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>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
109 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
110 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
111 unsigned long nr_segs
, loff_t pos
);
112 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
113 unsigned long nr_segs
, loff_t pos
);
114 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
116 static int sock_close(struct inode
*inode
, struct file
*file
);
117 static unsigned int sock_poll(struct file
*file
,
118 struct poll_table_struct
*wait
);
119 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
121 static long compat_sock_ioctl(struct file
*file
,
122 unsigned int cmd
, unsigned long arg
);
124 static int sock_fasync(int fd
, struct file
*filp
, int on
);
125 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
126 int offset
, size_t size
, loff_t
*ppos
, int more
);
127 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
128 struct pipe_inode_info
*pipe
, size_t len
,
132 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
133 * in the operation structures but are done directly via the socketcall() multiplexor.
136 static const struct file_operations socket_file_ops
= {
137 .owner
= THIS_MODULE
,
139 .aio_read
= sock_aio_read
,
140 .aio_write
= sock_aio_write
,
142 .unlocked_ioctl
= sock_ioctl
,
144 .compat_ioctl
= compat_sock_ioctl
,
147 .open
= sock_no_open
, /* special open code to disallow open via /proc */
148 .release
= sock_close
,
149 .fasync
= sock_fasync
,
150 .sendpage
= sock_sendpage
,
151 .splice_write
= generic_splice_sendpage
,
152 .splice_read
= sock_splice_read
,
156 * The protocol list. Each protocol is registered in here.
159 static DEFINE_SPINLOCK(net_family_lock
);
160 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
163 * Statistics counters of the socket lists
166 static DEFINE_PER_CPU(int, sockets_in_use
);
170 * Move socket addresses back and forth across the kernel/user
171 * divide and look after the messy bits.
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
185 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
187 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
191 if (copy_from_user(kaddr
, uaddr
, ulen
))
193 return audit_sockaddr(ulen
, kaddr
);
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
213 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
214 void __user
*uaddr
, int __user
*ulen
)
219 err
= get_user(len
, ulen
);
224 if (len
< 0 || len
> sizeof(struct sockaddr_storage
))
227 if (audit_sockaddr(klen
, kaddr
))
229 if (copy_to_user(uaddr
, kaddr
, len
))
233 * "fromlen shall refer to the value before truncation.."
236 return __put_user(klen
, ulen
);
239 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
241 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
243 struct socket_alloc
*ei
;
244 struct socket_wq
*wq
;
246 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
249 wq
= kmalloc(sizeof(*wq
), GFP_KERNEL
);
251 kmem_cache_free(sock_inode_cachep
, ei
);
254 init_waitqueue_head(&wq
->wait
);
255 wq
->fasync_list
= NULL
;
256 RCU_INIT_POINTER(ei
->socket
.wq
, wq
);
258 ei
->socket
.state
= SS_UNCONNECTED
;
259 ei
->socket
.flags
= 0;
260 ei
->socket
.ops
= NULL
;
261 ei
->socket
.sk
= NULL
;
262 ei
->socket
.file
= NULL
;
264 return &ei
->vfs_inode
;
267 static void sock_destroy_inode(struct inode
*inode
)
269 struct socket_alloc
*ei
;
270 struct socket_wq
*wq
;
272 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
273 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
275 kmem_cache_free(sock_inode_cachep
, ei
);
278 static void init_once(void *foo
)
280 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
282 inode_init_once(&ei
->vfs_inode
);
285 static int init_inodecache(void)
287 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
288 sizeof(struct socket_alloc
),
290 (SLAB_HWCACHE_ALIGN
|
291 SLAB_RECLAIM_ACCOUNT
|
294 if (sock_inode_cachep
== NULL
)
299 static const struct super_operations sockfs_ops
= {
300 .alloc_inode
= sock_alloc_inode
,
301 .destroy_inode
= sock_destroy_inode
,
302 .statfs
= simple_statfs
,
306 * sockfs_dname() is called from d_path().
308 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
310 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
311 dentry
->d_inode
->i_ino
);
314 static const struct dentry_operations sockfs_dentry_operations
= {
315 .d_dname
= sockfs_dname
,
318 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
319 int flags
, const char *dev_name
, void *data
)
321 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
322 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
325 static struct vfsmount
*sock_mnt __read_mostly
;
327 static struct file_system_type sock_fs_type
= {
329 .mount
= sockfs_mount
,
330 .kill_sb
= kill_anon_super
,
334 * Obtains the first available file descriptor and sets it up for use.
336 * These functions create file structures and maps them to fd space
337 * of the current process. On success it returns file descriptor
338 * and file struct implicitly stored in sock->file.
339 * Note that another thread may close file descriptor before we return
340 * from this function. We use the fact that now we do not refer
341 * to socket after mapping. If one day we will need it, this
342 * function will increment ref. count on file by 1.
344 * In any case returned fd MAY BE not valid!
345 * This race condition is unavoidable
346 * with shared fd spaces, we cannot solve it inside kernel,
347 * but we take care of internal coherence yet.
350 static int sock_alloc_file(struct socket
*sock
, struct file
**f
, int flags
,
353 struct qstr name
= { .name
= "" };
358 fd
= get_unused_fd_flags(flags
);
359 if (unlikely(fd
< 0))
364 name
.len
= strlen(name
.name
);
365 } else if (sock
->sk
) {
366 name
.name
= sock
->sk
->sk_prot_creator
->name
;
367 name
.len
= strlen(name
.name
);
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
, NULL
);
404 fd_install(fd
, newfile
);
408 EXPORT_SYMBOL(sock_map_fd
);
410 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 */
418 EXPORT_SYMBOL(sock_from_file
);
421 * sockfd_lookup - Go from a file number to its socket slot
423 * @err: pointer to an error code return
425 * The file handle passed in is locked and the socket it is bound
426 * too is returned. If an error occurs the err pointer is overwritten
427 * with a negative errno code and NULL is returned. The function checks
428 * for both invalid handles and passing a handle which is not a socket.
430 * On a success the socket object pointer is returned.
433 struct socket
*sockfd_lookup(int fd
, int *err
)
444 sock
= sock_from_file(file
, err
);
449 EXPORT_SYMBOL(sockfd_lookup
);
451 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
457 file
= fget_light(fd
, fput_needed
);
459 sock
= sock_from_file(file
, err
);
462 fput_light(file
, *fput_needed
);
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t
sockfs_getxattr(struct dentry
*dentry
,
471 const char *name
, void *value
, size_t size
)
473 const char *proto_name
;
478 if (!strncmp(name
, XATTR_NAME_SOCKPROTONAME
, XATTR_NAME_SOCKPROTONAME_LEN
)) {
479 proto_name
= dentry
->d_name
.name
;
480 proto_size
= strlen(proto_name
);
484 if (proto_size
+ 1 > size
)
487 strncpy(value
, proto_name
, proto_size
+ 1);
489 error
= proto_size
+ 1;
496 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
502 len
= security_inode_listsecurity(dentry
->d_inode
, buffer
, size
);
512 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
517 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
524 static const struct inode_operations sockfs_inode_ops
= {
525 .getxattr
= sockfs_getxattr
,
526 .listxattr
= sockfs_listxattr
,
530 * sock_alloc - allocate a socket
532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
537 static struct socket
*sock_alloc(void)
542 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
546 sock
= SOCKET_I(inode
);
548 kmemcheck_annotate_bitfield(sock
, type
);
549 inode
->i_ino
= get_next_ino();
550 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
551 inode
->i_uid
= current_fsuid();
552 inode
->i_gid
= current_fsgid();
553 inode
->i_op
= &sockfs_inode_ops
;
555 this_cpu_add(sockets_in_use
, 1);
560 * In theory you can't get an open on this inode, but /proc provides
561 * a back door. Remember to keep it shut otherwise you'll let the
562 * creepy crawlies in.
