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>
107 #include <net/busy_poll.h>
109 #ifdef CONFIG_NET_RX_BUSY_POLL
110 unsigned int sysctl_net_busy_read __read_mostly
;
111 unsigned int sysctl_net_busy_poll __read_mostly
;
114 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
115 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
116 unsigned long nr_segs
, loff_t pos
);
117 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
118 unsigned long nr_segs
, loff_t pos
);
119 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
121 static int sock_close(struct inode
*inode
, struct file
*file
);
122 static unsigned int sock_poll(struct file
*file
,
123 struct poll_table_struct
*wait
);
124 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
126 static long compat_sock_ioctl(struct file
*file
,
127 unsigned int cmd
, unsigned long arg
);
129 static int sock_fasync(int fd
, struct file
*filp
, int on
);
130 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
131 int offset
, size_t size
, loff_t
*ppos
, int more
);
132 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
133 struct pipe_inode_info
*pipe
, size_t len
,
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
141 static const struct file_operations socket_file_ops
= {
142 .owner
= THIS_MODULE
,
144 .aio_read
= sock_aio_read
,
145 .aio_write
= sock_aio_write
,
147 .unlocked_ioctl
= sock_ioctl
,
149 .compat_ioctl
= compat_sock_ioctl
,
152 .open
= sock_no_open
, /* special open code to disallow open via /proc */
153 .release
= sock_close
,
154 .fasync
= sock_fasync
,
155 .sendpage
= sock_sendpage
,
156 .splice_write
= generic_splice_sendpage
,
157 .splice_read
= sock_splice_read
,
161 * The protocol list. Each protocol is registered in here.
164 static DEFINE_SPINLOCK(net_family_lock
);
165 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
168 * Statistics counters of the socket lists
171 static DEFINE_PER_CPU(int, sockets_in_use
);
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
192 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
196 if (copy_from_user(kaddr
, uaddr
, ulen
))
198 return audit_sockaddr(ulen
, kaddr
);
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
218 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
219 void __user
*uaddr
, int __user
*ulen
)
224 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
225 err
= get_user(len
, ulen
);
233 if (audit_sockaddr(klen
, kaddr
))
235 if (copy_to_user(uaddr
, kaddr
, len
))
239 * "fromlen shall refer to the value before truncation.."
242 return __put_user(klen
, ulen
);
245 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
247 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
249 struct socket_alloc
*ei
;
250 struct socket_wq
*wq
;
252 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
255 wq
= kmalloc(sizeof(*wq
), GFP_KERNEL
);
257 kmem_cache_free(sock_inode_cachep
, ei
);
260 init_waitqueue_head(&wq
->wait
);
261 wq
->fasync_list
= NULL
;
262 RCU_INIT_POINTER(ei
->socket
.wq
, wq
);
264 ei
->socket
.state
= SS_UNCONNECTED
;
265 ei
->socket
.flags
= 0;
266 ei
->socket
.ops
= NULL
;
267 ei
->socket
.sk
= NULL
;
268 ei
->socket
.file
= NULL
;
270 return &ei
->vfs_inode
;
273 static void sock_destroy_inode(struct inode
*inode
)
275 struct socket_alloc
*ei
;
276 struct socket_wq
*wq
;
278 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
279 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
281 kmem_cache_free(sock_inode_cachep
, ei
);
284 static void init_once(void *foo
)
286 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
288 inode_init_once(&ei
->vfs_inode
);
291 static int init_inodecache(void)
293 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
294 sizeof(struct socket_alloc
),
296 (SLAB_HWCACHE_ALIGN
|
297 SLAB_RECLAIM_ACCOUNT
|
300 if (sock_inode_cachep
== NULL
)
305 static const struct super_operations sockfs_ops
= {
306 .alloc_inode
= sock_alloc_inode
,
307 .destroy_inode
= sock_destroy_inode
,
308 .statfs
= simple_statfs
,
312 * sockfs_dname() is called from d_path().
