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 __rcu
*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 static int move_addr_to_user(struct sockaddr
*kaddr
, int klen
,
213 void __user
*uaddr
, int __user
*ulen
)
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
;
243 struct socket_wq
*wq
;
245 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
248 wq
= kmalloc(sizeof(*wq
), GFP_KERNEL
);
250 kmem_cache_free(sock_inode_cachep
, ei
);
253 init_waitqueue_head(&wq
->wait
);
254 wq
->fasync_list
= NULL
;
255 RCU_INIT_POINTER(ei
->socket
.wq
, wq
);
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
;
268 static void wq_free_rcu(struct rcu_head
*head
)
270 struct socket_wq
*wq
= container_of(head
, struct socket_wq
, rcu
);
275 static void sock_destroy_inode(struct inode
*inode
)
277 struct socket_alloc
*ei
;
278 struct socket_wq
*wq
;
280 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
281 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
282 call_rcu(&wq
->rcu
, wq_free_rcu
);
283 kmem_cache_free(sock_inode_cachep
, ei
);
286 static void init_once(void *foo
)
288 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
290 inode_init_once(&ei
->vfs_inode
);
293 static int init_inodecache(void)
295 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
296 sizeof(struct socket_alloc
),
298 (SLAB_HWCACHE_ALIGN
|
299 SLAB_RECLAIM_ACCOUNT
|
302 if (sock_inode_cachep
== NULL
)
307 static const struct super_operations sockfs_ops
= {
308 .alloc_inode
= sock_alloc_inode
,
309 .destroy_inode
= sock_destroy_inode
,
310 .statfs
= simple_statfs
,
314 * sockfs_dname() is called from d_path().
316 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
318 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
319 dentry
->d_inode
->i_ino
);
322 static const struct dentry_operations sockfs_dentry_operations
= {
323 .d_dname
= sockfs_dname
,
326 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
327 int flags
, const char *dev_name
, void *data
)
329 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
330 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
333 static struct vfsmount
*sock_mnt __read_mostly
;
335 static struct file_system_type sock_fs_type
= {
337 .mount
= sockfs_mount
,
338 .kill_sb
= kill_anon_super
,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 static int sock_alloc_file(struct socket
*sock
, struct file
**f
, int flags
)
360 struct qstr name
= { .name
= "" };
365 fd
= get_unused_fd_flags(flags
);
366 if (unlikely(fd
< 0))
369 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
370 if (unlikely(!path
.dentry
)) {
374 path
.mnt
= mntget(sock_mnt
);
376 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
377 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
379 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
381 if (unlikely(!file
)) {
382 /* drop dentry, keep inode */
383 ihold(path
.dentry
->d_inode
);
390 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
392 file
->private_data
= sock
;
398 int sock_map_fd(struct socket
*sock
, int flags
)
400 struct file
*newfile
;
401 int fd
= sock_alloc_file(sock
, &newfile
, flags
);
404 fd_install(fd
, newfile
);
408 EXPORT_SYMBOL(sock_map_fd
);
410 static struct socket
*sock_from_file(struct file
*file
, int *err
)
412 if (file
->f_op
== &socket_file_ops
)
413 return file
->private_data
; /* set in sock_map_fd */
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket
*sockfd_lookup(int fd
, int *err
)
443 sock
= sock_from_file(file
, err
);
448 EXPORT_SYMBOL(sockfd_lookup
);
450 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
456 file
= fget_light(fd
, fput_needed
);
458 sock
= sock_from_file(file
, err
);
461 fput_light(file
, *fput_needed
);
467 * sock_alloc - allocate a socket
469 * Allocate a new inode and socket object. The two are bound together
470 * and initialised. The socket is then returned. If we are out of inodes
474 static struct socket
*sock_alloc(void)
479 inode
= new_inode(sock_mnt
->mnt_sb
);
483 sock
= SOCKET_I(inode
);
485 kmemcheck_annotate_bitfield(sock
, type
);
486 inode
->i_ino
= get_next_ino();
487 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
488 inode
->i_uid
= current_fsuid();
489 inode
->i_gid
= current_fsgid();
491 percpu_add(sockets_in_use
, 1);
496 * In theory you can't get an open on this inode, but /proc provides
497 * a back door. Remember to keep it shut otherwise you'll let the
498 * creepy crawlies in.
501 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
506 const struct file_operations bad_sock_fops
= {
507 .owner
= THIS_MODULE
,
508 .open
= sock_no_open
,
509 .llseek
= noop_llseek
,
513 * sock_release - close a socket
514 * @sock: socket to close
516 * The socket is released from the protocol stack if it has a release
517 * callback, and the inode is then released if the socket is bound to
518 * an inode not a file.
521 void sock_release(struct socket
*sock
)
524 struct module
*owner
= sock
->ops
->owner
;
526 sock
->ops
->release(sock
);
531 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
532 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
534 percpu_sub(sockets_in_use
, 1);
536 iput(SOCK_INODE(sock
));
541 EXPORT_SYMBOL(sock_release
);
543 int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
)
546 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
547 *tx_flags
|= SKBTX_HW_TSTAMP
;
548 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
549 *tx_flags
|= SKBTX_SW_TSTAMP
;
552 EXPORT_SYMBOL(sock_tx_timestamp
);
554 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
555 struct msghdr
*msg
, size_t size
)
557 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
559 sock_update_classid(sock
->sk
);
566 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
569 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
570 struct msghdr
*msg
, size_t size
)
572 int err
= security_socket_sendmsg(sock
, msg
, size
);
574 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
577 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
580 struct sock_iocb siocb
;
583 init_sync_kiocb(&iocb
, NULL
);
584 iocb
.private = &siocb
;
585 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
586 if (-EIOCBQUEUED
== ret
)
587 ret
= wait_on_sync_kiocb(&iocb
);
590 EXPORT_SYMBOL(sock_sendmsg
);
592 int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
595 struct sock_iocb siocb
;
598 init_sync_kiocb(&iocb
, NULL
);
599 iocb
.private = &siocb
;
600 ret
= __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
);
601 if (-EIOCBQUEUED
== ret
)
602 ret
= wait_on_sync_kiocb(&iocb
);
606 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
607 struct kvec
*vec
, size_t num
, size_t size
)
609 mm_segment_t oldfs
= get_fs();
614 * the following is safe, since for compiler definitions of kvec and
615 * iovec are identical, yielding the same in-core layout and alignment
617 msg
->msg_iov
= (struct iovec
*)vec
;
618 msg
->msg_iovlen
= num
;
619 result
= sock_sendmsg(sock
, msg
, size
);
623 EXPORT_SYMBOL(kernel_sendmsg
);
625 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
628 *ts
= ktime_to_timespec(kt
);
636 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
638 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
641 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
642 struct timespec ts
[3];
644 struct skb_shared_hwtstamps
*shhwtstamps
=
647 /* Race occurred between timestamp enabling and packet
648 receiving. Fill in the current time for now. */
649 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
650 __net_timestamp(skb
);
652 if (need_software_tstamp
) {
653 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
655 skb_get_timestamp(skb
, &tv
);
656 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
659 skb_get_timestampns(skb
, &ts
[0]);
660 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
661 sizeof(ts
[0]), &ts
[0]);
666 memset(ts
, 0, sizeof(ts
));
667 if (skb
->tstamp
.tv64
&&
668 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
669 skb_get_timestampns(skb
, ts
+ 0);
673 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
674 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
676 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
677 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
681 put_cmsg(msg
, SOL_SOCKET
,
682 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
684 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
686 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
689 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
690 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
691 sizeof(__u32
), &skb
->dropcount
);
694 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
697 sock_recv_timestamp(msg
, sk
, skb
);
698 sock_recv_drops(msg
, sk
, skb
);
700 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
702 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
703 struct msghdr
*msg
, size_t size
, int flags
)
705 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
707 sock_update_classid(sock
->sk
);
715 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
718 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
719 struct msghdr
*msg
, size_t size
, int flags
)
721 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
723 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
726 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
727 size_t size
, int flags
)
730 struct sock_iocb siocb
;
733 init_sync_kiocb(&iocb
, NULL
);
734 iocb
.private = &siocb
;
735 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
736 if (-EIOCBQUEUED
== ret
)
737 ret
= wait_on_sync_kiocb(&iocb
);
740 EXPORT_SYMBOL(sock_recvmsg
);
742 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
743 size_t size
, int flags
)
746 struct sock_iocb siocb
;
749 init_sync_kiocb(&iocb
, NULL
);
750 iocb
.private = &siocb
;
751 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
752 if (-EIOCBQUEUED
== ret
)
753 ret
= wait_on_sync_kiocb(&iocb
);
758 * kernel_recvmsg - Receive a message from a socket (kernel space)
759 * @sock: The socket to receive the message from
760 * @msg: Received message
761 * @vec: Input s/g array for message data
762 * @num: Size of input s/g array
763 * @size: Number of bytes to read
764 * @flags: Message flags (MSG_DONTWAIT, etc...)
