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/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
87 #include <linux/nsproxy.h>
88 #include <linux/magic.h>
89 #include <linux/slab.h>
90 #include <linux/xattr.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_storage
*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_storage
*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
;
266 static void sock_destroy_inode(struct inode
*inode
)
268 struct socket_alloc
*ei
;
269 struct socket_wq
*wq
;
271 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
272 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
274 kmem_cache_free(sock_inode_cachep
, ei
);
277 static void init_once(void *foo
)
279 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
281 inode_init_once(&ei
->vfs_inode
);
284 static int init_inodecache(void)
286 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
287 sizeof(struct socket_alloc
),
289 (SLAB_HWCACHE_ALIGN
|
290 SLAB_RECLAIM_ACCOUNT
|
293 if (sock_inode_cachep
== NULL
)
298 static const struct super_operations sockfs_ops
= {
299 .alloc_inode
= sock_alloc_inode
,
300 .destroy_inode
= sock_destroy_inode
,
301 .statfs
= simple_statfs
,
305 * sockfs_dname() is called from d_path().
307 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
309 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
310 dentry
->d_inode
->i_ino
);
313 static const struct dentry_operations sockfs_dentry_operations
= {
314 .d_dname
= sockfs_dname
,
317 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
318 int flags
, const char *dev_name
, void *data
)
320 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
321 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
324 static struct vfsmount
*sock_mnt __read_mostly
;
326 static struct file_system_type sock_fs_type
= {
328 .mount
= sockfs_mount
,
329 .kill_sb
= kill_anon_super
,
333 * Obtains the first available file descriptor and sets it up for use.
335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
349 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
351 struct qstr name
= { .name
= "" };
357 name
.len
= strlen(name
.name
);
358 } else if (sock
->sk
) {
359 name
.name
= sock
->sk
->sk_prot_creator
->name
;
360 name
.len
= strlen(name
.name
);
362 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
363 if (unlikely(!path
.dentry
))
364 return ERR_PTR(-ENOMEM
);
365 path
.mnt
= mntget(sock_mnt
);
367 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
368 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
370 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
372 if (unlikely(!file
)) {
373 /* drop dentry, keep inode */
374 ihold(path
.dentry
->d_inode
);
376 return ERR_PTR(-ENFILE
);
380 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
382 file
->private_data
= sock
;
385 EXPORT_SYMBOL(sock_alloc_file
);
387 static int sock_map_fd(struct socket
*sock
, int flags
)
389 struct file
*newfile
;
390 int fd
= get_unused_fd_flags(flags
);
391 if (unlikely(fd
< 0))
394 newfile
= sock_alloc_file(sock
, flags
, NULL
);
395 if (likely(!IS_ERR(newfile
))) {
396 fd_install(fd
, newfile
);
401 return PTR_ERR(newfile
);
404 struct socket
*sock_from_file(struct file
*file
, int *err
)
406 if (file
->f_op
== &socket_file_ops
)
407 return file
->private_data
; /* set in sock_map_fd */
412 EXPORT_SYMBOL(sock_from_file
);
415 * sockfd_lookup - Go from a file number to its socket slot
417 * @err: pointer to an error code return
419 * The file handle passed in is locked and the socket it is bound
420 * too is returned. If an error occurs the err pointer is overwritten
421 * with a negative errno code and NULL is returned. The function checks
422 * for both invalid handles and passing a handle which is not a socket.
424 * On a success the socket object pointer is returned.
427 struct socket
*sockfd_lookup(int fd
, int *err
)
438 sock
= sock_from_file(file
, err
);
443 EXPORT_SYMBOL(sockfd_lookup
);
445 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
451 file
= fget_light(fd
, fput_needed
);
453 sock
= sock_from_file(file
, err
);
456 fput_light(file
, *fput_needed
);
461 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
462 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
463 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
464 static ssize_t
sockfs_getxattr(struct dentry
*dentry
,
465 const char *name
, void *value
, size_t size
)
467 const char *proto_name
;
472 if (!strncmp(name
, XATTR_NAME_SOCKPROTONAME
, XATTR_NAME_SOCKPROTONAME_LEN
)) {
473 proto_name
= dentry
->d_name
.name
;
474 proto_size
= strlen(proto_name
);
478 if (proto_size
+ 1 > size
)
481 strncpy(value
, proto_name
, proto_size
+ 1);
483 error
= proto_size
+ 1;
490 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
496 len
= security_inode_listsecurity(dentry
->d_inode
, buffer
, size
);
506 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
511 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
518 static const struct inode_operations sockfs_inode_ops
= {
519 .getxattr
= sockfs_getxattr
,
520 .listxattr
= sockfs_listxattr
,
524 * sock_alloc - allocate a socket
526 * Allocate a new inode and socket object. The two are bound together
527 * and initialised. The socket is then returned. If we are out of inodes
531 static struct socket
*sock_alloc(void)
536 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
540 sock
= SOCKET_I(inode
);
542 kmemcheck_annotate_bitfield(sock
, type
);
543 inode
->i_ino
= get_next_ino();
544 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
545 inode
->i_uid
= current_fsuid();
546 inode
->i_gid
= current_fsgid();
547 inode
->i_op
= &sockfs_inode_ops
;
549 this_cpu_add(sockets_in_use
, 1);
554 * In theory you can't get an open on this inode, but /proc provides
555 * a back door. Remember to keep it shut otherwise you'll let the
556 * creepy crawlies in.
559 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
564 const struct file_operations bad_sock_fops
= {
565 .owner
= THIS_MODULE
,
566 .open
= sock_no_open
,
567 .llseek
= noop_llseek
,
571 * sock_release - close a socket
572 * @sock: socket to close
574 * The socket is released from the protocol stack if it has a release
575 * callback, and the inode is then released if the socket is bound to
576 * an inode not a file.