314 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
316 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
317 dentry
->d_inode
->i_ino
);
320 static const struct dentry_operations sockfs_dentry_operations
= {
321 .d_dname
= sockfs_dname
,
324 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
325 int flags
, const char *dev_name
, void *data
)
327 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
328 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
331 static struct vfsmount
*sock_mnt __read_mostly
;
333 static struct file_system_type sock_fs_type
= {
335 .mount
= sockfs_mount
,
336 .kill_sb
= kill_anon_super
,
340 * Obtains the first available file descriptor and sets it up for use.
342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
356 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
358 struct qstr name
= { .name
= "" };
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
))
371 return ERR_PTR(-ENOMEM
);
372 path
.mnt
= mntget(sock_mnt
);
374 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
375 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
377 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
379 if (unlikely(IS_ERR(file
))) {
380 /* drop dentry, keep inode */
381 ihold(path
.dentry
->d_inode
);
387 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
388 file
->private_data
= sock
;
391 EXPORT_SYMBOL(sock_alloc_file
);
393 static int sock_map_fd(struct socket
*sock
, int flags
)
395 struct file
*newfile
;
396 int fd
= get_unused_fd_flags(flags
);
397 if (unlikely(fd
< 0))
400 newfile
= sock_alloc_file(sock
, flags
, NULL
);
401 if (likely(!IS_ERR(newfile
))) {
402 fd_install(fd
, newfile
);
407 return PTR_ERR(newfile
);
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 void 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
;
620 EXPORT_SYMBOL(sock_tx_timestamp
);
622 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
623 struct msghdr
*msg
, size_t size
)
625 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
632 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
635 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
636 struct msghdr
*msg
, size_t size
)
638 int err
= security_socket_sendmsg(sock
, msg
, size
);
640 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
643 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
646 struct sock_iocb siocb
;
649 init_sync_kiocb(&iocb
, NULL
);
650 iocb
.private = &siocb
;
651 ret
= __sock_sendmsg(&iocb
, sock
, msg
, size
);
652 if (-EIOCBQUEUED
== ret
)
653 ret
= wait_on_sync_kiocb(&iocb
);
656 EXPORT_SYMBOL(sock_sendmsg
);
658 static int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
661 struct sock_iocb siocb
;
664 init_sync_kiocb(&iocb
, NULL
);
665 iocb
.private = &siocb
;
666 ret
= __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
);
667 if (-EIOCBQUEUED
== ret
)
668 ret
= wait_on_sync_kiocb(&iocb
);
672 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
673 struct kvec
*vec
, size_t num
, size_t size
)
675 mm_segment_t oldfs
= get_fs();
680 * the following is safe, since for compiler definitions of kvec and
681 * iovec are identical, yielding the same in-core layout and alignment
683 msg
->msg_iov
= (struct iovec
*)vec
;
684 msg
->msg_iovlen
= num
;
685 result
= sock_sendmsg(sock
, msg
, size
);
689 EXPORT_SYMBOL(kernel_sendmsg
);
692 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
694 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
697 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
698 struct timespec ts
[3];
700 struct skb_shared_hwtstamps
*shhwtstamps
=
703 /* Race occurred between timestamp enabling and packet
704 receiving. Fill in the current time for now. */
705 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
706 __net_timestamp(skb
);
708 if (need_software_tstamp
) {
709 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
711 skb_get_timestamp(skb
, &tv
);
712 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
715 skb_get_timestampns(skb
, &ts
[0]);
716 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
717 sizeof(ts
[0]), &ts
[0]);
722 memset(ts
, 0, sizeof(ts
));
723 if (sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
) &&
724 ktime_to_timespec_cond(skb
->tstamp
, ts
+ 0))
727 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
) &&
728 ktime_to_timespec_cond(shhwtstamps
->syststamp
, ts
+ 1))
730 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
) &&
731 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, ts
+ 2))
735 put_cmsg(msg
, SOL_SOCKET
,
736 SCM_TIMESTAMPING
, sizeof(ts
), &ts
);
738 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
740 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
745 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
747 if (!skb
->wifi_acked_valid
)
750 ack
= skb
->wifi_acked
;
752 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
754 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
756 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
759 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
760 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
761 sizeof(__u32
), &skb
->dropcount
);
764 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
767 sock_recv_timestamp(msg
, sk
, skb
);
768 sock_recv_drops(msg
, sk
, skb
);
770 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
772 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
773 struct msghdr
*msg
, size_t size
, int flags
)
775 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
783 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
786 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
787 struct msghdr
*msg
, size_t size
, int flags
)
789 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
791 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
794 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
795 size_t size
, int flags
)
798 struct sock_iocb siocb
;
801 init_sync_kiocb(&iocb
, NULL
);
802 iocb
.private = &siocb
;
803 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
804 if (-EIOCBQUEUED
== ret
)
805 ret
= wait_on_sync_kiocb(&iocb
);
808 EXPORT_SYMBOL(sock_recvmsg
);
810 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
811 size_t size
, int flags
)
814 struct sock_iocb siocb
;
817 init_sync_kiocb(&iocb
, NULL
);
818 iocb
.private = &siocb
;
819 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
820 if (-EIOCBQUEUED
== ret
)
821 ret
= wait_on_sync_kiocb(&iocb
);
826 * kernel_recvmsg - Receive a message from a socket (kernel space)
827 * @sock: The socket to receive the message from
828 * @msg: Received message
829 * @vec: Input s/g array for message data
830 * @num: Size of input s/g array
831 * @size: Number of bytes to read
832 * @flags: Message flags (MSG_DONTWAIT, etc...)