766 * On return the msg structure contains the scatter/gather array passed in the
767 * vec argument. The array is modified so that it consists of the unfilled
768 * portion of the original array.
770 * The returned value is the total number of bytes received, or an error.
772 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
773 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
775 mm_segment_t oldfs
= get_fs();
780 * the following is safe, since for compiler definitions of kvec and
781 * iovec are identical, yielding the same in-core layout and alignment
783 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
784 result
= sock_recvmsg(sock
, msg
, size
, flags
);
788 EXPORT_SYMBOL(kernel_recvmsg
);
790 static void sock_aio_dtor(struct kiocb
*iocb
)
792 kfree(iocb
->private);
795 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
796 int offset
, size_t size
, loff_t
*ppos
, int more
)
801 sock
= file
->private_data
;
803 flags
= !(file
->f_flags
& O_NONBLOCK
) ? 0 : MSG_DONTWAIT
;
807 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
810 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
811 struct pipe_inode_info
*pipe
, size_t len
,
814 struct socket
*sock
= file
->private_data
;
816 if (unlikely(!sock
->ops
->splice_read
))
819 sock_update_classid(sock
->sk
);
821 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
824 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
825 struct sock_iocb
*siocb
)
827 if (!is_sync_kiocb(iocb
)) {
828 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
831 iocb
->ki_dtor
= sock_aio_dtor
;
835 iocb
->private = siocb
;
839 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
840 struct file
*file
, const struct iovec
*iov
,
841 unsigned long nr_segs
)
843 struct socket
*sock
= file
->private_data
;
847 for (i
= 0; i
< nr_segs
; i
++)
848 size
+= iov
[i
].iov_len
;
850 msg
->msg_name
= NULL
;
851 msg
->msg_namelen
= 0;
852 msg
->msg_control
= NULL
;
853 msg
->msg_controllen
= 0;
854 msg
->msg_iov
= (struct iovec
*)iov
;
855 msg
->msg_iovlen
= nr_segs
;
856 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
858 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
861 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
862 unsigned long nr_segs
, loff_t pos
)
864 struct sock_iocb siocb
, *x
;
869 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
873 x
= alloc_sock_iocb(iocb
, &siocb
);
876 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
879 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
880 struct file
*file
, const struct iovec
*iov
,
881 unsigned long nr_segs
)
883 struct socket
*sock
= file
->private_data
;
887 for (i
= 0; i
< nr_segs
; i
++)
888 size
+= iov
[i
].iov_len
;
890 msg
->msg_name
= NULL
;
891 msg
->msg_namelen
= 0;
892 msg
->msg_control
= NULL
;
893 msg
->msg_controllen
= 0;
894 msg
->msg_iov
= (struct iovec
*)iov
;
895 msg
->msg_iovlen
= nr_segs
;
896 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
897 if (sock
->type
== SOCK_SEQPACKET
)
898 msg
->msg_flags
|= MSG_EOR
;
900 return __sock_sendmsg(iocb
, sock
, msg
, size
);
903 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
904 unsigned long nr_segs
, loff_t pos
)
906 struct sock_iocb siocb
, *x
;
911 x
= alloc_sock_iocb(iocb
, &siocb
);
915 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
919 * Atomic setting of ioctl hooks to avoid race
920 * with module unload.
923 static DEFINE_MUTEX(br_ioctl_mutex
);
924 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
926 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
928 mutex_lock(&br_ioctl_mutex
);
929 br_ioctl_hook
= hook
;
930 mutex_unlock(&br_ioctl_mutex
);
932 EXPORT_SYMBOL(brioctl_set
);
934 static DEFINE_MUTEX(vlan_ioctl_mutex
);
935 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
937 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
939 mutex_lock(&vlan_ioctl_mutex
);
940 vlan_ioctl_hook
= hook
;
941 mutex_unlock(&vlan_ioctl_mutex
);
943 EXPORT_SYMBOL(vlan_ioctl_set
);
945 static DEFINE_MUTEX(dlci_ioctl_mutex
);
946 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
948 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
950 mutex_lock(&dlci_ioctl_mutex
);
951 dlci_ioctl_hook
= hook
;
952 mutex_unlock(&dlci_ioctl_mutex
);
954 EXPORT_SYMBOL(dlci_ioctl_set
);
956 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
957 unsigned int cmd
, unsigned long arg
)
960 void __user
*argp
= (void __user
*)arg
;
962 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
965 * If this ioctl is unknown try to hand it down
968 if (err
== -ENOIOCTLCMD
)
969 err
= dev_ioctl(net
, cmd
, argp
);
975 * With an ioctl, arg may well be a user mode pointer, but we don't know
976 * what to do with it - that's up to the protocol still.