579 void sock_release(struct socket
*sock
)
582 struct module
*owner
= sock
->ops
->owner
;
584 sock
->ops
->release(sock
);
589 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
590 printk(KERN_ERR
"sock_release: fasync list not empty!\n");
592 if (test_bit(SOCK_EXTERNALLY_ALLOCATED
, &sock
->flags
))
595 this_cpu_sub(sockets_in_use
, 1);
597 iput(SOCK_INODE(sock
));
602 EXPORT_SYMBOL(sock_release
);
604 int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
)
607 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
608 *tx_flags
|= SKBTX_HW_TSTAMP
;
609 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
610 *tx_flags
|= SKBTX_SW_TSTAMP
;
611 if (sock_flag(sk
, SOCK_WIFI_STATUS
))
612 *tx_flags
|= SKBTX_WIFI_STATUS
;
615 EXPORT_SYMBOL(sock_tx_timestamp
);
617 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
618 struct msghdr
*msg
, size_t size
)
620 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
627 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
630 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
631 struct msghdr
*msg
, size_t size
)
633 int err
= security_socket_sendmsg(sock
, msg
, size
);
635 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
638 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
641 struct sock_iocb siocb
;
644 init_sync_kiocb(&iocb
, NULL
);
645 iocb
.private = &siocb
;
646 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
647 if (-EIOCBQUEUED
== ret
)
648 ret
= wait_on_sync_kiocb(&iocb
);
651 EXPORT_SYMBOL(sock_sendmsg
);
653 static int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
656 struct sock_iocb siocb
;
659 init_sync_kiocb(&iocb
, NULL
);
660 iocb
.private = &siocb
;
661 ret
= __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
);
662 if (-EIOCBQUEUED
== ret
)
663 ret
= wait_on_sync_kiocb(&iocb
);
667 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
668 struct kvec
*vec
, size_t num
, size_t size
)
670 mm_segment_t oldfs
= get_fs();
675 * the following is safe, since for compiler definitions of kvec and
676 * iovec are identical, yielding the same in-core layout and alignment
678 msg
->msg_iov
= (struct iovec
*)vec
;
679 msg
->msg_iovlen
= num
;
680 result
= sock_sendmsg(sock
, msg
, size
);
684 EXPORT_SYMBOL(kernel_sendmsg
);
686 static int ktime2ts(ktime_t kt
, struct timespec
*ts
)
689 *ts
= ktime_to_timespec(kt
);
697 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
699 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
702 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
703 struct timespec ts
[3];
705 struct skb_shared_hwtstamps
*shhwtstamps
=
708 /* Race occurred between timestamp enabling and packet
709 receiving. Fill in the current time for now. */
710 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
711 __net_timestamp(skb
);
713 if (need_software_tstamp
) {
714 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
716 skb_get_timestamp(skb
, &tv
);
717 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
720 skb_get_timestampns(skb
, &ts
[0]);
721 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
722 sizeof(ts
[0]), &ts
[0]);
727 memset(ts
, 0, sizeof(ts
));
728 if (skb
->tstamp
.tv64
&&
729 sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) {
730 skb_get_timestampns(skb
, ts
+ 0);
734 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
735 ktime2ts(shhwtstamps
->syststamp
, ts
+ 1))
737 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
738 ktime2ts(shhwtstamps
->hwtstamp
, ts
+ 2))
742 put_cmsg(msg
, SOL_SOCKET
,
743 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
745 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
747 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
752 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
754 if (!skb
->wifi_acked_valid
)
757 ack
= skb
->wifi_acked
;
759 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
761 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
763 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
766 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
767 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
768 sizeof(__u32
), &skb
->dropcount
);
771 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
774 sock_recv_timestamp(msg
, sk
, skb
);
775 sock_recv_drops(msg
, sk
, skb
);
777 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
779 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
780 struct msghdr
*msg
, size_t size
, int flags
)
782 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
790 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
793 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
794 struct msghdr
*msg
, size_t size
, int flags
)
796 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
798 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
801 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
802 size_t size
, int flags
)
805 struct sock_iocb siocb
;
808 init_sync_kiocb(&iocb
, NULL
);
809 iocb
.private = &siocb
;
810 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
811 if (-EIOCBQUEUED
== ret
)
812 ret
= wait_on_sync_kiocb(&iocb
);
815 EXPORT_SYMBOL(sock_recvmsg
);
817 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
818 size_t size
, int flags
)
821 struct sock_iocb siocb
;
824 init_sync_kiocb(&iocb
, NULL
);
825 iocb
.private = &siocb
;
826 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
827 if (-EIOCBQUEUED
== ret
)
828 ret
= wait_on_sync_kiocb(&iocb
);
833 * kernel_recvmsg - Receive a message from a socket (kernel space)
834 * @sock: The socket to receive the message from
835 * @msg: Received message
836 * @vec: Input s/g array for message data
837 * @num: Size of input s/g array
838 * @size: Number of bytes to read
839 * @flags: Message flags (MSG_DONTWAIT, etc...)
841 * On return the msg structure contains the scatter/gather array passed in the
842 * vec argument. The array is modified so that it consists of the unfilled
843 * portion of the original array.
845 * The returned value is the total number of bytes received, or an error.
847 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
848 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
850 mm_segment_t oldfs
= get_fs();
855 * the following is safe, since for compiler definitions of kvec and
856 * iovec are identical, yielding the same in-core layout and alignment
858 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
859 result
= sock_recvmsg(sock
, msg
, size
, flags
);
863 EXPORT_SYMBOL(kernel_recvmsg
);
865 static void sock_aio_dtor(struct kiocb
*iocb
)
867 kfree(iocb
->private);
870 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
871 int offset
, size_t size
, loff_t
*ppos
, int more
)
876 sock
= file
->private_data
;
878 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
879 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
882 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
885 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
886 struct pipe_inode_info
*pipe
, size_t len
,
889 struct socket
*sock
= file
->private_data
;
891 if (unlikely(!sock
->ops
->splice_read
))
894 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
897 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
898 struct sock_iocb
*siocb
)
900 if (!is_sync_kiocb(iocb
)) {
901 siocb
= kmalloc(sizeof(*siocb
), GFP_KERNEL
);
904 iocb
->ki_dtor
= sock_aio_dtor
;
908 iocb
->private = siocb
;
912 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
913 struct file
*file
, const struct iovec
*iov
,
914 unsigned long nr_segs
)
916 struct socket
*sock
= file
->private_data
;
920 for (i
= 0; i
< nr_segs
; i
++)
921 size
+= iov
[i
].iov_len
;
923 msg
->msg_name
= NULL
;
924 msg
->msg_namelen
= 0;
925 msg
->msg_control
= NULL
;
926 msg
->msg_controllen
= 0;
927 msg
->msg_iov
= (struct iovec
*)iov
;
928 msg
->msg_iovlen
= nr_segs
;
929 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
931 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
934 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
935 unsigned long nr_segs
, loff_t pos
)
937 struct sock_iocb siocb
, *x
;
942 if (iocb
->ki_left
== 0) /* Match SYS5 behaviour */
946 x
= alloc_sock_iocb(iocb
, &siocb
);
949 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
952 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
953 struct file
*file
, const struct iovec
*iov
,
954 unsigned long nr_segs
)
956 struct socket
*sock
= file
->private_data
;
960 for (i
= 0; i
< nr_segs
; i
++)
961 size
+= iov
[i
].iov_len
;
963 msg
->msg_name
= NULL
;
964 msg
->msg_namelen
= 0;
965 msg
->msg_control
= NULL
;
966 msg
->msg_controllen
= 0;
967 msg
->msg_iov
= (struct iovec
*)iov
;
968 msg
->msg_iovlen
= nr_segs
;
969 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
970 if (sock
->type
== SOCK_SEQPACKET
)
971 msg
->msg_flags
|= MSG_EOR
;
973 return __sock_sendmsg(iocb
, sock
, msg
, size
);
976 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
977 unsigned long nr_segs
, loff_t pos
)
979 struct sock_iocb siocb
, *x
;
984 x
= alloc_sock_iocb(iocb
, &siocb
);
988 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
992 * Atomic setting of ioctl hooks to avoid race
993 * with module unload.