565 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
570 const struct file_operations bad_sock_fops
= {
571 .owner
= THIS_MODULE
,
572 .open
= sock_no_open
,
573 .llseek
= noop_llseek
,
577 * sock_release - close a socket
578 * @sock: socket to close
580 * The socket is released from the protocol stack if it has a release
581 * callback, and the inode is then released if the socket is bound to
582 * an inode not a file.
585 void sock_release(struct socket
*sock
)
588 struct module
*owner
= sock
->ops
->owner
;
590 sock
->ops
->release(sock
);
595 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
596 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
598 if (test_bit(SOCK_EXTERNALLY_ALLOCATED
, &sock
->flags
))
601 this_cpu_sub(sockets_in_use
, 1);
603 iput(SOCK_INODE(sock
));
608 EXPORT_SYMBOL(sock_release
);
610 int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
)
613 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
614 *tx_flags
|= SKBTX_HW_TSTAMP
;
615 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
616 *tx_flags
|= SKBTX_SW_TSTAMP
;
617 if (sock_flag(sk
, SOCK_WIFI_STATUS
))
618 *tx_flags
|= SKBTX_WIFI_STATUS
;
621 EXPORT_SYMBOL(sock_tx_timestamp
);
623 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
624 struct msghdr
*msg
, size_t size
)
626 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
628 sock_update_classid(sock
->sk
);
635 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
638 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
639 struct msghdr
*msg
, size_t size
)
641 int err
= security_socket_sendmsg(sock
, msg
, size
);
643 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
646 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
649 struct sock_iocb siocb
;
652 init_sync_kiocb(&iocb
, NULL
);
653 iocb
.private = &siocb
;
654 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
655 if (-EIOCBQUEUED
== ret
)
656 ret
= wait_on_sync_kiocb(&iocb
);
659 EXPORT_SYMBOL(sock_sendmsg
);
661 static int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
664 struct sock_iocb siocb
;
667 init_sync_kiocb(&iocb
, NULL
);
668 iocb
.private = &siocb
;
669 ret
= __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
);
670 if (-EIOCBQUEUED
== ret
)
671 ret
= wait_on_sync_kiocb(&iocb
);
675 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
676 struct kvec
*vec
, size_t num
, size_t size
)
678 mm_segment_t oldfs
= get_fs();
683 * the following is safe, since for compiler definitions of kvec and
684 * iovec are identical, yielding the same in-core layout and alignment
686 msg
->msg_iov
= (struct iovec
*)vec
;
687 msg
->msg_iovlen
= num
;
688 result
= sock_sendmsg(sock
, msg
, size
);
692 EXPORT_SYMBOL(kernel_sendmsg
);
694 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
697 *ts
= ktime_to_timespec(kt
);
705 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
707 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
710 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
711 struct timespec ts
[3];
713 struct skb_shared_hwtstamps
*shhwtstamps
=
716 /* Race occurred between timestamp enabling and packet
717 receiving. Fill in the current time for now. */
718 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
719 __net_timestamp(skb
);
721 if (need_software_tstamp
) {
722 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
724 skb_get_timestamp(skb
, &tv
);
725 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
728 skb_get_timestampns(skb
, &ts
[0]);
729 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
730 sizeof(ts
[0]), &ts
[0]);
735 memset(ts
, 0, sizeof(ts
));
736 if (skb
->tstamp
.tv64
&&
737 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
738 skb_get_timestampns(skb
, ts
+ 0);
742 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
743 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
745 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
746 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
750 put_cmsg(msg
, SOL_SOCKET
,
751 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
753 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
755 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
760 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
762 if (!skb
->wifi_acked_valid
)
765 ack
= skb
->wifi_acked
;
767 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
769 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
771 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
774 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
775 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
776 sizeof(__u32
), &skb
->dropcount
);
779 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
782 sock_recv_timestamp(msg
, sk
, skb
);
783 sock_recv_drops(msg
, sk
, skb
);
785 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
787 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
788 struct msghdr
*msg
, size_t size
, int flags
)
790 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
792 sock_update_classid(sock
->sk
);
800 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
803 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
804 struct msghdr
*msg
, size_t size
, int flags
)
806 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
808 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
811 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
812 size_t size
, int flags
)
815 struct sock_iocb siocb
;
818 init_sync_kiocb(&iocb
, NULL
);
819 iocb
.private = &siocb
;
820 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
821 if (-EIOCBQUEUED
== ret
)
822 ret
= wait_on_sync_kiocb(&iocb
);
825 EXPORT_SYMBOL(sock_recvmsg
);
827 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
828 size_t size
, int flags
)
831 struct sock_iocb siocb
;
834 init_sync_kiocb(&iocb
, NULL
);
835 iocb
.private = &siocb
;
836 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
837 if (-EIOCBQUEUED
== ret
)
838 ret
= wait_on_sync_kiocb(&iocb
);
843 * kernel_recvmsg - Receive a message from a socket (kernel space)
844 * @sock: The socket to receive the message from
845 * @msg: Received message
846 * @vec: Input s/g array for message data
847 * @num: Size of input s/g array
848 * @size: Number of bytes to read
849 * @flags: Message flags (MSG_DONTWAIT, etc...)
851 * On return the msg structure contains the scatter/gather array passed in the
852 * vec argument. The array is modified so that it consists of the unfilled
853 * portion of the original array.
855 * The returned value is the total number of bytes received, or an error.
857 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
858 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
860 mm_segment_t oldfs
= get_fs();
865 * the following is safe, since for compiler definitions of kvec and
866 * iovec are identical, yielding the same in-core layout and alignment
868 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
869 result
= sock_recvmsg(sock
, msg
, size
, flags
);
873 EXPORT_SYMBOL(kernel_recvmsg
);
875 static void sock_aio_dtor(struct kiocb
*iocb
)
877 kfree(iocb
->private);
880 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
881 int offset
, size_t size
, loff_t
*ppos
, int more
)
886 sock
= file
->private_data
;
888 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
889 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
892 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
895 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
896 struct pipe_inode_info
*pipe
, size_t len
,
899 struct socket
*sock
= file
->private_data
;
901 if (unlikely(!sock
->ops
->splice_read
))
904 sock_update_classid(sock
->sk
);
906 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
909 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
910 struct sock_iocb
*siocb
)
912 if (!is_sync_kiocb(iocb
)) {
913 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
916 iocb
->ki_dtor
= sock_aio_dtor
;
920 iocb
->private = siocb
;
924 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
925 struct file
*file
, const struct iovec
*iov
,
926 unsigned long nr_segs
)
928 struct socket
*sock
= file
->private_data
;
932 for (i
= 0; i
< nr_segs
; i
++)
933 size
+= iov
[i
].iov_len
;
935 msg
->msg_name
= NULL
;
936 msg
->msg_namelen
= 0;
937 msg
->msg_control
= NULL
;
938 msg
->msg_controllen
= 0;
939 msg
->msg_iov
= (struct iovec
*)iov
;
940 msg
->msg_iovlen
= nr_segs
;
941 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
943 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
946 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
947 unsigned long nr_segs
, loff_t pos
)
949 struct sock_iocb siocb
, *x
;
954 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
958 x
= alloc_sock_iocb(iocb
, &siocb
);
961 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
964 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
965 struct file
*file
, const struct iovec
*iov
,
966 unsigned long nr_segs
)
968 struct socket
*sock
= file
->private_data
;
972 for (i
= 0; i
< nr_segs
; i
++)
973 size
+= iov
[i
].iov_len
;
975 msg
->msg_name
= NULL
;
976 msg
->msg_namelen
= 0;
977 msg
->msg_control
= NULL
;
978 msg
->msg_controllen
= 0;
979 msg
->msg_iov
= (struct iovec
*)iov
;
980 msg
->msg_iovlen
= nr_segs
;
981 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
982 if (sock
->type
== SOCK_SEQPACKET
)
983 msg
->msg_flags
|= MSG_EOR
;
985 return __sock_sendmsg(iocb
, sock
, msg
, size
);
988 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
989 unsigned long nr_segs
, loff_t pos
)
991 struct sock_iocb siocb
, *x
;
996 x
= alloc_sock_iocb(iocb
, &siocb
);
1000 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
1004 * Atomic setting of ioctl hooks to avoid race
1005 * with module unload.