979 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
983 void __user
*argp
= (void __user
*)arg
;
987 sock
= file
->private_data
;
990 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
991 err
= dev_ioctl(net
, cmd
, argp
);
993 #ifdef CONFIG_WEXT_CORE
994 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
995 err
= dev_ioctl(net
, cmd
, argp
);
1002 if (get_user(pid
, (int __user
*)argp
))
1004 err
= f_setown(sock
->file
, pid
, 1);
1008 err
= put_user(f_getown(sock
->file
),
1009 (int __user
*)argp
);
1017 request_module("bridge");
1019 mutex_lock(&br_ioctl_mutex
);
1021 err
= br_ioctl_hook(net
, cmd
, argp
);
1022 mutex_unlock(&br_ioctl_mutex
);
1027 if (!vlan_ioctl_hook
)
1028 request_module("8021q");
1030 mutex_lock(&vlan_ioctl_mutex
);
1031 if (vlan_ioctl_hook
)
1032 err
= vlan_ioctl_hook(net
, argp
);
1033 mutex_unlock(&vlan_ioctl_mutex
);
1038 if (!dlci_ioctl_hook
)
1039 request_module("dlci");
1041 mutex_lock(&dlci_ioctl_mutex
);
1042 if (dlci_ioctl_hook
)
1043 err
= dlci_ioctl_hook(cmd
, argp
);
1044 mutex_unlock(&dlci_ioctl_mutex
);
1047 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1053 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1056 struct socket
*sock
= NULL
;
1058 err
= security_socket_create(family
, type
, protocol
, 1);
1062 sock
= sock_alloc();
1069 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1081 EXPORT_SYMBOL(sock_create_lite
);
1083 /* No kernel lock held - perfect */
1084 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1086 struct socket
*sock
;
1089 * We can't return errors to poll, so it's either yes or no.
1091 sock
= file
->private_data
;
1092 return sock
->ops
->poll(file
, sock
, wait
);
1095 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1097 struct socket
*sock
= file
->private_data
;
1099 return sock
->ops
->mmap(file
, sock
, vma
);
1102 static int sock_close(struct inode
*inode
, struct file
*filp
)
1105 * It was possible the inode is NULL we were
1106 * closing an unfinished socket.
1110 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1113 sock_release(SOCKET_I(inode
));
1118 * Update the socket async list
1120 * Fasync_list locking strategy.
1122 * 1. fasync_list is modified only under process context socket lock
1123 * i.e. under semaphore.
1124 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1125 * or under socket lock
1128 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1130 struct socket
*sock
= filp
->private_data
;
1131 struct sock
*sk
= sock
->sk
;
1132 struct socket_wq
*wq
;
1138 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1139 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1141 if (!wq
->fasync_list
)
1142 sock_reset_flag(sk
, SOCK_FASYNC
);
1144 sock_set_flag(sk
, SOCK_FASYNC
);
1150 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1152 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1154 struct socket_wq
*wq
;
1159 wq
= rcu_dereference(sock
->wq
);
1160 if (!wq
|| !wq
->fasync_list
) {
1165 case SOCK_WAKE_WAITD
:
1166 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1169 case SOCK_WAKE_SPACE
:
1170 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1175 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1178 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1183 EXPORT_SYMBOL(sock_wake_async
);
1185 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1186 struct socket
**res
, int kern
)
1189 struct socket
*sock
;
1190 const struct net_proto_family
*pf
;
1193 * Check protocol is in range
1195 if (family
< 0 || family
>= NPROTO
)
1196 return -EAFNOSUPPORT
;
1197 if (type
< 0 || type
>= SOCK_MAX
)
1202 This uglymoron is moved from INET layer to here to avoid
1203 deadlock in module load.
1205 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1209 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1215 err
= security_socket_create(family
, type
, protocol
, kern
);
1220 * Allocate the socket and allow the family to set things up. if
1221 * the protocol is 0, the family is instructed to select an appropriate
1224 sock
= sock_alloc();
1226 if (net_ratelimit())
1227 printk(KERN_WARNING
"socket: no more sockets\n");
1228 return -ENFILE
; /* Not exactly a match, but its the
1229 closest posix thing */
1234 #ifdef CONFIG_MODULES
1235 /* Attempt to load a protocol module if the find failed.
1237 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1238 * requested real, full-featured networking support upon configuration.
1239 * Otherwise module support will break!
1241 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1242 request_module("net-pf-%d", family
);
1246 pf
= rcu_dereference(net_families
[family
]);
1247 err
= -EAFNOSUPPORT
;
1252 * We will call the ->create function, that possibly is in a loadable
1253 * module, so we have to bump that loadable module refcnt first.
1255 if (!try_module_get(pf
->owner
))
1258 /* Now protected by module ref count */
1261 err
= pf
->create(net
, sock
, protocol
, kern
);
1263 goto out_module_put
;
1266 * Now to bump the refcnt of the [loadable] module that owns this
1267 * socket at sock_release time we decrement its refcnt.
1269 if (!try_module_get(sock
->ops
->owner
))
1270 goto out_module_busy
;
1273 * Now that we're done with the ->create function, the [loadable]
1274 * module can have its refcnt decremented
1276 module_put(pf
->owner
);
1277 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1279 goto out_sock_release
;
1285 err
= -EAFNOSUPPORT
;
1288 module_put(pf
->owner
);
1295 goto out_sock_release
;
1297 EXPORT_SYMBOL(__sock_create
);
1299 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1301 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1303 EXPORT_SYMBOL(sock_create
);
1305 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1307 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1309 EXPORT_SYMBOL(sock_create_kern
);
1311 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1314 struct socket
*sock
;
1317 /* Check the SOCK_* constants for consistency. */
1318 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1319 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1320 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1321 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1323 flags
= type
& ~SOCK_TYPE_MASK
;
1324 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1326 type
&= SOCK_TYPE_MASK
;
1328 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1329 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1331 retval
= sock_create(family
, type
, protocol
, &sock
);
1335 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1340 /* It may be already another descriptor 8) Not kernel problem. */
1349 * Create a pair of connected sockets.
1352 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1353 int __user
*, usockvec
)
1355 struct socket
*sock1
, *sock2
;
1357 struct file
*newfile1
, *newfile2
;
1360 flags
= type
& ~SOCK_TYPE_MASK
;
1361 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1363 type
&= SOCK_TYPE_MASK
;
1365 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1366 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1369 * Obtain the first socket and check if the underlying protocol
1370 * supports the socketpair call.
1373 err
= sock_create(family
, type
, protocol
, &sock1
);
1377 err
= sock_create(family
, type
, protocol
, &sock2
);
1381 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1383 goto out_release_both
;
1385 fd1
= sock_alloc_file(sock1
, &newfile1
, flags
);
1386 if (unlikely(fd1
< 0)) {
1388 goto out_release_both
;
1391 fd2
= sock_alloc_file(sock2
, &newfile2
, flags
);
1392 if (unlikely(fd2
< 0)) {
1396 sock_release(sock2
);
1400 audit_fd_pair(fd1
, fd2
);
1401 fd_install(fd1
, newfile1
);
1402 fd_install(fd2
, newfile2
);
1403 /* fd1 and fd2 may be already another descriptors.