996 static DEFINE_MUTEX(br_ioctl_mutex
);
997 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
999 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
1001 mutex_lock(&br_ioctl_mutex
);
1002 br_ioctl_hook
= hook
;
1003 mutex_unlock(&br_ioctl_mutex
);
1005 EXPORT_SYMBOL(brioctl_set
);
1007 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1008 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1010 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1012 mutex_lock(&vlan_ioctl_mutex
);
1013 vlan_ioctl_hook
= hook
;
1014 mutex_unlock(&vlan_ioctl_mutex
);
1016 EXPORT_SYMBOL(vlan_ioctl_set
);
1018 static DEFINE_MUTEX(dlci_ioctl_mutex
);
1019 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
1021 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
1023 mutex_lock(&dlci_ioctl_mutex
);
1024 dlci_ioctl_hook
= hook
;
1025 mutex_unlock(&dlci_ioctl_mutex
);
1027 EXPORT_SYMBOL(dlci_ioctl_set
);
1029 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1030 unsigned int cmd
, unsigned long arg
)
1033 void __user
*argp
= (void __user
*)arg
;
1035 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1038 * If this ioctl is unknown try to hand it down
1039 * to the NIC driver.
1041 if (err
== -ENOIOCTLCMD
)
1042 err
= dev_ioctl(net
, cmd
, argp
);
1048 * With an ioctl, arg may well be a user mode pointer, but we don't know
1049 * what to do with it - that's up to the protocol still.
1052 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1054 struct socket
*sock
;
1056 void __user
*argp
= (void __user
*)arg
;
1060 sock
= file
->private_data
;
1063 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
1064 err
= dev_ioctl(net
, cmd
, argp
);
1066 #ifdef CONFIG_WEXT_CORE
1067 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1068 err
= dev_ioctl(net
, cmd
, argp
);
1075 if (get_user(pid
, (int __user
*)argp
))
1077 err
= f_setown(sock
->file
, pid
, 1);
1081 err
= put_user(f_getown(sock
->file
),
1082 (int __user
*)argp
);
1090 request_module("bridge");
1092 mutex_lock(&br_ioctl_mutex
);
1094 err
= br_ioctl_hook(net
, cmd
, argp
);
1095 mutex_unlock(&br_ioctl_mutex
);
1100 if (!vlan_ioctl_hook
)
1101 request_module("8021q");
1103 mutex_lock(&vlan_ioctl_mutex
);
1104 if (vlan_ioctl_hook
)
1105 err
= vlan_ioctl_hook(net
, argp
);
1106 mutex_unlock(&vlan_ioctl_mutex
);
1111 if (!dlci_ioctl_hook
)
1112 request_module("dlci");
1114 mutex_lock(&dlci_ioctl_mutex
);
1115 if (dlci_ioctl_hook
)
1116 err
= dlci_ioctl_hook(cmd
, argp
);
1117 mutex_unlock(&dlci_ioctl_mutex
);
1120 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1126 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1129 struct socket
*sock
= NULL
;
1131 err
= security_socket_create(family
, type
, protocol
, 1);
1135 sock
= sock_alloc();
1142 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1154 EXPORT_SYMBOL(sock_create_lite
);
1156 /* No kernel lock held - perfect */
1157 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1159 struct socket
*sock
;
1162 * We can't return errors to poll, so it's either yes or no.
1164 sock
= file
->private_data
;
1165 return sock
->ops
->poll(file
, sock
, wait
);
1168 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1170 struct socket
*sock
= file
->private_data
;
1172 return sock
->ops
->mmap(file
, sock
, vma
);
1175 static int sock_close(struct inode
*inode
, struct file
*filp
)
1178 * It was possible the inode is NULL we were
1179 * closing an unfinished socket.
1183 printk(KERN_DEBUG
"sock_close: NULL inode\n");
1186 sock_release(SOCKET_I(inode
));
1191 * Update the socket async list
1193 * Fasync_list locking strategy.
1195 * 1. fasync_list is modified only under process context socket lock
1196 * i.e. under semaphore.
1197 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1198 * or under socket lock
1201 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1203 struct socket
*sock
= filp
->private_data
;
1204 struct sock
*sk
= sock
->sk
;
1205 struct socket_wq
*wq
;
1211 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1212 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1214 if (!wq
->fasync_list
)
1215 sock_reset_flag(sk
, SOCK_FASYNC
);
1217 sock_set_flag(sk
, SOCK_FASYNC
);
1223 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1225 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1227 struct socket_wq
*wq
;
1232 wq
= rcu_dereference(sock
->wq
);
1233 if (!wq
|| !wq
->fasync_list
) {
1238 case SOCK_WAKE_WAITD
:
1239 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1242 case SOCK_WAKE_SPACE
:
1243 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1248 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1251 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1256 EXPORT_SYMBOL(sock_wake_async
);
1258 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1259 struct socket
**res
, int kern
)
1262 struct socket
*sock
;
1263 const struct net_proto_family
*pf
;
1266 * Check protocol is in range
1268 if (family
< 0 || family
>= NPROTO
)
1269 return -EAFNOSUPPORT
;
1270 if (type
< 0 || type
>= SOCK_MAX
)
1275 This uglymoron is moved from INET layer to here to avoid
1276 deadlock in module load.
1278 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1282 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1288 err
= security_socket_create(family
, type
, protocol
, kern
);
1293 * Allocate the socket and allow the family to set things up. if
1294 * the protocol is 0, the family is instructed to select an appropriate
1297 sock
= sock_alloc();
1299 net_warn_ratelimited("socket: no more sockets\n");
1300 return -ENFILE
; /* Not exactly a match, but its the
1301 closest posix thing */
1306 #ifdef CONFIG_MODULES
1307 /* Attempt to load a protocol module if the find failed.
1309 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1310 * requested real, full-featured networking support upon configuration.
1311 * Otherwise module support will break!
1313 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1314 request_module("net-pf-%d", family
);
1318 pf
= rcu_dereference(net_families
[family
]);
1319 err
= -EAFNOSUPPORT
;
1324 * We will call the ->create function, that possibly is in a loadable
1325 * module, so we have to bump that loadable module refcnt first.
1327 if (!try_module_get(pf
->owner
))
1330 /* Now protected by module ref count */
1333 err
= pf
->create(net
, sock
, protocol
, kern
);
1335 goto out_module_put
;
1338 * Now to bump the refcnt of the [loadable] module that owns this
1339 * socket at sock_release time we decrement its refcnt.
1341 if (!try_module_get(sock
->ops
->owner
))
1342 goto out_module_busy
;
1345 * Now that we're done with the ->create function, the [loadable]
1346 * module can have its refcnt decremented
1348 module_put(pf
->owner
);
1349 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1351 goto out_sock_release
;
1357 err
= -EAFNOSUPPORT
;
1360 module_put(pf
->owner
);
1367 goto out_sock_release
;
1369 EXPORT_SYMBOL(__sock_create
);
1371 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1373 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1375 EXPORT_SYMBOL(sock_create
);
1377 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1379 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1381 EXPORT_SYMBOL(sock_create_kern
);
1383 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1386 struct socket
*sock
;
1389 /* Check the SOCK_* constants for consistency. */
1390 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1391 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1392 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1393 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1395 flags
= type
& ~SOCK_TYPE_MASK
;
1396 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1398 type
&= SOCK_TYPE_MASK
;
1400 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1401 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1403 retval
= sock_create(family
, type
, protocol
, &sock
);
1407 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1412 /* It may be already another descriptor 8) Not kernel problem. */
1421 * Create a pair of connected sockets.