1008 static DEFINE_MUTEX(br_ioctl_mutex
);
1009 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
1011 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
1013 mutex_lock(&br_ioctl_mutex
);
1014 br_ioctl_hook
= hook
;
1015 mutex_unlock(&br_ioctl_mutex
);
1017 EXPORT_SYMBOL(brioctl_set
);
1019 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1020 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1022 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1024 mutex_lock(&vlan_ioctl_mutex
);
1025 vlan_ioctl_hook
= hook
;
1026 mutex_unlock(&vlan_ioctl_mutex
);
1028 EXPORT_SYMBOL(vlan_ioctl_set
);
1030 static DEFINE_MUTEX(dlci_ioctl_mutex
);
1031 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
1033 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
1035 mutex_lock(&dlci_ioctl_mutex
);
1036 dlci_ioctl_hook
= hook
;
1037 mutex_unlock(&dlci_ioctl_mutex
);
1039 EXPORT_SYMBOL(dlci_ioctl_set
);
1041 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1042 unsigned int cmd
, unsigned long arg
)
1045 void __user
*argp
= (void __user
*)arg
;
1047 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1050 * If this ioctl is unknown try to hand it down
1051 * to the NIC driver.
1053 if (err
== -ENOIOCTLCMD
)
1054 err
= dev_ioctl(net
, cmd
, argp
);
1060 * With an ioctl, arg may well be a user mode pointer, but we don't know
1061 * what to do with it - that's up to the protocol still.
1064 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1066 struct socket
*sock
;
1068 void __user
*argp
= (void __user
*)arg
;
1072 sock
= file
->private_data
;
1075 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
1076 err
= dev_ioctl(net
, cmd
, argp
);
1078 #ifdef CONFIG_WEXT_CORE
1079 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1080 err
= dev_ioctl(net
, cmd
, argp
);
1087 if (get_user(pid
, (int __user
*)argp
))
1089 err
= f_setown(sock
->file
, pid
, 1);
1093 err
= put_user(f_getown(sock
->file
),
1094 (int __user
*)argp
);
1102 request_module("bridge");
1104 mutex_lock(&br_ioctl_mutex
);
1106 err
= br_ioctl_hook(net
, cmd
, argp
);
1107 mutex_unlock(&br_ioctl_mutex
);
1112 if (!vlan_ioctl_hook
)
1113 request_module("8021q");
1115 mutex_lock(&vlan_ioctl_mutex
);
1116 if (vlan_ioctl_hook
)
1117 err
= vlan_ioctl_hook(net
, argp
);
1118 mutex_unlock(&vlan_ioctl_mutex
);
1123 if (!dlci_ioctl_hook
)
1124 request_module("dlci");
1126 mutex_lock(&dlci_ioctl_mutex
);
1127 if (dlci_ioctl_hook
)
1128 err
= dlci_ioctl_hook(cmd
, argp
);
1129 mutex_unlock(&dlci_ioctl_mutex
);
1132 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1138 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1141 struct socket
*sock
= NULL
;
1143 err
= security_socket_create(family
, type
, protocol
, 1);
1147 sock
= sock_alloc();
1154 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1166 EXPORT_SYMBOL(sock_create_lite
);
1168 /* No kernel lock held - perfect */
1169 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1171 struct socket
*sock
;
1174 * We can't return errors to poll, so it's either yes or no.
1176 sock
= file
->private_data
;
1177 return sock
->ops
->poll(file
, sock
, wait
);
1180 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1182 struct socket
*sock
= file
->private_data
;
1184 return sock
->ops
->mmap(file
, sock
, vma
);
1187 static int sock_close(struct inode
*inode
, struct file
*filp
)
1190 * It was possible the inode is NULL we were
1191 * closing an unfinished socket.
1195 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1198 sock_release(SOCKET_I(inode
));
1203 * Update the socket async list
1205 * Fasync_list locking strategy.
1207 * 1. fasync_list is modified only under process context socket lock
1208 * i.e. under semaphore.
1209 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1210 * or under socket lock
1213 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1215 struct socket
*sock
= filp
->private_data
;
1216 struct sock
*sk
= sock
->sk
;
1217 struct socket_wq
*wq
;
1223 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1224 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1226 if (!wq
->fasync_list
)
1227 sock_reset_flag(sk
, SOCK_FASYNC
);
1229 sock_set_flag(sk
, SOCK_FASYNC
);
1235 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1237 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1239 struct socket_wq
*wq
;
1244 wq
= rcu_dereference(sock
->wq
);
1245 if (!wq
|| !wq
->fasync_list
) {
1250 case SOCK_WAKE_WAITD
:
1251 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1254 case SOCK_WAKE_SPACE
:
1255 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1260 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1263 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1268 EXPORT_SYMBOL(sock_wake_async
);
1270 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1271 struct socket
**res
, int kern
)
1274 struct socket
*sock
;
1275 const struct net_proto_family
*pf
;
1278 * Check protocol is in range
1280 if (family
< 0 || family
>= NPROTO
)
1281 return -EAFNOSUPPORT
;
1282 if (type
< 0 || type
>= SOCK_MAX
)
1287 This uglymoron is moved from INET layer to here to avoid
1288 deadlock in module load.
1290 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1294 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1300 err
= security_socket_create(family
, type
, protocol
, kern
);
1305 * Allocate the socket and allow the family to set things up. if
1306 * the protocol is 0, the family is instructed to select an appropriate
1309 sock
= sock_alloc();
1311 net_warn_ratelimited("socket: no more sockets\n");
1312 return -ENFILE
; /* Not exactly a match, but its the
1313 closest posix thing */
1318 #ifdef CONFIG_MODULES
1319 /* Attempt to load a protocol module if the find failed.
1321 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1322 * requested real, full-featured networking support upon configuration.
1323 * Otherwise module support will break!