1404 * Not kernel problem.
1407 err
= put_user(fd1
, &usockvec
[0]);
1409 err
= put_user(fd2
, &usockvec
[1]);
1418 sock_release(sock2
);
1420 sock_release(sock1
);
1426 * Bind a name to a socket. Nothing much to do here since it's
1427 * the protocol's responsibility to handle the local address.
1429 * We move the socket address to kernel space before we call
1430 * the protocol layer (having also checked the address is ok).
1433 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1435 struct socket
*sock
;
1436 struct sockaddr_storage address
;
1437 int err
, fput_needed
;
1439 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1441 err
= move_addr_to_kernel(umyaddr
, addrlen
, (struct sockaddr
*)&address
);
1443 err
= security_socket_bind(sock
,
1444 (struct sockaddr
*)&address
,
1447 err
= sock
->ops
->bind(sock
,
1451 fput_light(sock
->file
, fput_needed
);
1457 * Perform a listen. Basically, we allow the protocol to do anything
1458 * necessary for a listen, and if that works, we mark the socket as
1459 * ready for listening.
1462 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1464 struct socket
*sock
;
1465 int err
, fput_needed
;
1468 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1470 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1471 if ((unsigned)backlog
> somaxconn
)
1472 backlog
= somaxconn
;
1474 err
= security_socket_listen(sock
, backlog
);
1476 err
= sock
->ops
->listen(sock
, backlog
);
1478 fput_light(sock
->file
, fput_needed
);
1484 * For accept, we attempt to create a new socket, set up the link
1485 * with the client, wake up the client, then return the new
1486 * connected fd. We collect the address of the connector in kernel
1487 * space and move it to user at the very end. This is unclean because
1488 * we open the socket then return an error.
1490 * 1003.1g adds the ability to recvmsg() to query connection pending
1491 * status to recvmsg. We need to add that support in a way thats
1492 * clean when we restucture accept also.
1495 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1496 int __user
*, upeer_addrlen
, int, flags
)
1498 struct socket
*sock
, *newsock
;
1499 struct file
*newfile
;
1500 int err
, len
, newfd
, fput_needed
;
1501 struct sockaddr_storage address
;
1503 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1506 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1507 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1509 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1514 newsock
= sock_alloc();
1518 newsock
->type
= sock
->type
;
1519 newsock
->ops
= sock
->ops
;
1522 * We don't need try_module_get here, as the listening socket (sock)
1523 * has the protocol module (sock->ops->owner) held.
1525 __module_get(newsock
->ops
->owner
);
1527 newfd
= sock_alloc_file(newsock
, &newfile
, flags
);
1528 if (unlikely(newfd
< 0)) {
1530 sock_release(newsock
);
1534 err
= security_socket_accept(sock
, newsock
);
1538 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1542 if (upeer_sockaddr
) {
1543 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1545 err
= -ECONNABORTED
;
1548 err
= move_addr_to_user((struct sockaddr
*)&address
,
1549 len
, upeer_sockaddr
, upeer_addrlen
);
1554 /* File flags are not inherited via accept() unlike another OSes. */
1556 fd_install(newfd
, newfile
);
1560 fput_light(sock
->file
, fput_needed
);
1565 put_unused_fd(newfd
);
1569 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1570 int __user
*, upeer_addrlen
)
1572 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1576 * Attempt to connect to a socket with the server address. The address
1577 * is in user space so we verify it is OK and move it to kernel space.
1579 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1582 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1583 * other SEQPACKET protocols that take time to connect() as it doesn't
1584 * include the -EINPROGRESS status for such sockets.
1587 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1590 struct socket
*sock
;
1591 struct sockaddr_storage address
;
1592 int err
, fput_needed
;
1594 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1597 err
= move_addr_to_kernel(uservaddr
, addrlen
, (struct sockaddr
*)&address
);
1602 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1606 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1607 sock
->file
->f_flags
);
1609 fput_light(sock
->file
, fput_needed
);
1615 * Get the local address ('name') of a socket object. Move the obtained
1616 * name to user space.
1619 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1620 int __user
*, usockaddr_len
)
1622 struct socket
*sock
;
1623 struct sockaddr_storage address
;
1624 int len
, err
, fput_needed
;
1626 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1630 err
= security_socket_getsockname(sock
);
1634 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1637 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
, usockaddr_len
);
1640 fput_light(sock
->file
, fput_needed
);
1646 * Get the remote address ('name') of a socket object. Move the obtained
1647 * name to user space.
1650 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1651 int __user
*, usockaddr_len
)
1653 struct socket
*sock
;
1654 struct sockaddr_storage address
;
1655 int len
, err
, fput_needed
;
1657 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1659 err
= security_socket_getpeername(sock
);
1661 fput_light(sock
->file
, fput_needed
);
1666 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1669 err
= move_addr_to_user((struct sockaddr
*)&address
, len
, usockaddr
,
1671 fput_light(sock
->file
, fput_needed
);
1677 * Send a datagram to a given address. We move the address into kernel
1678 * space and check the user space data area is readable before invoking
1682 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1683 unsigned, flags
, struct sockaddr __user
*, addr
,
1686 struct socket
*sock
;
1687 struct sockaddr_storage address
;
1695 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1699 iov
.iov_base
= buff
;
1701 msg
.msg_name
= NULL
;
1704 msg
.msg_control
= NULL
;
1705 msg
.msg_controllen
= 0;
1706 msg
.msg_namelen
= 0;
1708 err
= move_addr_to_kernel(addr
, addr_len
, (struct sockaddr
*)&address
);
1711 msg
.msg_name
= (struct sockaddr
*)&address
;
1712 msg
.msg_namelen
= addr_len
;
1714 if (sock
->file
->f_flags
& O_NONBLOCK
)
1715 flags
|= MSG_DONTWAIT
;
1716 msg
.msg_flags
= flags
;
1717 err
= sock_sendmsg(sock
, &msg
, len
);
1720 fput_light(sock
->file
, fput_needed
);
1726 * Send a datagram down a socket.
1729 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1732 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1736 * Receive a frame from the socket and optionally record the address of the
1737 * sender. We verify the buffers are writable and if needed move the
1738 * sender address from kernel to user space.
1741 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1742 unsigned, flags
, struct sockaddr __user
*, addr
,
1743 int __user
*, addr_len
)
1745 struct socket
*sock
;
1748 struct sockaddr_storage address
;
1754 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1758 msg
.msg_control
= NULL
;
1759 msg
.msg_controllen
= 0;
1763 iov
.iov_base
= ubuf
;
1764 msg
.msg_name
= (struct sockaddr
*)&address
;
1765 msg
.msg_namelen
= sizeof(address
);
1766 if (sock
->file
->f_flags
& O_NONBLOCK
)
1767 flags
|= MSG_DONTWAIT
;
1768 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1770 if (err
>= 0 && addr
!= NULL
) {
1771 err2
= move_addr_to_user((struct sockaddr
*)&address
,
1772 msg
.msg_namelen
, addr
, addr_len
);
1777 fput_light(sock
->file
, fput_needed
);
1783 * Receive a datagram from a socket.