1424 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1425 int __user
*, usockvec
)
1427 struct socket
*sock1
, *sock2
;
1429 struct file
*newfile1
, *newfile2
;
1432 flags
= type
& ~SOCK_TYPE_MASK
;
1433 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1435 type
&= SOCK_TYPE_MASK
;
1437 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1438 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1441 * Obtain the first socket and check if the underlying protocol
1442 * supports the socketpair call.
1445 err
= sock_create(family
, type
, protocol
, &sock1
);
1449 err
= sock_create(family
, type
, protocol
, &sock2
);
1453 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1455 goto out_release_both
;
1457 fd1
= get_unused_fd_flags(flags
);
1458 if (unlikely(fd1
< 0)) {
1460 goto out_release_both
;
1462 fd2
= get_unused_fd_flags(flags
);
1463 if (unlikely(fd2
< 0)) {
1466 goto out_release_both
;
1469 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1470 if (unlikely(IS_ERR(newfile1
))) {
1471 err
= PTR_ERR(newfile1
);
1474 goto out_release_both
;
1477 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1478 if (IS_ERR(newfile2
)) {
1479 err
= PTR_ERR(newfile2
);
1483 sock_release(sock2
);
1487 audit_fd_pair(fd1
, fd2
);
1488 fd_install(fd1
, newfile1
);
1489 fd_install(fd2
, newfile2
);
1490 /* fd1 and fd2 may be already another descriptors.
1491 * Not kernel problem.
1494 err
= put_user(fd1
, &usockvec
[0]);
1496 err
= put_user(fd2
, &usockvec
[1]);
1505 sock_release(sock2
);
1507 sock_release(sock1
);
1513 * Bind a name to a socket. Nothing much to do here since it's
1514 * the protocol's responsibility to handle the local address.
1516 * We move the socket address to kernel space before we call
1517 * the protocol layer (having also checked the address is ok).
1520 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1522 struct socket
*sock
;
1523 struct sockaddr_storage address
;
1524 int err
, fput_needed
;
1526 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1528 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1530 err
= security_socket_bind(sock
,
1531 (struct sockaddr
*)&address
,
1534 err
= sock
->ops
->bind(sock
,
1538 fput_light(sock
->file
, fput_needed
);
1544 * Perform a listen. Basically, we allow the protocol to do anything
1545 * necessary for a listen, and if that works, we mark the socket as
1546 * ready for listening.
1549 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1551 struct socket
*sock
;
1552 int err
, fput_needed
;
1555 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1557 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1558 if ((unsigned int)backlog
> somaxconn
)
1559 backlog
= somaxconn
;
1561 err
= security_socket_listen(sock
, backlog
);
1563 err
= sock
->ops
->listen(sock
, backlog
);
1565 fput_light(sock
->file
, fput_needed
);
1571 * For accept, we attempt to create a new socket, set up the link
1572 * with the client, wake up the client, then return the new
1573 * connected fd. We collect the address of the connector in kernel
1574 * space and move it to user at the very end. This is unclean because
1575 * we open the socket then return an error.
1577 * 1003.1g adds the ability to recvmsg() to query connection pending
1578 * status to recvmsg. We need to add that support in a way thats
1579 * clean when we restucture accept also.
1582 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1583 int __user
*, upeer_addrlen
, int, flags
)
1585 struct socket
*sock
, *newsock
;
1586 struct file
*newfile
;
1587 int err
, len
, newfd
, fput_needed
;
1588 struct sockaddr_storage address
;
1590 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1593 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1594 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1596 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1601 newsock
= sock_alloc();
1605 newsock
->type
= sock
->type
;
1606 newsock
->ops
= sock
->ops
;
1609 * We don't need try_module_get here, as the listening socket (sock)
1610 * has the protocol module (sock->ops->owner) held.
1612 __module_get(newsock
->ops
->owner
);
1614 newfd
= get_unused_fd_flags(flags
);
1615 if (unlikely(newfd
< 0)) {
1617 sock_release(newsock
);
1620 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1621 if (unlikely(IS_ERR(newfile
))) {
1622 err
= PTR_ERR(newfile
);
1623 put_unused_fd(newfd
);
1624 sock_release(newsock
);
1628 err
= security_socket_accept(sock
, newsock
);
1632 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1636 if (upeer_sockaddr
) {
1637 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1639 err
= -ECONNABORTED
;
1642 err
= move_addr_to_user(&address
,
1643 len
, upeer_sockaddr
, upeer_addrlen
);
1648 /* File flags are not inherited via accept() unlike another OSes. */
1650 fd_install(newfd
, newfile
);
1654 fput_light(sock
->file
, fput_needed
);
1659 put_unused_fd(newfd
);
1663 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1664 int __user
*, upeer_addrlen
)
1666 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1670 * Attempt to connect to a socket with the server address. The address
1671 * is in user space so we verify it is OK and move it to kernel space.
1673 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1676 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1677 * other SEQPACKET protocols that take time to connect() as it doesn't
1678 * include the -EINPROGRESS status for such sockets.
1681 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1684 struct socket
*sock
;
1685 struct sockaddr_storage address
;
1686 int err
, fput_needed
;
1688 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1691 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1696 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1700 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1701 sock
->file
->f_flags
);
1703 fput_light(sock
->file
, fput_needed
);
1709 * Get the local address ('name') of a socket object. Move the obtained
1710 * name to user space.
1713 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1714 int __user
*, usockaddr_len
)
1716 struct socket
*sock
;
1717 struct sockaddr_storage address
;
1718 int len
, err
, fput_needed
;
1720 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1724 err
= security_socket_getsockname(sock
);
1728 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1731 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1734 fput_light(sock
->file
, fput_needed
);
1740 * Get the remote address ('name') of a socket object. Move the obtained
1741 * name to user space.
1744 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1745 int __user
*, usockaddr_len
)
1747 struct socket
*sock
;
1748 struct sockaddr_storage address
;
1749 int len
, err
, fput_needed
;
1751 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1753 err
= security_socket_getpeername(sock
);
1755 fput_light(sock
->file
, fput_needed
);
1760 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1763 err
= move_addr_to_user(&address
, len
, usockaddr
,
1765 fput_light(sock
->file
, fput_needed
);
1771 * Send a datagram to a given address. We move the address into kernel
1772 * space and check the user space data area is readable before invoking
1776 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1777 unsigned int, flags
, struct sockaddr __user
*, addr
,
1780 struct socket
*sock
;
1781 struct sockaddr_storage address
;
1789 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1793 iov
.iov_base
= buff
;
1795 msg
.msg_name
= NULL
;
1798 msg
.msg_control
= NULL
;
1799 msg
.msg_controllen
= 0;
1800 msg
.msg_namelen
= 0;
1802 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1805 msg
.msg_name
= (struct sockaddr
*)&address
;
1806 msg
.msg_namelen
= addr_len
;
1808 if (sock
->file
->f_flags
& O_NONBLOCK
)
1809 flags
|= MSG_DONTWAIT
;
1810 msg
.msg_flags
= flags
;
1811 err
= sock_sendmsg(sock
, &msg
, len
);
1814 fput_light(sock
->file
, fput_needed
);
1820 * Send a datagram down a socket.