1325 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1326 request_module("net-pf-%d", family
);
1330 pf
= rcu_dereference(net_families
[family
]);
1331 err
= -EAFNOSUPPORT
;
1336 * We will call the ->create function, that possibly is in a loadable
1337 * module, so we have to bump that loadable module refcnt first.
1339 if (!try_module_get(pf
->owner
))
1342 /* Now protected by module ref count */
1345 err
= pf
->create(net
, sock
, protocol
, kern
);
1347 goto out_module_put
;
1350 * Now to bump the refcnt of the [loadable] module that owns this
1351 * socket at sock_release time we decrement its refcnt.
1353 if (!try_module_get(sock
->ops
->owner
))
1354 goto out_module_busy
;
1357 * Now that we're done with the ->create function, the [loadable]
1358 * module can have its refcnt decremented
1360 module_put(pf
->owner
);
1361 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1363 goto out_sock_release
;
1369 err
= -EAFNOSUPPORT
;
1372 module_put(pf
->owner
);
1379 goto out_sock_release
;
1381 EXPORT_SYMBOL(__sock_create
);
1383 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1385 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1387 EXPORT_SYMBOL(sock_create
);
1389 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1391 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1393 EXPORT_SYMBOL(sock_create_kern
);
1395 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1398 struct socket
*sock
;
1401 /* Check the SOCK_* constants for consistency. */
1402 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1403 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1404 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1405 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1407 flags
= type
& ~SOCK_TYPE_MASK
;
1408 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1410 type
&= SOCK_TYPE_MASK
;
1412 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1413 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1415 retval
= sock_create(family
, type
, protocol
, &sock
);
1419 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1424 /* It may be already another descriptor 8) Not kernel problem. */
1433 * Create a pair of connected sockets.
1436 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1437 int __user
*, usockvec
)
1439 struct socket
*sock1
, *sock2
;
1441 struct file
*newfile1
, *newfile2
;
1444 flags
= type
& ~SOCK_TYPE_MASK
;
1445 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1447 type
&= SOCK_TYPE_MASK
;
1449 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1450 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1453 * Obtain the first socket and check if the underlying protocol
1454 * supports the socketpair call.
1457 err
= sock_create(family
, type
, protocol
, &sock1
);
1461 err
= sock_create(family
, type
, protocol
, &sock2
);
1465 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1467 goto out_release_both
;
1469 fd1
= sock_alloc_file(sock1
, &newfile1
, flags
, NULL
);
1470 if (unlikely(fd1
< 0)) {
1472 goto out_release_both
;
1475 fd2
= sock_alloc_file(sock2
, &newfile2
, flags
, NULL
);
1476 if (unlikely(fd2
< 0)) {
1480 sock_release(sock2
);
1484 audit_fd_pair(fd1
, fd2
);
1485 fd_install(fd1
, newfile1
);
1486 fd_install(fd2
, newfile2
);
1487 /* fd1 and fd2 may be already another descriptors.
1488 * Not kernel problem.
1491 err
= put_user(fd1
, &usockvec
[0]);
1493 err
= put_user(fd2
, &usockvec
[1]);
1502 sock_release(sock2
);
1504 sock_release(sock1
);
1510 * Bind a name to a socket. Nothing much to do here since it's
1511 * the protocol's responsibility to handle the local address.
1513 * We move the socket address to kernel space before we call
1514 * the protocol layer (having also checked the address is ok).
1517 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1519 struct socket
*sock
;
1520 struct sockaddr_storage address
;
1521 int err
, fput_needed
;
1523 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1525 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1527 err
= security_socket_bind(sock
,
1528 (struct sockaddr
*)&address
,
1531 err
= sock
->ops
->bind(sock
,
1535 fput_light(sock
->file
, fput_needed
);
1541 * Perform a listen. Basically, we allow the protocol to do anything
1542 * necessary for a listen, and if that works, we mark the socket as
1543 * ready for listening.
1546 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1548 struct socket
*sock
;
1549 int err
, fput_needed
;
1552 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1554 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1555 if ((unsigned int)backlog
> somaxconn
)
1556 backlog
= somaxconn
;
1558 err
= security_socket_listen(sock
, backlog
);
1560 err
= sock
->ops
->listen(sock
, backlog
);
1562 fput_light(sock
->file
, fput_needed
);
1568 * For accept, we attempt to create a new socket, set up the link
1569 * with the client, wake up the client, then return the new
1570 * connected fd. We collect the address of the connector in kernel
1571 * space and move it to user at the very end. This is unclean because
1572 * we open the socket then return an error.
1574 * 1003.1g adds the ability to recvmsg() to query connection pending
1575 * status to recvmsg. We need to add that support in a way thats
1576 * clean when we restucture accept also.
1579 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1580 int __user
*, upeer_addrlen
, int, flags
)
1582 struct socket
*sock
, *newsock
;
1583 struct file
*newfile
;
1584 int err
, len
, newfd
, fput_needed
;
1585 struct sockaddr_storage address
;
1587 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1590 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1591 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1593 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1598 newsock
= sock_alloc();
1602 newsock
->type
= sock
->type
;
1603 newsock
->ops
= sock
->ops
;
1606 * We don't need try_module_get here, as the listening socket (sock)
1607 * has the protocol module (sock->ops->owner) held.
1609 __module_get(newsock
->ops
->owner
);
1611 newfd
= sock_alloc_file(newsock
, &newfile
, flags
,
1612 sock
->sk
->sk_prot_creator
->name
);
1613 if (unlikely(newfd
< 0)) {
1615 sock_release(newsock
);
1619 err
= security_socket_accept(sock
, newsock
);
1623 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1627 if (upeer_sockaddr
) {
1628 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1630 err
= -ECONNABORTED
;
1633 err
= move_addr_to_user(&address
,
1634 len
, upeer_sockaddr
, upeer_addrlen
);
1639 /* File flags are not inherited via accept() unlike another OSes. */
1641 fd_install(newfd
, newfile
);
1645 fput_light(sock
->file
, fput_needed
);
1650 put_unused_fd(newfd
);
1654 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1655 int __user
*, upeer_addrlen
)
1657 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1661 * Attempt to connect to a socket with the server address. The address
1662 * is in user space so we verify it is OK and move it to kernel space.
1664 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1667 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1668 * other SEQPACKET protocols that take time to connect() as it doesn't
1669 * include the -EINPROGRESS status for such sockets.
1672 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1675 struct socket
*sock
;
1676 struct sockaddr_storage address
;
1677 int err
, fput_needed
;
1679 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1682 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1687 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1691 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1692 sock
->file
->f_flags
);
1694 fput_light(sock
->file
, fput_needed
);
1700 * Get the local address ('name') of a socket object. Move the obtained
1701 * name to user space.
1704 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1705 int __user
*, usockaddr_len
)
1707 struct socket
*sock
;
1708 struct sockaddr_storage address
;
1709 int len
, err
, fput_needed
;
1711 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1715 err
= security_socket_getsockname(sock
);
1719 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1722 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1725 fput_light(sock
->file
, fput_needed
);
1731 * Get the remote address ('name') of a socket object. Move the obtained
1732 * name to user space.