1786 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1789 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1793 * Set a socket option. Because we don't know the option lengths we have
1794 * to pass the user mode parameter for the protocols to sort out.
1797 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1798 char __user
*, optval
, int, optlen
)
1800 int err
, fput_needed
;
1801 struct socket
*sock
;
1806 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1808 err
= security_socket_setsockopt(sock
, level
, optname
);
1812 if (level
== SOL_SOCKET
)
1814 sock_setsockopt(sock
, level
, optname
, optval
,
1818 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1821 fput_light(sock
->file
, fput_needed
);
1827 * Get a socket option. Because we don't know the option lengths we have
1828 * to pass a user mode parameter for the protocols to sort out.
1831 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1832 char __user
*, optval
, int __user
*, optlen
)
1834 int err
, fput_needed
;
1835 struct socket
*sock
;
1837 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1839 err
= security_socket_getsockopt(sock
, level
, optname
);
1843 if (level
== SOL_SOCKET
)
1845 sock_getsockopt(sock
, level
, optname
, optval
,
1849 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1852 fput_light(sock
->file
, fput_needed
);
1858 * Shutdown a socket.
1861 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1863 int err
, fput_needed
;
1864 struct socket
*sock
;
1866 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1868 err
= security_socket_shutdown(sock
, how
);
1870 err
= sock
->ops
->shutdown(sock
, how
);
1871 fput_light(sock
->file
, fput_needed
);
1876 /* A couple of helpful macros for getting the address of the 32/64 bit
1877 * fields which are the same type (int / unsigned) on our platforms.
1879 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1880 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1881 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1883 static int __sys_sendmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1884 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
1886 struct compat_msghdr __user
*msg_compat
=
1887 (struct compat_msghdr __user
*)msg
;
1888 struct sockaddr_storage address
;
1889 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1890 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1891 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1892 /* 20 is size of ipv6_pktinfo */
1893 unsigned char *ctl_buf
= ctl
;
1894 int err
, ctl_len
, iov_size
, total_len
;
1897 if (MSG_CMSG_COMPAT
& flags
) {
1898 if (get_compat_msghdr(msg_sys
, msg_compat
))
1900 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
1903 /* do not move before msg_sys is valid */
1905 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
1908 /* Check whether to allocate the iovec area */
1910 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
1911 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
1912 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
1917 /* This will also move the address data into kernel space */
1918 if (MSG_CMSG_COMPAT
& flags
) {
1919 err
= verify_compat_iovec(msg_sys
, iov
,
1920 (struct sockaddr
*)&address
,
1923 err
= verify_iovec(msg_sys
, iov
,
1924 (struct sockaddr
*)&address
,
1932 if (msg_sys
->msg_controllen
> INT_MAX
)
1934 ctl_len
= msg_sys
->msg_controllen
;
1935 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1937 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
1941 ctl_buf
= msg_sys
->msg_control
;
1942 ctl_len
= msg_sys
->msg_controllen
;
1943 } else if (ctl_len
) {
1944 if (ctl_len
> sizeof(ctl
)) {
1945 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1946 if (ctl_buf
== NULL
)
1951 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1952 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1953 * checking falls down on this.
1955 if (copy_from_user(ctl_buf
,
1956 (void __user __force
*)msg_sys
->msg_control
,
1959 msg_sys
->msg_control
= ctl_buf
;
1961 msg_sys
->msg_flags
= flags
;
1963 if (sock
->file
->f_flags
& O_NONBLOCK
)
1964 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
1965 err
= (nosec
? sock_sendmsg_nosec
: sock_sendmsg
)(sock
, msg_sys
,
1970 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1972 if (iov
!= iovstack
)
1973 sock_kfree_s(sock
->sk
, iov
, iov_size
);
1979 * BSD sendmsg interface
1982 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned, flags
)
1984 int fput_needed
, err
;
1985 struct msghdr msg_sys
;
1986 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1991 err
= __sys_sendmsg(sock
, msg
, &msg_sys
, flags
, 0);
1993 fput_light(sock
->file
, fput_needed
);
1999 * Linux sendmmsg interface
2002 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2005 int fput_needed
, err
, datagrams
;
2006 struct socket
*sock
;
2007 struct mmsghdr __user
*entry
;
2008 struct compat_mmsghdr __user
*compat_entry
;
2009 struct msghdr msg_sys
;
2013 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2017 err
= sock_error(sock
->sk
);
2022 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2024 while (datagrams
< vlen
) {
2026 * No need to ask LSM for more than the first datagram.
2028 if (MSG_CMSG_COMPAT
& flags
) {
2029 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)compat_entry
,
2030 &msg_sys
, flags
, datagrams
);
2033 err
= __put_user(err
, &compat_entry
->msg_len
);
2036 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)entry
,
2037 &msg_sys
, flags
, datagrams
);
2040 err
= put_user(err
, &entry
->msg_len
);
2050 fput_light(sock
->file
, fput_needed
);
2055 if (datagrams
!= 0) {
2057 * We may send less entries than requested (vlen) if the
2058 * sock is non blocking...
2060 if (err
!= -EAGAIN
) {
2062 * ... or if sendmsg returns an error after we
2063 * send some datagrams, where we record the
2064 * error to return on the next call or if the
2065 * app asks about it using getsockopt(SO_ERROR).
2067 sock
->sk
->sk_err
= -err
;
2076 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2077 unsigned int, vlen
, unsigned int, flags
)
2079 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2082 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
2083 struct msghdr
*msg_sys
, unsigned flags
, int nosec
)
2085 struct compat_msghdr __user
*msg_compat
=
2086 (struct compat_msghdr __user
*)msg
;
2087 struct iovec iovstack
[UIO_FASTIOV
];
2088 struct iovec
*iov
= iovstack
;
2089 unsigned long cmsg_ptr
;
2090 int err
, iov_size
, total_len
, len
;
2092 /* kernel mode address */
2093 struct sockaddr_storage addr
;
2095 /* user mode address pointers */
2096 struct sockaddr __user
*uaddr
;
2097 int __user
*uaddr_len
;
2099 if (MSG_CMSG_COMPAT
& flags
) {
2100 if (get_compat_msghdr(msg_sys
, msg_compat
))
2102 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2106 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2109 /* Check whether to allocate the iovec area */
2111 iov_size
= msg_sys
->msg_iovlen
* sizeof(struct iovec
);
2112 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2113 iov
= sock_kmalloc(sock
->sk
, iov_size
, GFP_KERNEL
);
2119 * Save the user-mode address (verify_iovec will change the
2120 * kernel msghdr to use the kernel address space)
2123 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2124 uaddr_len
= COMPAT_NAMELEN(msg
);
2125 if (MSG_CMSG_COMPAT
& flags
) {
2126 err
= verify_compat_iovec(msg_sys
, iov
,
2127 (struct sockaddr
*)&addr
,
2130 err
= verify_iovec(msg_sys
, iov
,
2131 (struct sockaddr
*)&addr
,
2137 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2138 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2140 if (sock
->file
->f_flags
& O_NONBLOCK
)
2141 flags
|= MSG_DONTWAIT
;
2142 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2148 if (uaddr
!= NULL
) {
2149 err
= move_addr_to_user((struct sockaddr
*)&addr
,
2150 msg_sys
->msg_namelen
, uaddr
,
2155 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2159 if (MSG_CMSG_COMPAT
& flags
)
2160 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2161 &msg_compat
->msg_controllen
);
2163 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2164 &msg
->msg_controllen
);
2170 if (iov
!= iovstack
)
2171 sock_kfree_s(sock
->sk
, iov
, iov_size
);
2177 * BSD recvmsg interface
2180 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2181 unsigned int, flags
)
2183 int fput_needed
, err
;
2184 struct msghdr msg_sys
;
2185 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2190 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2192 fput_light(sock
->file
, fput_needed
);
2198 * Linux recvmmsg interface
2201 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2202 unsigned int flags
, struct timespec
*timeout
)
2204 int fput_needed
, err
, datagrams
;
2205 struct socket
*sock
;
2206 struct mmsghdr __user
*entry
;
2207 struct compat_mmsghdr __user
*compat_entry
;
2208 struct msghdr msg_sys
;
2209 struct timespec end_time
;
2212 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2218 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2222 err
= sock_error(sock
->sk
);
2227 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2229 while (datagrams
< vlen
) {
2231 * No need to ask LSM for more than the first datagram.