834 * On return the msg structure contains the scatter/gather array passed in the
835 * vec argument. The array is modified so that it consists of the unfilled
836 * portion of the original array.
838 * The returned value is the total number of bytes received, or an error.
840 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
841 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
843 mm_segment_t oldfs
= get_fs();
848 * the following is safe, since for compiler definitions of kvec and
849 * iovec are identical, yielding the same in-core layout and alignment
851 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
852 result
= sock_recvmsg(sock
, msg
, size
, flags
);
856 EXPORT_SYMBOL(kernel_recvmsg
);
858 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
859 int offset
, size_t size
, loff_t
*ppos
, int more
)
864 sock
= file
->private_data
;
866 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
867 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
870 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
873 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
874 struct pipe_inode_info
*pipe
, size_t len
,
877 struct socket
*sock
= file
->private_data
;
879 if (unlikely(!sock
->ops
->splice_read
))
882 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
885 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
886 struct sock_iocb
*siocb
)
888 if (!is_sync_kiocb(iocb
))
892 iocb
->private = siocb
;
896 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
897 struct file
*file
, const struct iovec
*iov
,
898 unsigned long nr_segs
)
900 struct socket
*sock
= file
->private_data
;
904 for (i
= 0; i
< nr_segs
; i
++)
905 size
+= iov
[i
].iov_len
;
907 msg
->msg_name
= NULL
;
908 msg
->msg_namelen
= 0;
909 msg
->msg_control
= NULL
;
910 msg
->msg_controllen
= 0;
911 msg
->msg_iov
= (struct iovec
*)iov
;
912 msg
->msg_iovlen
= nr_segs
;
913 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
915 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
918 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
919 unsigned long nr_segs
, loff_t pos
)
921 struct sock_iocb siocb
, *x
;
926 if (iocb
->ki_nbytes
== 0) /* Match SYS5 behaviour */
930 x
= alloc_sock_iocb(iocb
, &siocb
);
933 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
936 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
937 struct file
*file
, const struct iovec
*iov
,
938 unsigned long nr_segs
)
940 struct socket
*sock
= file
->private_data
;
944 for (i
= 0; i
< nr_segs
; i
++)
945 size
+= iov
[i
].iov_len
;
947 msg
->msg_name
= NULL
;
948 msg
->msg_namelen
= 0;
949 msg
->msg_control
= NULL
;
950 msg
->msg_controllen
= 0;
951 msg
->msg_iov
= (struct iovec
*)iov
;
952 msg
->msg_iovlen
= nr_segs
;
953 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
954 if (sock
->type
== SOCK_SEQPACKET
)
955 msg
->msg_flags
|= MSG_EOR
;
957 return __sock_sendmsg(iocb
, sock
, msg
, size
);
960 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
961 unsigned long nr_segs
, loff_t pos
)
963 struct sock_iocb siocb
, *x
;
968 x
= alloc_sock_iocb(iocb
, &siocb
);
972 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
976 * Atomic setting of ioctl hooks to avoid race
977 * with module unload.