1735 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1736 int __user
*, usockaddr_len
)
1738 struct socket
*sock
;
1739 struct sockaddr_storage address
;
1740 int len
, err
, fput_needed
;
1742 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1744 err
= security_socket_getpeername(sock
);
1746 fput_light(sock
->file
, fput_needed
);
1751 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1754 err
= move_addr_to_user(&address
, len
, usockaddr
,
1756 fput_light(sock
->file
, fput_needed
);
1762 * Send a datagram to a given address. We move the address into kernel
1763 * space and check the user space data area is readable before invoking
1767 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1768 unsigned int, flags
, struct sockaddr __user
*, addr
,
1771 struct socket
*sock
;
1772 struct sockaddr_storage address
;
1780 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1784 iov
.iov_base
= buff
;
1786 msg
.msg_name
= NULL
;
1789 msg
.msg_control
= NULL
;
1790 msg
.msg_controllen
= 0;
1791 msg
.msg_namelen
= 0;
1793 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1796 msg
.msg_name
= (struct sockaddr
*)&address
;
1797 msg
.msg_namelen
= addr_len
;
1799 if (sock
->file
->f_flags
& O_NONBLOCK
)
1800 flags
|= MSG_DONTWAIT
;
1801 msg
.msg_flags
= flags
;
1802 err
= sock_sendmsg(sock
, &msg
, len
);
1805 fput_light(sock
->file
, fput_needed
);
1811 * Send a datagram down a socket.
1814 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1815 unsigned int, flags
)
1817 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1821 * Receive a frame from the socket and optionally record the address of the
1822 * sender. We verify the buffers are writable and if needed move the
1823 * sender address from kernel to user space.
1826 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1827 unsigned int, flags
, struct sockaddr __user
*, addr
,
1828 int __user
*, addr_len
)
1830 struct socket
*sock
;
1833 struct sockaddr_storage address
;
1839 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1843 msg
.msg_control
= NULL
;
1844 msg
.msg_controllen
= 0;
1848 iov
.iov_base
= ubuf
;
1849 msg
.msg_name
= (struct sockaddr
*)&address
;
1850 msg
.msg_namelen
= sizeof(address
);
1851 if (sock
->file
->f_flags
& O_NONBLOCK
)
1852 flags
|= MSG_DONTWAIT
;
1853 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1855 if (err
>= 0 && addr
!= NULL
) {
1856 err2
= move_addr_to_user(&address
,
1857 msg
.msg_namelen
, addr
, addr_len
);
1862 fput_light(sock
->file
, fput_needed
);
1868 * Receive a datagram from a socket.
1871 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1874 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1878 * Set a socket option. Because we don't know the option lengths we have
1879 * to pass the user mode parameter for the protocols to sort out.
1882 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1883 char __user
*, optval
, int, optlen
)
1885 int err
, fput_needed
;
1886 struct socket
*sock
;
1891 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1893 err
= security_socket_setsockopt(sock
, level
, optname
);
1897 if (level
== SOL_SOCKET
)
1899 sock_setsockopt(sock
, level
, optname
, optval
,
1903 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1906 fput_light(sock
->file
, fput_needed
);
1912 * Get a socket option. Because we don't know the option lengths we have
1913 * to pass a user mode parameter for the protocols to sort out.
1916 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1917 char __user
*, optval
, int __user
*, optlen
)
1919 int err
, fput_needed
;
1920 struct socket
*sock
;
1922 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1924 err
= security_socket_getsockopt(sock
, level
, optname
);
1928 if (level
== SOL_SOCKET
)
1930 sock_getsockopt(sock
, level
, optname
, optval
,
1934 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1937 fput_light(sock
->file
, fput_needed
);
1943 * Shutdown a socket.
1946 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1948 int err
, fput_needed
;
1949 struct socket
*sock
;
1951 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1953 err
= security_socket_shutdown(sock
, how
);
1955 err
= sock
->ops
->shutdown(sock
, how
);
1956 fput_light(sock
->file
, fput_needed
);
1961 /* A couple of helpful macros for getting the address of the 32/64 bit
1962 * fields which are the same type (int / unsigned) on our platforms.
1964 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1965 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1966 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1968 struct used_address
{
1969 struct sockaddr_storage name
;
1970 unsigned int name_len
;
1973 static int __sys_sendmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1974 struct msghdr
*msg_sys
, unsigned int flags
,
1975 struct used_address
*used_address
)
1977 struct compat_msghdr __user
*msg_compat
=
1978 (struct compat_msghdr __user
*)msg
;
1979 struct sockaddr_storage address
;
1980 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1981 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1982 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1983 /* 20 is size of ipv6_pktinfo */
1984 unsigned char *ctl_buf
= ctl
;
1985 int err
, ctl_len
, total_len
;
1988 if (MSG_CMSG_COMPAT
& flags
) {
1989 if (get_compat_msghdr(msg_sys
, msg_compat
))
1991 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
1994 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
1996 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
1999 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2005 /* This will also move the address data into kernel space */
2006 if (MSG_CMSG_COMPAT
& flags
) {
2007 err
= verify_compat_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2009 err
= verify_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2016 if (msg_sys
->msg_controllen
> INT_MAX
)
2018 ctl_len
= msg_sys
->msg_controllen
;
2019 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2021 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2025 ctl_buf
= msg_sys
->msg_control
;
2026 ctl_len
= msg_sys
->msg_controllen
;
2027 } else if (ctl_len
) {
2028 if (ctl_len
> sizeof(ctl
)) {
2029 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2030 if (ctl_buf
== NULL
)
2035 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2036 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2037 * checking falls down on this.
2039 if (copy_from_user(ctl_buf
,
2040 (void __user __force
*)msg_sys
->msg_control
,
2043 msg_sys
->msg_control
= ctl_buf
;
2045 msg_sys
->msg_flags
= flags
;
2047 if (sock
->file
->f_flags
& O_NONBLOCK
)
2048 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2050 * If this is sendmmsg() and current destination address is same as
2051 * previously succeeded address, omit asking LSM's decision.
2052 * used_address->name_len is initialized to UINT_MAX so that the first
2053 * destination address never matches.
2055 if (used_address
&& msg_sys
->msg_name
&&
2056 used_address
->name_len
== msg_sys
->msg_namelen
&&
2057 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2058 used_address
->name_len
)) {
2059 err
= sock_sendmsg_nosec(sock
, msg_sys
, total_len
);
2062 err
= sock_sendmsg(sock
, msg_sys
, total_len
);
2064 * If this is sendmmsg() and sending to current destination address was
2065 * successful, remember it.
2067 if (used_address
&& err
>= 0) {
2068 used_address
->name_len
= msg_sys
->msg_namelen
;
2069 if (msg_sys
->msg_name
)
2070 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2071 used_address
->name_len
);
2076 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2078 if (iov
!= iovstack
)
2085 * BSD sendmsg interface
2088 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned int, flags
)
2090 int fput_needed
, err
;
2091 struct msghdr msg_sys
;
2092 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2097 err
= __sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
);
2099 fput_light(sock
->file
, fput_needed
);
2105 * Linux sendmmsg interface
2108 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2111 int fput_needed
, err
, datagrams
;
2112 struct socket
*sock
;
2113 struct mmsghdr __user
*entry
;
2114 struct compat_mmsghdr __user
*compat_entry
;
2115 struct msghdr msg_sys
;
2116 struct used_address used_address
;
2118 if (vlen
> UIO_MAXIOV
)
2123 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2127 used_address
.name_len
= UINT_MAX
;
2129 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2132 while (datagrams
< vlen
) {
2133 if (MSG_CMSG_COMPAT
& flags
) {
2134 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)compat_entry
,
2135 &msg_sys
, flags
, &used_address
);
2138 err
= __put_user(err
, &compat_entry
->msg_len
);
2141 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)entry
,
2142 &msg_sys
, flags
, &used_address
);
2145 err
= put_user(err
, &entry
->msg_len
);
2154 fput_light(sock
->file
, fput_needed
);
2156 /* We only return an error if no datagrams were able to be sent */
2163 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2164 unsigned int, vlen
, unsigned int, flags
)
2166 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2169 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
2170 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2172 struct compat_msghdr __user
*msg_compat
=
2173 (struct compat_msghdr __user
*)msg
;
2174 struct iovec iovstack
[UIO_FASTIOV
];
2175 struct iovec
*iov
= iovstack
;
2176 unsigned long cmsg_ptr
;
2177 int err
, total_len
, len
;
2179 /* kernel mode address */
2180 struct sockaddr_storage addr
;
2182 /* user mode address pointers */
2183 struct sockaddr __user
*uaddr
;
2184 int __user
*uaddr_len
;
2186 if (MSG_CMSG_COMPAT
& flags
) {
2187 if (get_compat_msghdr(msg_sys
, msg_compat
))
2189 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2192 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2194 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2197 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2204 * Save the user-mode address (verify_iovec will change the
2205 * kernel msghdr to use the kernel address space)
2208 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2209 uaddr_len
= COMPAT_NAMELEN(msg
);
2210 if (MSG_CMSG_COMPAT
& flags
) {
2211 err
= verify_compat_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2213 err
= verify_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2218 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2219 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2221 if (sock
->file
->f_flags
& O_NONBLOCK
)
2222 flags
|= MSG_DONTWAIT
;
2223 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2229 if (uaddr
!= NULL
) {
2230 err
= move_addr_to_user(&addr
,
2231 msg_sys
->msg_namelen
, uaddr
,
2236 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2240 if (MSG_CMSG_COMPAT
& flags
)
2241 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2242 &msg_compat
->msg_controllen
);
2244 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2245 &msg
->msg_controllen
);
2251 if (iov
!= iovstack
)
2258 * BSD recvmsg interface
2261 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2262 unsigned int, flags
)
2264 int fput_needed
, err
;
2265 struct msghdr msg_sys
;
2266 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2271 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2273 fput_light(sock
->file
, fput_needed
);
2279 * Linux recvmmsg interface
2282 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2283 unsigned int flags
, struct timespec
*timeout
)
2285 int fput_needed
, err
, datagrams
;
2286 struct socket
*sock
;
2287 struct mmsghdr __user
*entry
;
2288 struct compat_mmsghdr __user
*compat_entry
;
2289 struct msghdr msg_sys
;
2290 struct timespec end_time
;
2293 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2299 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2303 err
= sock_error(sock
->sk
);
2308 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2310 while (datagrams
< vlen
) {
2312 * No need to ask LSM for more than the first datagram.
2314 if (MSG_CMSG_COMPAT
& flags
) {
2315 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2316 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2320 err
= __put_user(err
, &compat_entry
->msg_len
);
2323 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2324 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2328 err
= put_user(err
, &entry
->msg_len
);
2336 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2337 if (flags
& MSG_WAITFORONE
)
2338 flags
|= MSG_DONTWAIT
;
2341 ktime_get_ts(timeout
);
2342 *timeout
= timespec_sub(end_time
, *timeout
);
2343 if (timeout
->tv_sec
< 0) {
2344 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2348 /* Timeout, return less than vlen datagrams */
2349 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2353 /* Out of band data, return right away */
2354 if (msg_sys
.msg_flags
& MSG_OOB
)
2359 fput_light(sock
->file
, fput_needed
);
2364 if (datagrams
!= 0) {
2366 * We may return less entries than requested (vlen) if the
2367 * sock is non block and there aren't enough datagrams...
2369 if (err
!= -EAGAIN
) {
2371 * ... or if recvmsg returns an error after we
2372 * received some datagrams, where we record the
2373 * error to return on the next call or if the
2374 * app asks about it using getsockopt(SO_ERROR).
2376 sock
->sk
->sk_err
= -err
;
2385 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2386 unsigned int, vlen
, unsigned int, flags
,
2387 struct timespec __user
*, timeout
)
2390 struct timespec timeout_sys
;
2393 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2395 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2398 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2400 if (datagrams
> 0 &&
2401 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2402 datagrams
= -EFAULT
;
2407 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2408 /* Argument list sizes for sys_socketcall */
2409 #define AL(x) ((x) * sizeof(unsigned long))
2410 static const unsigned char nargs
[21] = {
2411 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2412 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2413 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2420 * System call vectors.
2422 * Argument checking cleaned up. Saved 20% in size.
2423 * This function doesn't need to set the kernel lock because
2424 * it is set by the callees.
2427 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2430 unsigned long a0
, a1
;
2434 if (call
< 1 || call
> SYS_SENDMMSG
)
2438 if (len
> sizeof(a
))
2441 /* copy_from_user should be SMP safe. */
2442 if (copy_from_user(a
, args
, len
))
2445 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2452 err
= sys_socket(a0
, a1
, a
[2]);
2455 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2458 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2461 err
= sys_listen(a0
, a1
);
2464 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2465 (int __user
*)a
[2], 0);
2467 case SYS_GETSOCKNAME
:
2469 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2470 (int __user
*)a
[2]);
2472 case SYS_GETPEERNAME
:
2474 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2475 (int __user
*)a
[2]);
2477 case SYS_SOCKETPAIR
:
2478 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2481 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2484 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2485 (struct sockaddr __user
*)a
[4], a
[5]);
2488 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2491 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2492 (struct sockaddr __user
*)a
[4],
2493 (int __user
*)a
[5]);
2496 err
= sys_shutdown(a0
, a1
);
2498 case SYS_SETSOCKOPT
:
2499 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2501 case SYS_GETSOCKOPT
:
2503 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2504 (int __user
*)a
[4]);
2507 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2510 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2513 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2516 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2517 (struct timespec __user
*)a
[4]);
2520 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2521 (int __user
*)a
[2], a
[3]);
2530 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2533 * sock_register - add a socket protocol handler
2534 * @ops: description of protocol
2536 * This function is called by a protocol handler that wants to
2537 * advertise its address family, and have it linked into the
2538 * socket interface. The value ops->family coresponds to the
2539 * socket system call protocol family.
2541 int sock_register(const struct net_proto_family
*ops
)
2545 if (ops
->family
>= NPROTO
) {
2546 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2551 spin_lock(&net_family_lock
);
2552 if (rcu_dereference_protected(net_families
[ops
->family
],
2553 lockdep_is_held(&net_family_lock
)))
2556 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2559 spin_unlock(&net_family_lock
);
2561 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2564 EXPORT_SYMBOL(sock_register
);
2567 * sock_unregister - remove a protocol handler
2568 * @family: protocol family to remove
2570 * This function is called by a protocol handler that wants to
2571 * remove its address family, and have it unlinked from the
2572 * new socket creation.
2574 * If protocol handler is a module, then it can use module reference
2575 * counts to protect against new references. If protocol handler is not
2576 * a module then it needs to provide its own protection in
2577 * the ops->create routine.
2579 void sock_unregister(int family
)
2581 BUG_ON(family
< 0 || family
>= NPROTO
);
2583 spin_lock(&net_family_lock
);
2584 RCU_INIT_POINTER(net_families
[family
], NULL
);
2585 spin_unlock(&net_family_lock
);
2589 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2591 EXPORT_SYMBOL(sock_unregister
);
2593 static int __init
sock_init(void)
2597 * Initialize the network sysctl infrastructure.
2599 err
= net_sysctl_init();
2604 * Initialize sock SLAB cache.
2610 * Initialize skbuff SLAB cache
2615 * Initialize the protocols module.
2620 err
= register_filesystem(&sock_fs_type
);
2623 sock_mnt
= kern_mount(&sock_fs_type
);
2624 if (IS_ERR(sock_mnt
)) {
2625 err
= PTR_ERR(sock_mnt
);
2629 /* The real protocol initialization is performed in later initcalls.
2632 #ifdef CONFIG_NETFILTER
2636 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2637 skb_timestamping_init();
2644 unregister_filesystem(&sock_fs_type
);
2649 core_initcall(sock_init
); /* early initcall */
2651 #ifdef CONFIG_PROC_FS
2652 void socket_seq_show(struct seq_file
*seq
)
2657 for_each_possible_cpu(cpu
)
2658 counter
+= per_cpu(sockets_in_use
, cpu
);
2660 /* It can be negative, by the way. 8) */
2664 seq_printf(seq
, "sockets: used %d\n", counter
);
2666 #endif /* CONFIG_PROC_FS */
2668 #ifdef CONFIG_COMPAT
2669 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2670 unsigned int cmd
, void __user
*up
)
2672 mm_segment_t old_fs
= get_fs();
2677 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2680 err
= compat_put_timeval(&ktv
, up
);
2685 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2686 unsigned int cmd
, void __user
*up
)
2688 mm_segment_t old_fs
= get_fs();
2689 struct timespec kts
;
2693 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2696 err
= compat_put_timespec(&kts
, up
);
2701 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2703 struct ifreq __user
*uifr
;
2706 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2707 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2710 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2714 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2720 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2722 struct compat_ifconf ifc32
;
2724 struct ifconf __user
*uifc
;
2725 struct compat_ifreq __user
*ifr32
;
2726 struct ifreq __user
*ifr
;
2730 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2733 memset(&ifc
, 0, sizeof(ifc
));
2734 if (ifc32
.ifcbuf
== 0) {
2738 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2740 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2741 sizeof(struct ifreq
);
2742 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2744 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2745 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2746 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2747 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2753 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2756 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2760 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2764 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2766 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2767 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2768 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2774 if (ifc32
.ifcbuf
== 0) {
2775 /* Translate from 64-bit structure multiple to
2779 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2784 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2790 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2792 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2793 bool convert_in
= false, convert_out
= false;
2794 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2795 struct ethtool_rxnfc __user
*rxnfc
;
2796 struct ifreq __user
*ifr
;
2797 u32 rule_cnt
= 0, actual_rule_cnt
;
2802 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2805 compat_rxnfc
= compat_ptr(data
);
2807 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2810 /* Most ethtool structures are defined without padding.
2811 * Unfortunately struct ethtool_rxnfc is an exception.
2816 case ETHTOOL_GRXCLSRLALL
:
2817 /* Buffer size is variable */
2818 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2820 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2822 buf_size
+= rule_cnt
* sizeof(u32
);
2824 case ETHTOOL_GRXRINGS
:
2825 case ETHTOOL_GRXCLSRLCNT
:
2826 case ETHTOOL_GRXCLSRULE
:
2827 case ETHTOOL_SRXCLSRLINS
:
2830 case ETHTOOL_SRXCLSRLDEL
:
2831 buf_size
+= sizeof(struct ethtool_rxnfc
);
2836 ifr
= compat_alloc_user_space(buf_size
);
2837 rxnfc
= (void *)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2839 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2842 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2843 &ifr
->ifr_ifru
.ifru_data
))
2847 /* We expect there to be holes between fs.m_ext and
2848 * fs.ring_cookie and at the end of fs, but nowhere else.
2850 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2851 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2852 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2853 sizeof(rxnfc
->fs
.m_ext
));
2855 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2856 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2857 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2858 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2860 if (copy_in_user(rxnfc
, compat_rxnfc
,
2861 (void *)(&rxnfc
->fs
.m_ext
+ 1) -
2863 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2864 &compat_rxnfc
->fs
.ring_cookie
,
2865 (void *)(&rxnfc
->fs
.location
+ 1) -
2866 (void *)&rxnfc
->fs
.ring_cookie
) ||
2867 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2868 sizeof(rxnfc
->rule_cnt
)))
2872 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2877 if (copy_in_user(compat_rxnfc
, rxnfc
,
2878 (const void *)(&rxnfc
->fs
.m_ext
+ 1) -
2879 (const void *)rxnfc
) ||
2880 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2881 &rxnfc
->fs
.ring_cookie
,
2882 (const void *)(&rxnfc
->fs
.location
+ 1) -
2883 (const void *)&rxnfc
->fs
.ring_cookie
) ||
2884 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2885 sizeof(rxnfc
->rule_cnt
)))
2888 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2889 /* As an optimisation, we only copy the actual
2890 * number of rules that the underlying
2891 * function returned. Since Mallory might
2892 * change the rule count in user memory, we
2893 * check that it is less than the rule count
2894 * originally given (as the user buffer size),
2895 * which has been range-checked.
2897 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2899 if (actual_rule_cnt
< rule_cnt
)
2900 rule_cnt
= actual_rule_cnt
;
2901 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2902 &rxnfc
->rule_locs
[0],
2903 rule_cnt
* sizeof(u32
)))
2911 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2914 compat_uptr_t uptr32
;
2915 struct ifreq __user
*uifr
;
2917 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2918 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2921 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2924 uptr
= compat_ptr(uptr32
);
2926 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2929 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2932 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2933 struct compat_ifreq __user
*ifr32
)
2936 struct ifreq __user
*uifr
;
2937 mm_segment_t old_fs
;
2943 case SIOCBONDENSLAVE
:
2944 case SIOCBONDRELEASE
:
2945 case SIOCBONDSETHWADDR
:
2946 case SIOCBONDCHANGEACTIVE
:
2947 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2952 err
= dev_ioctl(net
, cmd
,
2953 (struct ifreq __user __force
*) &kifr
);
2957 case SIOCBONDSLAVEINFOQUERY
:
2958 case SIOCBONDINFOQUERY
:
2959 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2960 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2963 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2966 datap
= compat_ptr(data
);
2967 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2970 return dev_ioctl(net
, cmd
, uifr
);
2972 return -ENOIOCTLCMD
;
2976 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2977 struct compat_ifreq __user
*u_ifreq32
)
2979 struct ifreq __user
*u_ifreq64
;
2980 char tmp_buf
[IFNAMSIZ
];
2981 void __user
*data64
;
2984 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2987 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2989 data64
= compat_ptr(data32
);
2991 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2993 /* Don't check these user accesses, just let that get trapped
2994 * in the ioctl handler instead.