2233 if (MSG_CMSG_COMPAT
& flags
) {
2234 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2235 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2239 err
= __put_user(err
, &compat_entry
->msg_len
);
2242 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2243 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2247 err
= put_user(err
, &entry
->msg_len
);
2255 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2256 if (flags
& MSG_WAITFORONE
)
2257 flags
|= MSG_DONTWAIT
;
2260 ktime_get_ts(timeout
);
2261 *timeout
= timespec_sub(end_time
, *timeout
);
2262 if (timeout
->tv_sec
< 0) {
2263 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2267 /* Timeout, return less than vlen datagrams */
2268 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2272 /* Out of band data, return right away */
2273 if (msg_sys
.msg_flags
& MSG_OOB
)
2278 fput_light(sock
->file
, fput_needed
);
2283 if (datagrams
!= 0) {
2285 * We may return less entries than requested (vlen) if the
2286 * sock is non block and there aren't enough datagrams...
2288 if (err
!= -EAGAIN
) {
2290 * ... or if recvmsg returns an error after we
2291 * received some datagrams, where we record the
2292 * error to return on the next call or if the
2293 * app asks about it using getsockopt(SO_ERROR).
2295 sock
->sk
->sk_err
= -err
;
2304 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2305 unsigned int, vlen
, unsigned int, flags
,
2306 struct timespec __user
*, timeout
)
2309 struct timespec timeout_sys
;
2312 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2314 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2317 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2319 if (datagrams
> 0 &&
2320 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2321 datagrams
= -EFAULT
;
2326 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2327 /* Argument list sizes for sys_socketcall */
2328 #define AL(x) ((x) * sizeof(unsigned long))
2329 static const unsigned char nargs
[21] = {
2330 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2331 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2332 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2339 * System call vectors.
2341 * Argument checking cleaned up. Saved 20% in size.
2342 * This function doesn't need to set the kernel lock because
2343 * it is set by the callees.
2346 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2349 unsigned long a0
, a1
;
2353 if (call
< 1 || call
> SYS_SENDMMSG
)
2357 if (len
> sizeof(a
))
2360 /* copy_from_user should be SMP safe. */
2361 if (copy_from_user(a
, args
, len
))
2364 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2371 err
= sys_socket(a0
, a1
, a
[2]);
2374 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2377 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2380 err
= sys_listen(a0
, a1
);
2383 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2384 (int __user
*)a
[2], 0);
2386 case SYS_GETSOCKNAME
:
2388 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2389 (int __user
*)a
[2]);
2391 case SYS_GETPEERNAME
:
2393 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2394 (int __user
*)a
[2]);
2396 case SYS_SOCKETPAIR
:
2397 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2400 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2403 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2404 (struct sockaddr __user
*)a
[4], a
[5]);
2407 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2410 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2411 (struct sockaddr __user
*)a
[4],
2412 (int __user
*)a
[5]);
2415 err
= sys_shutdown(a0
, a1
);
2417 case SYS_SETSOCKOPT
:
2418 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2420 case SYS_GETSOCKOPT
:
2422 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2423 (int __user
*)a
[4]);
2426 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2429 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2432 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2435 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2436 (struct timespec __user
*)a
[4]);
2439 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2440 (int __user
*)a
[2], a
[3]);
2449 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2452 * sock_register - add a socket protocol handler
2453 * @ops: description of protocol
2455 * This function is called by a protocol handler that wants to
2456 * advertise its address family, and have it linked into the
2457 * socket interface. The value ops->family coresponds to the
2458 * socket system call protocol family.
2460 int sock_register(const struct net_proto_family
*ops
)
2464 if (ops
->family
>= NPROTO
) {
2465 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2470 spin_lock(&net_family_lock
);
2471 if (rcu_dereference_protected(net_families
[ops
->family
],
2472 lockdep_is_held(&net_family_lock
)))
2475 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2478 spin_unlock(&net_family_lock
);
2480 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2483 EXPORT_SYMBOL(sock_register
);
2486 * sock_unregister - remove a protocol handler
2487 * @family: protocol family to remove
2489 * This function is called by a protocol handler that wants to
2490 * remove its address family, and have it unlinked from the
2491 * new socket creation.
2493 * If protocol handler is a module, then it can use module reference
2494 * counts to protect against new references. If protocol handler is not
2495 * a module then it needs to provide its own protection in
2496 * the ops->create routine.
2498 void sock_unregister(int family
)
2500 BUG_ON(family
< 0 || family
>= NPROTO
);
2502 spin_lock(&net_family_lock
);
2503 rcu_assign_pointer(net_families
[family
], NULL
);
2504 spin_unlock(&net_family_lock
);
2508 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2510 EXPORT_SYMBOL(sock_unregister
);
2512 static int __init
sock_init(void)
2517 * Initialize sock SLAB cache.
2523 * Initialize skbuff SLAB cache
2528 * Initialize the protocols module.
2533 err
= register_filesystem(&sock_fs_type
);
2536 sock_mnt
= kern_mount(&sock_fs_type
);
2537 if (IS_ERR(sock_mnt
)) {
2538 err
= PTR_ERR(sock_mnt
);
2542 /* The real protocol initialization is performed in later initcalls.