980 static DEFINE_MUTEX(br_ioctl_mutex
);
981 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
983 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
985 mutex_lock(&br_ioctl_mutex
);
986 br_ioctl_hook
= hook
;
987 mutex_unlock(&br_ioctl_mutex
);
989 EXPORT_SYMBOL(brioctl_set
);
991 static DEFINE_MUTEX(vlan_ioctl_mutex
);
992 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
994 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
996 mutex_lock(&vlan_ioctl_mutex
);
997 vlan_ioctl_hook
= hook
;
998 mutex_unlock(&vlan_ioctl_mutex
);
1000 EXPORT_SYMBOL(vlan_ioctl_set
);
1002 static DEFINE_MUTEX(dlci_ioctl_mutex
);
1003 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
1005 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
1007 mutex_lock(&dlci_ioctl_mutex
);
1008 dlci_ioctl_hook
= hook
;
1009 mutex_unlock(&dlci_ioctl_mutex
);
1011 EXPORT_SYMBOL(dlci_ioctl_set
);
1013 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1014 unsigned int cmd
, unsigned long arg
)
1017 void __user
*argp
= (void __user
*)arg
;
1019 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1022 * If this ioctl is unknown try to hand it down
1023 * to the NIC driver.
1025 if (err
== -ENOIOCTLCMD
)
1026 err
= dev_ioctl(net
, cmd
, argp
);
1032 * With an ioctl, arg may well be a user mode pointer, but we don't know
1033 * what to do with it - that's up to the protocol still.
1036 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1038 struct socket
*sock
;
1040 void __user
*argp
= (void __user
*)arg
;
1044 sock
= file
->private_data
;
1047 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
1048 err
= dev_ioctl(net
, cmd
, argp
);
1050 #ifdef CONFIG_WEXT_CORE
1051 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1052 err
= dev_ioctl(net
, cmd
, argp
);
1059 if (get_user(pid
, (int __user
*)argp
))
1061 err
= f_setown(sock
->file
, pid
, 1);
1065 err
= put_user(f_getown(sock
->file
),
1066 (int __user
*)argp
);
1074 request_module("bridge");
1076 mutex_lock(&br_ioctl_mutex
);
1078 err
= br_ioctl_hook(net
, cmd
, argp
);
1079 mutex_unlock(&br_ioctl_mutex
);
1084 if (!vlan_ioctl_hook
)
1085 request_module("8021q");
1087 mutex_lock(&vlan_ioctl_mutex
);
1088 if (vlan_ioctl_hook
)
1089 err
= vlan_ioctl_hook(net
, argp
);
1090 mutex_unlock(&vlan_ioctl_mutex
);
1095 if (!dlci_ioctl_hook
)
1096 request_module("dlci");
1098 mutex_lock(&dlci_ioctl_mutex
);
1099 if (dlci_ioctl_hook
)
1100 err
= dlci_ioctl_hook(cmd
, argp
);
1101 mutex_unlock(&dlci_ioctl_mutex
);
1104 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1110 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1113 struct socket
*sock
= NULL
;
1115 err
= security_socket_create(family
, type
, protocol
, 1);
1119 sock
= sock_alloc();
1126 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1138 EXPORT_SYMBOL(sock_create_lite
);
1140 /* No kernel lock held - perfect */
1141 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1143 unsigned int busy_flag
= 0;
1144 struct socket
*sock
;
1147 * We can't return errors to poll, so it's either yes or no.
1149 sock
= file
->private_data
;
1151 if (sk_can_busy_loop(sock
->sk
)) {
1152 /* this socket can poll_ll so tell the system call */
1153 busy_flag
= POLL_BUSY_LOOP
;
1155 /* once, only if requested by syscall */
1156 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1157 sk_busy_loop(sock
->sk
, 1);
1160 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1163 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1165 struct socket
*sock
= file
->private_data
;
1167 return sock
->ops
->mmap(file
, sock
, vma
);
1170 static int sock_close(struct inode
*inode
, struct file
*filp
)
1172 sock_release(SOCKET_I(inode
));
1177 * Update the socket async list
1179 * