, cpu
);
2414 /* It can be negative, by the way. 8) */
2418 seq_printf(seq
, "sockets: used %d\n", counter
);
2420 #endif /* CONFIG_PROC_FS */
2422 #ifdef CONFIG_COMPAT
2423 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2424 unsigned int cmd
, struct compat_timeval __user
*up
)
2426 mm_segment_t old_fs
= get_fs();
2431 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2434 err
= put_user(ktv
.tv_sec
, &up
->tv_sec
);
2435 err
|= __put_user(ktv
.tv_usec
, &up
->tv_usec
);
2440 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2441 unsigned int cmd
, struct compat_timespec __user
*up
)
2443 mm_segment_t old_fs
= get_fs();
2444 struct timespec kts
;
2448 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2451 err
= put_user(kts
.tv_sec
, &up
->tv_sec
);
2452 err
|= __put_user(kts
.tv_nsec
, &up
->tv_nsec
);
2457 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2459 struct ifreq __user
*uifr
;
2462 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2463 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2466 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2470 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2476 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2478 struct compat_ifconf ifc32
;
2480 struct ifconf __user
*uifc
;
2481 struct compat_ifreq __user
*ifr32
;
2482 struct ifreq __user
*ifr
;
2486 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2489 if (ifc32
.ifcbuf
== 0) {
2493 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2495 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2496 sizeof(struct ifreq
);
2497 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2499 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2500 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2501 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2502 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2508 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2511 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2515 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2519 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2521 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2522 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2523 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2529 if (ifc32
.ifcbuf
== 0) {
2530 /* Translate from 64-bit structure multiple to
2534 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2539 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2545 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2547 struct ifreq __user
*ifr
;
2551 ifr
= compat_alloc_user_space(sizeof(*ifr
));
2553 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2556 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2559 datap
= compat_ptr(data
);
2560 if (put_user(datap
, &ifr
->ifr_ifru
.ifru_data
))
2563 return dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2566 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2569 compat_uptr_t uptr32
;
2570 struct ifreq __user
*uifr
;
2572 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2573 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2576 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2579 uptr
= compat_ptr(uptr32
);
2581 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2584 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2587 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2588 struct compat_ifreq __user
*ifr32
)
2591 struct ifreq __user
*uifr
;
2592 mm_segment_t old_fs
;
2598 case SIOCBONDENSLAVE
:
2599 case SIOCBONDRELEASE
:
2600 case SIOCBONDSETHWADDR
:
2601 case SIOCBONDCHANGEACTIVE
:
2602 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2607 err
= dev_ioctl(net
, cmd
, &kifr
);
2611 case SIOCBONDSLAVEINFOQUERY
:
2612 case SIOCBONDINFOQUERY
:
2613 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2614 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2617 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2620 datap
= compat_ptr(data
);
2621 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2624 return dev_ioctl(net
, cmd
, uifr
);
2630 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2631 struct compat_ifreq __user
*u_ifreq32
)
2633 struct ifreq __user
*u_ifreq64
;
2634 char tmp_buf
[IFNAMSIZ
];
2635 void __user
*data64
;
2638 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2641 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2643 data64
= compat_ptr(data32
);
2645 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2647 /* Don't check these user accesses, just let that get trapped
2648 * in the ioctl handler instead.
2650 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2653 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2656 return dev_ioctl(net
, cmd
, u_ifreq64
);
2659 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2660 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2662 struct ifreq __user
*uifr
;
2665 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2666 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2669 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2680 case SIOCGIFBRDADDR
:
2681 case SIOCGIFDSTADDR
:
2682 case SIOCGIFNETMASK
:
2687 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2695 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2696 struct compat_ifreq __user
*uifr32
)
2699 struct compat_ifmap __user
*uifmap32
;
2700 mm_segment_t old_fs
;
2703 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2704 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2705 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2706 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2707 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2708 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2709 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2710 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2716 err
= dev_ioctl(net
, cmd
, (void __user
*)&ifr
);
2719 if (cmd
== SIOCGIFMAP
&& !err
) {
2720 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2721 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2722 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2723 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2724 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2725 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2726 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2733 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2736 compat_uptr_t uptr32
;
2737 struct ifreq __user
*uifr
;
2739 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2740 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2743 if (get_user(uptr32
, &uifr32
->ifr_data
))
2746 uptr
= compat_ptr(uptr32
);
2748 if (put_user(uptr
, &uifr
->ifr_data
))
2751 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
2756 struct sockaddr rt_dst
; /* target address */
2757 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2758 struct sockaddr rt_genmask
; /* target network mask (IP) */
2759 unsigned short rt_flags
;
2762 unsigned char rt_tos
;
2763 unsigned char rt_class
;
2765 short rt_metric
; /* +1 for binary compatibility! */
2766 /* char * */ u32 rt_dev
; /* forcing the device at add */
2767 u32 rt_mtu
; /* per route MTU/Window */
2768 u32 rt_window
; /* Window clamping */
2769 unsigned short rt_irtt
; /* Initial RTT */
2772 struct in6_rtmsg32
{
2773 struct in6_addr rtmsg_dst
;
2774 struct in6_addr rtmsg_src
;
2775 struct in6_addr rtmsg_gateway
;
2785 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
2786 unsigned int cmd
, void __user
*argp
)
2790 struct in6_rtmsg r6
;
2794 mm_segment_t old_fs
= get_fs();
2796 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
2797 struct in6_rtmsg32 __user
*ur6
= argp
;
2798 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
2799 3 * sizeof(struct in6_addr
));
2800 ret
|= __get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
2801 ret
|= __get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
2802 ret
|= __get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
2803 ret
|= __get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
2804 ret
|= __get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
2805 ret
|= __get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
2806 ret
|= __get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
2810 struct rtentry32 __user
*ur4
= argp
;
2811 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
2812 3 * sizeof(struct sockaddr
));
2813 ret
|= __get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
2814 ret
|= __get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
2815 ret
|= __get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
2816 ret
|= __get_user(r4
.rt_window
, &(ur4
->rt_window
));
2817 ret
|= __get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
2818 ret
|= __get_user(rtdev
, &(ur4
->rt_dev
));
2820 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
2821 r4
.rt_dev
= devname
; devname
[15] = 0;
2834 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
2841 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2842 * for some operations; this forces use of the newer bridge-utils that
2843 * use compatiable ioctls
2845 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
2849 if (get_user(tmp
, argp
))
2851 if (tmp
== BRCTL_GET_VERSION
)
2852 return BRCTL_VERSION
+ 1;
2856 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
2857 unsigned int cmd
, unsigned long arg
)
2859 void __user
*argp
= compat_ptr(arg
);
2860 struct sock
*sk
= sock
->sk
;
2861 struct net
*net
= sock_net(sk
);
2863 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
2864 return siocdevprivate_ioctl(net
, cmd
, argp
);
2869 return old_bridge_ioctl(argp
);
2871 return dev_ifname32(net
, argp
);
2873 return dev_ifconf(net
, argp
);
2875 return ethtool_ioctl(net
, argp
);
2877 return compat_siocwandev(net
, argp
);
2880 return compat_sioc_ifmap(net
, cmd
, argp
);
2881 case SIOCBONDENSLAVE
:
2882 case SIOCBONDRELEASE
:
2883 case SIOCBONDSETHWADDR
:
2884 case SIOCBONDSLAVEINFOQUERY
:
2885 case SIOCBONDINFOQUERY
:
2886 case SIOCBONDCHANGEACTIVE
:
2887 return bond_ioctl(net
, cmd
, argp
);
2890 return routing_ioctl(net
, sock
, cmd
, argp
);
2892 return do_siocgstamp(net
, sock
, cmd
, argp
);
2894 return do_siocgstampns(net
, sock
, cmd
, argp
);
2896 return compat_siocshwtstamp(net
, argp
);
2908 return sock_ioctl(file
, cmd
, arg
);
2925 case SIOCSIFHWBROADCAST
:
2927 case SIOCGIFBRDADDR
:
2928 case SIOCSIFBRDADDR
:
2929 case SIOCGIFDSTADDR
:
2930 case SIOCSIFDSTADDR
:
2931 case SIOCGIFNETMASK
:
2932 case SIOCSIFNETMASK
:
2943 return dev_ifsioc(net
, sock
, cmd
, argp
);
2949 return sock_do_ioctl(net
, sock
, cmd
, arg
);
2952 /* Prevent warning from compat_sys_ioctl, these always
2953 * result in -EINVAL in the native case anyway. */
2966 return -ENOIOCTLCMD
;
2969 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
2972 struct socket
*sock
= file
->private_data
;
2973 int ret
= -ENOIOCTLCMD
;
2980 if (sock
->ops
->compat_ioctl
)
2981 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
2983 if (ret
== -ENOIOCTLCMD
&&
2984 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
2985 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
2987 if (ret
== -ENOIOCTLCMD
)
2988 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
2994 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
2996 return sock
->ops
->bind(sock
, addr
, addrlen
);
2998 EXPORT_SYMBOL(kernel_bind
);
3000 int kernel_listen(struct socket
*sock
, int backlog
)
3002 return sock
->ops
->listen(sock
, backlog
);
3004 EXPORT_SYMBOL(kernel_listen
);
3006 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3008 struct sock
*sk
= sock
->sk
;
3011 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3016 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3018 sock_release(*newsock
);
3023 (*newsock
)->ops
= sock
->ops
;
3024 __module_get((*newsock
)->ops
->owner
);
3029 EXPORT_SYMBOL(kernel_accept
);
3031 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3034 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3036 EXPORT_SYMBOL(kernel_connect
);
3038 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3041 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3043 EXPORT_SYMBOL(kernel_getsockname
);
3045 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3048 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3050 EXPORT_SYMBOL(kernel_getpeername
);
3052 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3053 char *optval
, int *optlen
)
3055 mm_segment_t oldfs
= get_fs();
3059 if (level
== SOL_SOCKET
)
3060 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
3062 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
3067 EXPORT_SYMBOL(kernel_getsockopt
);
3069 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3070 char *optval
, unsigned int optlen
)
3072 mm_segment_t oldfs
= get_fs();
3076 if (level
== SOL_SOCKET
)
3077 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
3079 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
3084 EXPORT_SYMBOL(kernel_setsockopt
);
3086 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3087 size_t size
, int flags
)
3089 sock_update_classid(sock
->sk
);
3091 if (sock
->ops
->sendpage
)
3092 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3094 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3096 EXPORT_SYMBOL(kernel_sendpage
);
3098 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3100 mm_segment_t oldfs
= get_fs();
3104 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3109 EXPORT_SYMBOL(kernel_sock_ioctl
);
3111 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3113 return sock
->ops
->shutdown(sock
, how
);
3115 EXPORT_SYMBOL(kernel_sock_shutdown
);