1823 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1824 unsigned int, flags
)
1826 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1830 * Receive a frame from the socket and optionally record the address of the
1831 * sender. We verify the buffers are writable and if needed move the
1832 * sender address from kernel to user space.
1835 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1836 unsigned int, flags
, struct sockaddr __user
*, addr
,
1837 int __user
*, addr_len
)
1839 struct socket
*sock
;
1842 struct sockaddr_storage address
;
1848 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1852 msg
.msg_control
= NULL
;
1853 msg
.msg_controllen
= 0;
1857 iov
.iov_base
= ubuf
;
1858 msg
.msg_name
= (struct sockaddr
*)&address
;
1859 msg
.msg_namelen
= sizeof(address
);
1860 if (sock
->file
->f_flags
& O_NONBLOCK
)
1861 flags
|= MSG_DONTWAIT
;
1862 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1864 if (err
>= 0 && addr
!= NULL
) {
1865 err2
= move_addr_to_user(&address
,
1866 msg
.msg_namelen
, addr
, addr_len
);
1871 fput_light(sock
->file
, fput_needed
);
1877 * Receive a datagram from a socket.
1880 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1883 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1887 * Set a socket option. Because we don't know the option lengths we have
1888 * to pass the user mode parameter for the protocols to sort out.
1891 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1892 char __user
*, optval
, int, optlen
)
1894 int err
, fput_needed
;
1895 struct socket
*sock
;
1900 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1902 err
= security_socket_setsockopt(sock
, level
, optname
);
1906 if (level
== SOL_SOCKET
)
1908 sock_setsockopt(sock
, level
, optname
, optval
,
1912 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1915 fput_light(sock
->file
, fput_needed
);
1921 * Get a socket option. Because we don't know the option lengths we have
1922 * to pass a user mode parameter for the protocols to sort out.
1925 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1926 char __user
*, optval
, int __user
*, optlen
)
1928 int err
, fput_needed
;
1929 struct socket
*sock
;
1931 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1933 err
= security_socket_getsockopt(sock
, level
, optname
);
1937 if (level
== SOL_SOCKET
)
1939 sock_getsockopt(sock
, level
, optname
, optval
,
1943 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1946 fput_light(sock
->file
, fput_needed
);
1952 * Shutdown a socket.
1955 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1957 int err
, fput_needed
;
1958 struct socket
*sock
;
1960 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1962 err
= security_socket_shutdown(sock
, how
);
1964 err
= sock
->ops
->shutdown(sock
, how
);
1965 fput_light(sock
->file
, fput_needed
);
1970 /* A couple of helpful macros for getting the address of the 32/64 bit
1971 * fields which are the same type (int / unsigned) on our platforms.
1973 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1974 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1975 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1977 struct used_address
{
1978 struct sockaddr_storage name
;
1979 unsigned int name_len
;
1982 static int __sys_sendmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1983 struct msghdr
*msg_sys
, unsigned int flags
,
1984 struct used_address
*used_address
)
1986 struct compat_msghdr __user
*msg_compat
=
1987 (struct compat_msghdr __user
*)msg
;
1988 struct sockaddr_storage address
;
1989 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1990 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1991 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1992 /* 20 is size of ipv6_pktinfo */
1993 unsigned char *ctl_buf
= ctl
;
1994 int err
, ctl_len
, total_len
;
1997 if (MSG_CMSG_COMPAT
& flags
) {
1998 if (get_compat_msghdr(msg_sys
, msg_compat
))
2000 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2003 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2005 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2008 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2014 /* This will also move the address data into kernel space */
2015 if (MSG_CMSG_COMPAT
& flags
) {
2016 err
= verify_compat_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2018 err
= verify_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2025 if (msg_sys
->msg_controllen
> INT_MAX
)
2027 ctl_len
= msg_sys
->msg_controllen
;
2028 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2030 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2034 ctl_buf
= msg_sys
->msg_control
;
2035 ctl_len
= msg_sys
->msg_controllen
;
2036 } else if (ctl_len
) {
2037 if (ctl_len
> sizeof(ctl
)) {
2038 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2039 if (ctl_buf
== NULL
)
2044 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2045 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2046 * checking falls down on this.
2048 if (copy_from_user(ctl_buf
,
2049 (void __user __force
*)msg_sys
->msg_control
,
2052 msg_sys
->msg_control
= ctl_buf
;
2054 msg_sys
->msg_flags
= flags
;
2056 if (sock
->file
->f_flags
& O_NONBLOCK
)
2057 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2059 * If this is sendmmsg() and current destination address is same as
2060 * previously succeeded address, omit asking LSM's decision.
2061 * used_address->name_len is initialized to UINT_MAX so that the first
2062 * destination address never matches.
2064 if (used_address
&& msg_sys
->msg_name
&&
2065 used_address
->name_len
== msg_sys
->msg_namelen
&&
2066 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2067 used_address
->name_len
)) {
2068 err
= sock_sendmsg_nosec(sock
, msg_sys
, total_len
);
2071 err
= sock_sendmsg(sock
, msg_sys
, total_len
);
2073 * If this is sendmmsg() and sending to current destination address was
2074 * successful, remember it.
2076 if (used_address
&& err
>= 0) {
2077 used_address
->name_len
= msg_sys
->msg_namelen
;
2078 if (msg_sys
->msg_name
)
2079 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2080 used_address
->name_len
);
2085 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2087 if (iov
!= iovstack
)
2094 * BSD sendmsg interface
2097 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned int, flags
)
2099 int fput_needed
, err
;
2100 struct msghdr msg_sys
;
2101 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2106 err
= __sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
);
2108 fput_light(sock
->file
, fput_needed
);
2114 * Linux sendmmsg interface
2117 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2120 int fput_needed
, err
, datagrams
;
2121 struct socket
*sock
;
2122 struct mmsghdr __user
*entry
;
2123 struct compat_mmsghdr __user
*compat_entry
;
2124 struct msghdr msg_sys
;
2125 struct used_address used_address
;
2127 if (vlen
> UIO_MAXIOV
)
2132 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2136 used_address
.name_len
= UINT_MAX
;
2138 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2141 while (datagrams
< vlen
) {
2142 if (MSG_CMSG_COMPAT
& flags
) {
2143 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)compat_entry
,
2144 &msg_sys
, flags
, &used_address
);
2147 err
= __put_user(err
, &compat_entry
->msg_len
);
2150 err
= __sys_sendmsg(sock
, (struct msghdr __user
*)entry
,
2151 &msg_sys
, flags
, &used_address
);
2154 err
= put_user(err
, &entry
->msg_len
);
2163 fput_light(sock
->file
, fput_needed
);
2165 /* We only return an error if no datagrams were able to be sent */
2172 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2173 unsigned int, vlen
, unsigned int, flags
)
2175 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2178 static int __sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
2179 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2181 struct compat_msghdr __user
*msg_compat
=
2182 (struct compat_msghdr __user
*)msg
;
2183 struct iovec iovstack
[UIO_FASTIOV
];
2184 struct iovec
*iov
= iovstack
;
2185 unsigned long cmsg_ptr
;
2186 int err
, total_len
, len
;
2188 /* kernel mode address */
2189 struct sockaddr_storage addr
;
2191 /* user mode address pointers */
2192 struct sockaddr __user
*uaddr
;
2193 int __user
*uaddr_len
;
2195 if (MSG_CMSG_COMPAT
& flags
) {
2196 if (get_compat_msghdr(msg_sys
, msg_compat
))
2198 } else if (copy_from_user(msg_sys
, msg
, sizeof(struct msghdr
)))
2201 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2203 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2206 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2213 * Save the user-mode address (verify_iovec will change the
2214 * kernel msghdr to use the kernel address space)
2217 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2218 uaddr_len
= COMPAT_NAMELEN(msg
);
2219 if (MSG_CMSG_COMPAT
& flags
) {
2220 err
= verify_compat_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2222 err
= verify_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2227 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2228 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2230 if (sock
->file
->f_flags
& O_NONBLOCK
)
2231 flags
|= MSG_DONTWAIT
;
2232 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2238 if (uaddr
!= NULL
) {
2239 err
= move_addr_to_user(&addr
,
2240 msg_sys
->msg_namelen
, uaddr
,
2245 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2249 if (MSG_CMSG_COMPAT
& flags
)
2250 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2251 &msg_compat
->msg_controllen
);
2253 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2254 &msg
->msg_controllen
);
2260 if (iov
!= iovstack
)
2267 * BSD recvmsg interface
2270 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2271 unsigned int, flags
)
2273 int fput_needed
, err
;
2274 struct msghdr msg_sys
;
2275 struct socket
*sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2280 err
= __sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2282 fput_light(sock
->file
, fput_needed
);
2288 * Linux recvmmsg interface
2291 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2292 unsigned int flags
, struct timespec
*timeout
)
2294 int fput_needed
, err
, datagrams
;
2295 struct socket
*sock
;
2296 struct mmsghdr __user
*entry
;
2297 struct compat_mmsghdr __user
*compat_entry
;
2298 struct msghdr msg_sys
;
2299 struct timespec end_time
;
2302 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2308 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2312 err
= sock_error(sock
->sk
);
2317 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2319 while (datagrams
< vlen
) {
2321 * No need to ask LSM for more than the first datagram.
2323 if (MSG_CMSG_COMPAT
& flags
) {
2324 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2325 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2329 err
= __put_user(err
, &compat_entry
->msg_len
);
2332 err
= __sys_recvmsg(sock
, (struct msghdr __user
*)entry
,
2333 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2337 err
= put_user(err
, &entry
->msg_len
);
2345 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2346 if (flags
& MSG_WAITFORONE
)
2347 flags
|= MSG_DONTWAIT
;
2350 ktime_get_ts(timeout
);
2351 *timeout
= timespec_sub(end_time
, *timeout
);
2352 if (timeout
->tv_sec
< 0) {
2353 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2357 /* Timeout, return less than vlen datagrams */
2358 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2362 /* Out of band data, return right away */
2363 if (msg_sys
.msg_flags
& MSG_OOB
)
2368 fput_light(sock
->file
, fput_needed
);
2373 if (datagrams
!= 0) {
2375 * We may return less entries than requested (vlen) if the
2376 * sock is non block and there aren't enough datagrams...
2378 if (err
!= -EAGAIN
) {
2380 * ... or if recvmsg returns an error after we
2381 * received some datagrams, where we record the
2382 * error to return on the next call or if the
2383 * app asks about it using getsockopt(SO_ERROR).
2385 sock
->sk
->sk_err
= -err
;
2394 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2395 unsigned int, vlen
, unsigned int, flags
,
2396 struct timespec __user
*, timeout
)
2399 struct timespec timeout_sys
;
2402 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2404 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2407 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2409 if (datagrams
> 0 &&
2410 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2411 datagrams
= -EFAULT
;
2416 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2417 /* Argument list sizes for sys_socketcall */
2418 #define AL(x) ((x) * sizeof(unsigned long))
2419 static const unsigned char nargs
[21] = {
2420 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2421 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2422 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2429 * System call vectors.
2431 * Argument checking cleaned up. Saved 20% in size.
2432 * This function doesn't need to set the kernel lock because
2433 * it is set by the callees.
2436 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2439 unsigned long a0
, a1
;
2443 if (call
< 1 || call
> SYS_SENDMMSG
)
2447 if (len
> sizeof(a
))
2450 /* copy_from_user should be SMP safe. */
2451 if (copy_from_user(a
, args
, len
))
2454 audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2461 err
= sys_socket(a0
, a1
, a
[2]);
2464 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2467 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2470 err
= sys_listen(a0
, a1
);
2473 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2474 (int __user
*)a
[2], 0);
2476 case SYS_GETSOCKNAME
:
2478 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2479 (int __user
*)a
[2]);
2481 case SYS_GETPEERNAME
:
2483 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2484 (int __user
*)a
[2]);
2486 case SYS_SOCKETPAIR
:
2487 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2490 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2493 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2494 (struct sockaddr __user
*)a
[4], a
[5]);
2497 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2500 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2501 (struct sockaddr __user
*)a
[4],
2502 (int __user
*)a
[5]);
2505 err
= sys_shutdown(a0
, a1
);
2507 case SYS_SETSOCKOPT
:
2508 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2510 case SYS_GETSOCKOPT
:
2512 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2513 (int __user
*)a
[4]);
2516 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2519 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2522 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2525 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2526 (struct timespec __user
*)a
[4]);
2529 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2530 (int __user
*)a
[2], a
[3]);
2539 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2542 * sock_register - add a socket protocol handler
2543 * @ops: description of protocol
2545 * This function is called by a protocol handler that wants to
2546 * advertise its address family, and have it linked into the
2547 * socket interface. The value ops->family coresponds to the
2548 * socket system call protocol family.
2550 int sock_register(const struct net_proto_family
*ops
)
2554 if (ops
->family
>= NPROTO
) {
2555 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2560 spin_lock(&net_family_lock
);
2561 if (rcu_dereference_protected(net_families
[ops
->family
],
2562 lockdep_is_held(&net_family_lock
)))
2565 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2568 spin_unlock(&net_family_lock
);
2570 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2573 EXPORT_SYMBOL(sock_register
);
2576 * sock_unregister - remove a protocol handler
2577 * @family: protocol family to remove
2579 * This function is called by a protocol handler that wants to
2580 * remove its address family, and have it unlinked from the
2581 * new socket creation.
2583 * If protocol handler is a module, then it can use module reference
2584 * counts to protect against new references. If protocol handler is not
2585 * a module then it needs to provide its own protection in
2586 * the ops->create routine.
2588 void sock_unregister(int family
)
2590 BUG_ON(family
< 0 || family
>= NPROTO
);
2592 spin_lock(&net_family_lock
);
2593 RCU_INIT_POINTER(net_families
[family
], NULL
);
2594 spin_unlock(&net_family_lock
);
2598 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2600 EXPORT_SYMBOL(sock_unregister
);
2602 static int __init
sock_init(void)
2606 * Initialize the network sysctl infrastructure.
2608 err
= net_sysctl_init();
2613 * Initialize skbuff SLAB cache
2618 * Initialize the protocols module.
2623 err
= register_filesystem(&sock_fs_type
);
2626 sock_mnt
= kern_mount(&sock_fs_type
);
2627 if (IS_ERR(sock_mnt
)) {
2628 err
= PTR_ERR(sock_mnt
);
2632 /* The real protocol initialization is performed in later initcalls.
2635 #ifdef CONFIG_NETFILTER
2639 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2640 skb_timestamping_init();
2647 unregister_filesystem(&sock_fs_type
);
2652 core_initcall(sock_init
); /* early initcall */
2654 #ifdef CONFIG_PROC_FS
2655 void socket_seq_show(struct seq_file
*seq
)
2660 for_each_possible_cpu(cpu
)
2661 counter
+= per_cpu(sockets_in_use
, cpu
);
2663 /* It can be negative, by the way. 8) */
2667 seq_printf(seq
, "sockets: used %d\n", counter
);
2669 #endif /* CONFIG_PROC_FS */
2671 #ifdef CONFIG_COMPAT
2672 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2673 unsigned int cmd
, void __user
*up
)
2675 mm_segment_t old_fs
= get_fs();
2680 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2683 err
= compat_put_timeval(&ktv
, up
);
2688 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2689 unsigned int cmd
, void __user
*up
)
2691 mm_segment_t old_fs
= get_fs();
2692 struct timespec kts
;
2696 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2699 err
= compat_put_timespec(&kts
, up
);
2704 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2706 struct ifreq __user
*uifr
;
2709 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2710 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2713 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2717 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2723 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2725 struct compat_ifconf ifc32
;
2727 struct ifconf __user
*uifc
;
2728 struct compat_ifreq __user
*ifr32
;
2729 struct ifreq __user
*ifr
;
2733 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2736 memset(&ifc
, 0, sizeof(ifc
));
2737 if (ifc32
.ifcbuf
== 0) {
2741 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2743 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2744 sizeof(struct ifreq
);
2745 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2747 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2748 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2749 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2750 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2756 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2759 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2763 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2767 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2769 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2770 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2771 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2777 if (ifc32
.ifcbuf
== 0) {
2778 /* Translate from 64-bit structure multiple to
2782 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2787 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2793 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2795 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2796 bool convert_in
= false, convert_out
= false;
2797 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2798 struct ethtool_rxnfc __user
*rxnfc
;
2799 struct ifreq __user
*ifr
;
2800 u32 rule_cnt
= 0, actual_rule_cnt
;
2805 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2808 compat_rxnfc
= compat_ptr(data
);
2810 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2813 /* Most ethtool structures are defined without padding.
2814 * Unfortunately struct ethtool_rxnfc is an exception.
2819 case ETHTOOL_GRXCLSRLALL
:
2820 /* Buffer size is variable */
2821 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2823 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2825 buf_size
+= rule_cnt
* sizeof(u32
);
2827 case ETHTOOL_GRXRINGS
:
2828 case ETHTOOL_GRXCLSRLCNT
:
2829 case ETHTOOL_GRXCLSRULE
:
2830 case ETHTOOL_SRXCLSRLINS
:
2833 case ETHTOOL_SRXCLSRLDEL
:
2834 buf_size
+= sizeof(struct ethtool_rxnfc
);
2839 ifr
= compat_alloc_user_space(buf_size
);
2840 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2842 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2845 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2846 &ifr
->ifr_ifru
.ifru_data
))
2850 /* We expect there to be holes between fs.m_ext and
2851 * fs.ring_cookie and at the end of fs, but nowhere else.
2853 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2854 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2855 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2856 sizeof(rxnfc
->fs
.m_ext
));
2858 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2859 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2860 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2861 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2863 if (copy_in_user(rxnfc
, compat_rxnfc
,
2864 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2865 (void __user
*)rxnfc
) ||
2866 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2867 &compat_rxnfc
->fs
.ring_cookie
,
2868 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2869 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2870 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2871 sizeof(rxnfc
->rule_cnt
)))
2875 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2880 if (copy_in_user(compat_rxnfc
, rxnfc
,
2881 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2882 (const void __user
*)rxnfc
) ||
2883 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2884 &rxnfc
->fs
.ring_cookie
,
2885 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2886 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2887 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2888 sizeof(rxnfc
->rule_cnt
)))
2891 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2892 /* As an optimisation, we only copy the actual
2893 * number of rules that the underlying
2894 * function returned. Since Mallory might
2895 * change the rule count in user memory, we
2896 * check that it is less than the rule count
2897 * originally given (as the user buffer size),
2898 * which has been range-checked.
2900 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2902 if (actual_rule_cnt
< rule_cnt
)
2903 rule_cnt
= actual_rule_cnt
;
2904 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2905 &rxnfc
->rule_locs
[0],
2906 rule_cnt
* sizeof(u32
)))
2914 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2917 compat_uptr_t uptr32
;
2918 struct ifreq __user
*uifr
;
2920 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2921 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2924 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2927 uptr
= compat_ptr(uptr32
);
2929 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2932 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2935 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2936 struct compat_ifreq __user
*ifr32
)
2939 struct ifreq __user
*uifr
;
2940 mm_segment_t old_fs
;
2946 case SIOCBONDENSLAVE
:
2947 case SIOCBONDRELEASE
:
2948 case SIOCBONDSETHWADDR
:
2949 case SIOCBONDCHANGEACTIVE
:
2950 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2955 err
= dev_ioctl(net
, cmd
,
2956 (struct ifreq __user __force
*) &kifr
);
2960 case SIOCBONDSLAVEINFOQUERY
:
2961 case SIOCBONDINFOQUERY
:
2962 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2963 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2966 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2969 datap
= compat_ptr(data
);
2970 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
2973 return dev_ioctl(net
, cmd
, uifr
);
2975 return -ENOIOCTLCMD
;
2979 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
2980 struct compat_ifreq __user
*u_ifreq32
)
2982 struct ifreq __user
*u_ifreq64
;
2983 char tmp_buf
[IFNAMSIZ
];
2984 void __user
*data64
;
2987 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2990 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2992 data64
= compat_ptr(data32
);
2994 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2996 /* Don't check these user accesses, just let that get trapped
2997 * in the ioctl handler instead.