2545 #ifdef CONFIG_NETFILTER
2549 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2550 skb_timestamping_init();
2557 unregister_filesystem(&sock_fs_type
);
2562 core_initcall(sock_init
); /* early initcall */
2564 #ifdef CONFIG_PROC_FS
2565 void socket_seq_show(struct seq_file
*seq
)
2570 for_each_possible_cpu(cpu
)
2571 counter
+= per_cpu(sockets_in_use
, cpu
);
2573 /* It can be negative, by the way. 8) */
2577 seq_printf(seq
, "sockets: used %d\n", counter
);
2579 #endif /* CONFIG_PROC_FS */
2581 #ifdef CONFIG_COMPAT
2582 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2583 unsigned int cmd
, struct compat_timeval __user
*up
)
2585 mm_segment_t old_fs
= get_fs();
2590 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2593 err
= put_user(ktv
.tv_sec
, &up
->tv_sec
);
2594 err
|= __put_user(ktv
.tv_usec
, &up
->tv_usec
);
2599 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2600 unsigned int cmd
, struct compat_timespec __user
*up
)
2602 mm_segment_t old_fs
= get_fs();
2603 struct timespec kts
;
2607 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2610 err
= put_user(kts
.tv_sec
, &up
->tv_sec
);
2611 err
|= __put_user(kts
.tv_nsec
, &up
->tv_nsec
);
2616 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2618 struct ifreq __user
*uifr
;
2621 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2622 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2625 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2629 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2635 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2637 struct compat_ifconf ifc32
;
2639 struct ifconf __user
*uifc
;
2640 struct compat_ifreq __user
*ifr32
;
2641 struct ifreq __user
*ifr
;
2645 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2648 if (ifc32
.ifcbuf
== 0) {
2652 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2654 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2655 sizeof(struct ifreq
);
2656 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2658 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2659 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2660 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2661 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2667 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2670 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2674 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2678 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2680 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2681 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2682 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2688 if (ifc32
.ifcbuf
== 0) {
2689 /* Translate from 64-bit structure multiple to
2693 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2698 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2704 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2706 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2707 bool convert_in
= false, convert_out
= false;
2708 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2709 struct ethtool_rxnfc __user
*rxnfc
;
2710 struct ifreq __user
*ifr
;
2711 u32 rule_cnt
= 0, actual_rule_cnt
;
2716 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2719 compat_rxnfc
= compat_ptr(data
);
2721 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2724 /* Most ethtool structures are defined without padding.
2725 * Unfortunately struct ethtool_rxnfc is an exception.
2730 case ETHTOOL_GRXCLSRLALL
:
2731 /* Buffer size is variable */
2732 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2734 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2736 buf_size
+= rule_cnt
* sizeof(u32
);
2738 case ETHTOOL_GRXRINGS
:
2739 case ETHTOOL_GRXCLSRLCNT
:
2740 case ETHTOOL_GRXCLSRULE
:
2743 case ETHTOOL_SRXCLSRLDEL
:
2744 case ETHTOOL_SRXCLSRLINS
:
2745 buf_size
+= sizeof(struct ethtool_rxnfc
);
2750 ifr
= compat_alloc_user_space(buf_size
);
2751 rxnfc
= (void *)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2753 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2756 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2757 &ifr
->ifr_ifru
.ifru_data
))
2761 /* We expect there to be holes between fs.m_ext and
2762 * fs.ring_cookie and at the end of fs, but nowhere else.
2764 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2765 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2766 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2767 sizeof(rxnfc
->fs
.m_ext
));
2769 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2770 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2771 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2772 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2774 if (copy_in_user(rxnfc
, compat_rxnfc
,
2775 (void *)(&rxnfc
->fs
.m_ext
+ 1) -
2777 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2778 &compat_rxnfc
->fs
.ring_cookie
,
2779 (void *)(&rxnfc
->fs
.location
+ 1) -
2780 (void *)&rxnfc
->fs
.ring_cookie
) ||
2781 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2782 sizeof(rxnfc
->rule_cnt
)))
2786 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2791 if (copy_in_user(compat_rxnfc
, rxnfc
,
2792 (const void *)(&rxnfc
->fs
.m_ext
+ 1) -
2793 (const void *)rxnfc
) ||
2794 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2795 &rxnfc
->fs
.ring_cookie
,
2796 (const void *)(&rxnfc
->fs
.location
+ 1) -
2797 (const void *)&rxnfc
->fs
.ring_cookie
) ||
2798 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2799 sizeof(rxnfc
->rule_cnt
)))
2802 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2803 /* As an optimisation, we only copy the actual
2804 * number of rules that the underlying
2805 * function returned. Since Mallory might
2806 * change the rule count in user memory, we
2807 * check that it is less than the rule count
2808 * originally given (as the user buffer size),
2809 * which has been range-checked.
2811 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2813 if (actual_rule_cnt
< rule_cnt
)
2814 rule_cnt
= actual_rule_cnt
;
2815 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2816 &rxnfc
->rule_locs
[0],
2817 rule_cnt
* sizeof(u32
)))
2825 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2828 compat_uptr_t uptr32
;
2829 struct ifreq __user
*uifr
;
2831 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2832 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2835 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2838 uptr
= compat_ptr(uptr32
);
2840 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2843 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2846 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2847 struct compat_ifreq __user
*ifr32
)
2850 struct ifreq __user
*uifr
;
2851 mm_segment_t old_fs
;
2857 case SIOCBONDENSLAVE
:
2858 case SIOCBONDRELEASE
:
2859 case SIOCBONDSETHWADDR
:
2860 case SIOCBONDCHANGEACTIVE
:
2861 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2866 err
= dev_ioctl(net
, cmd
,
2867 (struct ifreq __user __force
*) &kifr
);
2871 case SIOCBONDSLAVEINFOQUERY
:
2872 case SIOCBONDINFOQUERY
:
2873 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2874 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2877 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2880 datap
= compat_ptr(data
);
2881 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2884 return dev_ioctl(net
, cmd
, uifr
);
2890 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2891 struct compat_ifreq __user
*u_ifreq32
)
2893 struct ifreq __user
*u_ifreq64
;
2894 char tmp_buf
[IFNAMSIZ
];
2895 void __user
*data64
;
2898 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2901 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2903 data64
= compat_ptr(data32
);
2905 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2907 /* Don't check these user accesses, just let that get trapped
2908 * in the ioctl handler instead.