2996 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2999 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
3002 return dev_ioctl(net
, cmd
, u_ifreq64
);
3005 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
3006 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
3008 struct ifreq __user
*uifr
;
3011 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3012 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
3015 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3026 case SIOCGIFBRDADDR
:
3027 case SIOCGIFDSTADDR
:
3028 case SIOCGIFNETMASK
:
3033 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3041 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3042 struct compat_ifreq __user
*uifr32
)
3045 struct compat_ifmap __user
*uifmap32
;
3046 mm_segment_t old_fs
;
3049 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3050 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3051 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3052 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3053 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3054 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3055 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3056 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3062 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3065 if (cmd
== SIOCGIFMAP
&& !err
) {
3066 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3067 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3068 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3069 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3070 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3071 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3072 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3079 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3082 compat_uptr_t uptr32
;
3083 struct ifreq __user
*uifr
;
3085 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3086 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
3089 if (get_user(uptr32
, &uifr32
->ifr_data
))
3092 uptr
= compat_ptr(uptr32
);
3094 if (put_user(uptr
, &uifr
->ifr_data
))
3097 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
3102 struct sockaddr rt_dst
; /* target address */
3103 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3104 struct sockaddr rt_genmask
; /* target network mask (IP) */
3105 unsigned short rt_flags
;
3108 unsigned char rt_tos
;
3109 unsigned char rt_class
;
3111 short rt_metric
; /* +1 for binary compatibility! */
3112 /* char * */ u32 rt_dev
; /* forcing the device at add */
3113 u32 rt_mtu
; /* per route MTU/Window */
3114 u32 rt_window
; /* Window clamping */
3115 unsigned short rt_irtt
; /* Initial RTT */
3118 struct in6_rtmsg32
{
3119 struct in6_addr rtmsg_dst
;
3120 struct in6_addr rtmsg_src
;
3121 struct in6_addr rtmsg_gateway
;
3131 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3132 unsigned int cmd
, void __user
*argp
)
3136 struct in6_rtmsg r6
;
3140 mm_segment_t old_fs
= get_fs();
3142 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3143 struct in6_rtmsg32 __user
*ur6
= argp
;
3144 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3145 3 * sizeof(struct in6_addr
));
3146 ret
|= __get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3147 ret
|= __get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3148 ret
|= __get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3149 ret
|= __get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3150 ret
|= __get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3151 ret
|= __get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3152 ret
|= __get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3156 struct rtentry32 __user
*ur4
= argp
;
3157 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3158 3 * sizeof(struct sockaddr
));
3159 ret
|= __get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3160 ret
|= __get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3161 ret
|= __get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3162 ret
|= __get_user(r4
.rt_window
, &(ur4
->rt_window
));
3163 ret
|= __get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3164 ret
|= __get_user(rtdev
, &(ur4
->rt_dev
));
3166 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3167 r4
.rt_dev
= (char __user __force
*)devname
;
3181 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3188 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3189 * for some operations; this forces use of the newer bridge-utils that
3190 * use compatible ioctls
3192 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3196 if (get_user(tmp
, argp
))
3198 if (tmp
== BRCTL_GET_VERSION
)
3199 return BRCTL_VERSION
+ 1;
3203 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3204 unsigned int cmd
, unsigned long arg
)
3206 void __user
*argp
= compat_ptr(arg
);
3207 struct sock
*sk
= sock
->sk
;
3208 struct net
*net
= sock_net(sk
);
3210 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3211 return siocdevprivate_ioctl(net
, cmd
, argp
);
3216 return old_bridge_ioctl(argp
);
3218 return dev_ifname32(net
, argp
);
3220 return dev_ifconf(net
, argp
);
3222 return ethtool_ioctl(net
, argp
);
3224 return compat_siocwandev(net
, argp
);
3227 return compat_sioc_ifmap(net
, cmd
, argp
);
3228 case SIOCBONDENSLAVE
:
3229 case SIOCBONDRELEASE
:
3230 case SIOCBONDSETHWADDR
:
3231 case SIOCBONDSLAVEINFOQUERY
:
3232 case SIOCBONDINFOQUERY
:
3233 case SIOCBONDCHANGEACTIVE
:
3234 return bond_ioctl(net
, cmd
, argp
);
3237 return routing_ioctl(net
, sock
, cmd
, argp
);
3239 return do_siocgstamp(net
, sock
, cmd
, argp
);
3241 return do_siocgstampns(net
, sock
, cmd
, argp
);
3243 return compat_siocshwtstamp(net
, argp
);
3255 return sock_ioctl(file
, cmd
, arg
);
3272 case SIOCSIFHWBROADCAST
:
3274 case SIOCGIFBRDADDR
:
3275 case SIOCSIFBRDADDR
:
3276 case SIOCGIFDSTADDR
:
3277 case SIOCSIFDSTADDR
:
3278 case SIOCGIFNETMASK
:
3279 case SIOCSIFNETMASK
:
3290 return dev_ifsioc(net
, sock
, cmd
, argp
);
3296 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3299 return -ENOIOCTLCMD
;
3302 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3305 struct socket
*sock
= file
->private_data
;
3306 int ret
= -ENOIOCTLCMD
;
3313 if (sock
->ops
->compat_ioctl
)
3314 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3316 if (ret
== -ENOIOCTLCMD
&&
3317 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3318 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3320 if (ret
== -ENOIOCTLCMD
)
3321 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3327 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3329 return sock
->ops
->bind(sock
, addr
, addrlen
);
3331 EXPORT_SYMBOL(kernel_bind
);
3333 int kernel_listen(struct socket
*sock
, int backlog
)
3335 return sock
->ops
->listen(sock
, backlog
);
3337 EXPORT_SYMBOL(kernel_listen
);
3339 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3341 struct sock
*sk
= sock
->sk
;
3344 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3349 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3351 sock_release(*newsock
);
3356 (*newsock
)->ops
= sock
->ops
;
3357 __module_get((*newsock
)->ops
->owner
);
3362 EXPORT_SYMBOL(kernel_accept
);
3364 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3367 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3369 EXPORT_SYMBOL(kernel_connect
);
3371 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3374 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3376 EXPORT_SYMBOL(kernel_getsockname
);
3378 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3381 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3383 EXPORT_SYMBOL(kernel_getpeername
);
3385 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3386 char *optval
, int *optlen
)
3388 mm_segment_t oldfs
= get_fs();
3389 char __user
*uoptval
;
3390 int __user
*uoptlen
;
3393 uoptval
= (char __user __force
*) optval
;
3394 uoptlen
= (int __user __force
*) optlen
;
3397 if (level
== SOL_SOCKET
)
3398 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3400 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3405 EXPORT_SYMBOL(kernel_getsockopt
);
3407 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3408 char *optval
, unsigned int optlen
)
3410 mm_segment_t oldfs
= get_fs();
3411 char __user
*uoptval
;
3414 uoptval
= (char __user __force
*) optval
;
3417 if (level
== SOL_SOCKET
)
3418 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3420 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3425 EXPORT_SYMBOL(kernel_setsockopt
);
3427 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3428 size_t size
, int flags
)
3430 sock_update_classid(sock
->sk
);
3432 if (sock
->ops
->sendpage
)
3433 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3435 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3437 EXPORT_SYMBOL(kernel_sendpage
);
3439 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3441 mm_segment_t oldfs
= get_fs();
3445 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3450 EXPORT_SYMBOL(kernel_sock_ioctl
);
3452 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3454 return sock
->ops
->shutdown(sock
, how
);
3456 EXPORT_SYMBOL(kernel_sock_shutdown
);