2999 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
3002 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
3005 return dev_ioctl(net
, cmd
, u_ifreq64
);
3008 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
3009 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
3011 struct ifreq __user
*uifr
;
3014 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3015 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
3018 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3029 case SIOCGIFBRDADDR
:
3030 case SIOCGIFDSTADDR
:
3031 case SIOCGIFNETMASK
:
3036 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3044 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3045 struct compat_ifreq __user
*uifr32
)
3048 struct compat_ifmap __user
*uifmap32
;
3049 mm_segment_t old_fs
;
3052 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3053 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3054 err
|= __get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3055 err
|= __get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3056 err
|= __get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3057 err
|= __get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3058 err
|= __get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3059 err
|= __get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3065 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3068 if (cmd
== SIOCGIFMAP
&& !err
) {
3069 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3070 err
|= __put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3071 err
|= __put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3072 err
|= __put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3073 err
|= __put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3074 err
|= __put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3075 err
|= __put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3082 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3085 compat_uptr_t uptr32
;
3086 struct ifreq __user
*uifr
;
3088 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3089 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
3092 if (get_user(uptr32
, &uifr32
->ifr_data
))
3095 uptr
= compat_ptr(uptr32
);
3097 if (put_user(uptr
, &uifr
->ifr_data
))
3100 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
3105 struct sockaddr rt_dst
; /* target address */
3106 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3107 struct sockaddr rt_genmask
; /* target network mask (IP) */
3108 unsigned short rt_flags
;
3111 unsigned char rt_tos
;
3112 unsigned char rt_class
;
3114 short rt_metric
; /* +1 for binary compatibility! */
3115 /* char * */ u32 rt_dev
; /* forcing the device at add */
3116 u32 rt_mtu
; /* per route MTU/Window */
3117 u32 rt_window
; /* Window clamping */
3118 unsigned short rt_irtt
; /* Initial RTT */
3121 struct in6_rtmsg32
{
3122 struct in6_addr rtmsg_dst
;
3123 struct in6_addr rtmsg_src
;
3124 struct in6_addr rtmsg_gateway
;
3134 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3135 unsigned int cmd
, void __user
*argp
)
3139 struct in6_rtmsg r6
;
3143 mm_segment_t old_fs
= get_fs();
3145 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3146 struct in6_rtmsg32 __user
*ur6
= argp
;
3147 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3148 3 * sizeof(struct in6_addr
));
3149 ret
|= __get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3150 ret
|= __get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3151 ret
|= __get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3152 ret
|= __get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3153 ret
|= __get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3154 ret
|= __get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3155 ret
|= __get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3159 struct rtentry32 __user
*ur4
= argp
;
3160 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3161 3 * sizeof(struct sockaddr
));
3162 ret
|= __get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3163 ret
|= __get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3164 ret
|= __get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3165 ret
|= __get_user(r4
.rt_window
, &(ur4
->rt_window
));
3166 ret
|= __get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3167 ret
|= __get_user(rtdev
, &(ur4
->rt_dev
));
3169 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3170 r4
.rt_dev
= (char __user __force
*)devname
;
3184 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3191 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3192 * for some operations; this forces use of the newer bridge-utils that
3193 * use compatible ioctls
3195 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3199 if (get_user(tmp
, argp
))
3201 if (tmp
== BRCTL_GET_VERSION
)
3202 return BRCTL_VERSION
+ 1;
3206 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3207 unsigned int cmd
, unsigned long arg
)
3209 void __user
*argp
= compat_ptr(arg
);
3210 struct sock
*sk
= sock
->sk
;
3211 struct net
*net
= sock_net(sk
);
3213 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3214 return siocdevprivate_ioctl(net
, cmd
, argp
);
3219 return old_bridge_ioctl(argp
);
3221 return dev_ifname32(net
, argp
);
3223 return dev_ifconf(net
, argp
);
3225 return ethtool_ioctl(net
, argp
);
3227 return compat_siocwandev(net
, argp
);
3230 return compat_sioc_ifmap(net
, cmd
, argp
);
3231 case SIOCBONDENSLAVE
:
3232 case SIOCBONDRELEASE
:
3233 case SIOCBONDSETHWADDR
:
3234 case SIOCBONDSLAVEINFOQUERY
:
3235 case SIOCBONDINFOQUERY
:
3236 case SIOCBONDCHANGEACTIVE
:
3237 return bond_ioctl(net
, cmd
, argp
);
3240 return routing_ioctl(net
, sock
, cmd
, argp
);
3242 return do_siocgstamp(net
, sock
, cmd
, argp
);
3244 return do_siocgstampns(net
, sock
, cmd
, argp
);
3246 return compat_siocshwtstamp(net
, argp
);
3258 return sock_ioctl(file
, cmd
, arg
);
3275 case SIOCSIFHWBROADCAST
:
3277 case SIOCGIFBRDADDR
:
3278 case SIOCSIFBRDADDR
:
3279 case SIOCGIFDSTADDR
:
3280 case SIOCSIFDSTADDR
:
3281 case SIOCGIFNETMASK
:
3282 case SIOCSIFNETMASK
:
3293 return dev_ifsioc(net
, sock
, cmd
, argp
);
3299 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3302 return -ENOIOCTLCMD
;
3305 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3308 struct socket
*sock
= file
->private_data
;
3309 int ret
= -ENOIOCTLCMD
;
3316 if (sock
->ops
->compat_ioctl
)
3317 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3319 if (ret
== -ENOIOCTLCMD
&&
3320 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3321 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3323 if (ret
== -ENOIOCTLCMD
)
3324 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3330 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3332 return sock
->ops
->bind(sock
, addr
, addrlen
);
3334 EXPORT_SYMBOL(kernel_bind
);
3336 int kernel_listen(struct socket
*sock
, int backlog
)
3338 return sock
->ops
->listen(sock
, backlog
);
3340 EXPORT_SYMBOL(kernel_listen
);
3342 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3344 struct sock
*sk
= sock
->sk
;
3347 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3352 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3354 sock_release(*newsock
);
3359 (*newsock
)->ops
= sock
->ops
;
3360 __module_get((*newsock
)->ops
->owner
);
3365 EXPORT_SYMBOL(kernel_accept
);
3367 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3370 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3372 EXPORT_SYMBOL(kernel_connect
);
3374 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3377 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3379 EXPORT_SYMBOL(kernel_getsockname
);
3381 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3384 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3386 EXPORT_SYMBOL(kernel_getpeername
);
3388 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3389 char *optval
, int *optlen
)
3391 mm_segment_t oldfs
= get_fs();
3392 char __user
*uoptval
;
3393 int __user
*uoptlen
;
3396 uoptval
= (char __user __force
*) optval
;
3397 uoptlen
= (int __user __force
*) optlen
;
3400 if (level
== SOL_SOCKET
)
3401 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3403 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3408 EXPORT_SYMBOL(kernel_getsockopt
);
3410 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3411 char *optval
, unsigned int optlen
)
3413 mm_segment_t oldfs
= get_fs();
3414 char __user
*uoptval
;
3417 uoptval
= (char __user __force
*) optval
;
3420 if (level
== SOL_SOCKET
)
3421 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3423 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3428 EXPORT_SYMBOL(kernel_setsockopt
);
3430 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3431 size_t size
, int flags
)
3433 if (sock
->ops
->sendpage
)
3434 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3436 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3438 EXPORT_SYMBOL(kernel_sendpage
);
3440 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3442 mm_segment_t oldfs
= get_fs();
3446 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3451 EXPORT_SYMBOL(kernel_sock_ioctl
);
3453 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3455 return sock
->ops
->shutdown(sock
, how
);
3457 EXPORT_SYMBOL(kernel_sock_shutdown
);