2910 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2913 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2916 return dev_ioctl(net
, cmd
, u_ifreq64
);
2919 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2920 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2922 struct ifreq __user
*uifr
;
2925 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2926 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2929 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2940 case SIOCGIFBRDADDR
:
2941 case SIOCGIFDSTADDR
:
2942 case SIOCGIFNETMASK
:
2947 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2955 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2956 struct compat_ifreq __user
*uifr32
)
2959 struct compat_ifmap __user
*uifmap32
;
2960 mm_segment_t old_fs
;
2963 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2964 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2965 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2966 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2967 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2968 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2969 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2970 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2976 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
2979 if (cmd
== SIOCGIFMAP
&& !err
) {
2980 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2981 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2982 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2983 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2984 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2985 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2986 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2993 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2996 compat_uptr_t uptr32
;
2997 struct ifreq __user
*uifr
;
2999 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3000 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
3003 if (get_user(uptr32
, &uifr32
->ifr_data
))
3006 uptr
= compat_ptr(uptr32
);
3008 if (put_user(uptr
, &uifr
->ifr_data
))
3011 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
3016 struct sockaddr rt_dst
; /* target address */
3017 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3018 struct sockaddr rt_genmask
; /* target network mask (IP) */
3019 unsigned short rt_flags
;
3022 unsigned char rt_tos
;
3023 unsigned char rt_class
;
3025 short rt_metric
; /* +1 for binary compatibility! */
3026 /* char * */ u32 rt_dev
; /* forcing the device at add */
3027 u32 rt_mtu
; /* per route MTU/Window */
3028 u32 rt_window
; /* Window clamping */
3029 unsigned short rt_irtt
; /* Initial RTT */
3032 struct in6_rtmsg32
{
3033 struct in6_addr rtmsg_dst
;
3034 struct in6_addr rtmsg_src
;
3035 struct in6_addr rtmsg_gateway
;
3045 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3046 unsigned int cmd
, void __user
*argp
)
3050 struct in6_rtmsg r6
;
3054 mm_segment_t old_fs
= get_fs();
3056 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3057 struct in6_rtmsg32 __user
*ur6
= argp
;
3058 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3059 3 * sizeof(struct in6_addr
));
3060 ret
|= __get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3061 ret
|= __get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3062 ret
|= __get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3063 ret
|= __get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3064 ret
|= __get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3065 ret
|= __get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3066 ret
|= __get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3070 struct rtentry32 __user
*ur4
= argp
;
3071 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3072 3 * sizeof(struct sockaddr
));
3073 ret
|= __get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3074 ret
|= __get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3075 ret
|= __get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3076 ret
|= __get_user(r4
.rt_window
, &(ur4
->rt_window
));
3077 ret
|= __get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3078 ret
|= __get_user(rtdev
, &(ur4
->rt_dev
));
3080 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3081 r4
.rt_dev
= (char __user __force
*)devname
;
3095 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3102 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3103 * for some operations; this forces use of the newer bridge-utils that
3104 * use compatible ioctls
3106 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3110 if (get_user(tmp
, argp
))
3112 if (tmp
== BRCTL_GET_VERSION
)
3113 return BRCTL_VERSION
+ 1;
3117 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3118 unsigned int cmd
, unsigned long arg
)
3120 void __user
*argp
= compat_ptr(arg
);
3121 struct sock
*sk
= sock
->sk
;
3122 struct net
*net
= sock_net(sk
);
3124 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3125 return siocdevprivate_ioctl(net
, cmd
, argp
);
3130 return old_bridge_ioctl(argp
);
3132 return dev_ifname32(net
, argp
);
3134 return dev_ifconf(net
, argp
);
3136 return ethtool_ioctl(net
, argp
);
3138 return compat_siocwandev(net
, argp
);
3141 return compat_sioc_ifmap(net
, cmd
, argp
);
3142 case SIOCBONDENSLAVE
:
3143 case SIOCBONDRELEASE
:
3144 case SIOCBONDSETHWADDR
:
3145 case SIOCBONDSLAVEINFOQUERY
:
3146 case SIOCBONDINFOQUERY
:
3147 case SIOCBONDCHANGEACTIVE
:
3148 return bond_ioctl(net
, cmd
, argp
);
3151 return routing_ioctl(net
, sock
, cmd
, argp
);
3153 return do_siocgstamp(net
, sock
, cmd
, argp
);
3155 return do_siocgstampns(net
, sock
, cmd
, argp
);
3157 return compat_siocshwtstamp(net
, argp
);
3169 return sock_ioctl(file
, cmd
, arg
);
3186 case SIOCSIFHWBROADCAST
:
3188 case SIOCGIFBRDADDR
:
3189 case SIOCSIFBRDADDR
:
3190 case SIOCGIFDSTADDR
:
3191 case SIOCSIFDSTADDR
:
3192 case SIOCGIFNETMASK
:
3193 case SIOCSIFNETMASK
:
3204 return dev_ifsioc(net
, sock
, cmd
, argp
);
3210 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3213 /* Prevent warning from compat_sys_ioctl, these always
3214 * result in -EINVAL in the native case anyway. */
3227 return -ENOIOCTLCMD
;
3230 static long compat_sock_ioctl(struct file
*file
, unsigned cmd
,
3233 struct socket
*sock
= file
->private_data
;
3234 int ret
= -ENOIOCTLCMD
;
3241 if (sock
->ops
->compat_ioctl
)
3242 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3244 if (ret
== -ENOIOCTLCMD
&&
3245 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3246 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3248 if (ret
== -ENOIOCTLCMD
)
3249 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3255 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3257 return sock
->ops
->bind(sock
, addr
, addrlen
);
3259 EXPORT_SYMBOL(kernel_bind
);
3261 int kernel_listen(struct socket
*sock
, int backlog
)
3263 return sock
->ops
->listen(sock
, backlog
);
3265 EXPORT_SYMBOL(kernel_listen
);
3267 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3269 struct sock
*sk
= sock
->sk
;
3272 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3277 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3279 sock_release(*newsock
);
3284 (*newsock
)->ops
= sock
->ops
;
3285 __module_get((*newsock
)->ops
->owner
);
3290 EXPORT_SYMBOL(kernel_accept
);
3292 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3295 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3297 EXPORT_SYMBOL(kernel_connect
);
3299 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3302 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3304 EXPORT_SYMBOL(kernel_getsockname
);
3306 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3309 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3311 EXPORT_SYMBOL(kernel_getpeername
);
3313 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3314 char *optval
, int *optlen
)
3316 mm_segment_t oldfs
= get_fs();
3317 char __user
*uoptval
;
3318 int __user
*uoptlen
;
3321 uoptval
= (char __user __force
*) optval
;
3322 uoptlen
= (int __user __force
*) optlen
;
3325 if (level
== SOL_SOCKET
)
3326 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3328 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3333 EXPORT_SYMBOL(kernel_getsockopt
);
3335 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3336 char *optval
, unsigned int optlen
)
3338 mm_segment_t oldfs
= get_fs();
3339 char __user
*uoptval
;
3342 uoptval
= (char __user __force
*) optval
;
3345 if (level
== SOL_SOCKET
)
3346 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3348 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3353 EXPORT_SYMBOL(kernel_setsockopt
);
3355 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3356 size_t size
, int flags
)
3358 sock_update_classid(sock
->sk
);
3360 if (sock
->ops
->sendpage
)
3361 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3363 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3365 EXPORT_SYMBOL(kernel_sendpage
);
3367 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3369 mm_segment_t oldfs
= get_fs();
3373 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3378 EXPORT_SYMBOL(kernel_sock_ioctl
);
3380 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3382 return sock
->ops
->shutdown(sock
, how
);
3384 EXPORT_SYMBOL(kernel_sock_shutdown
);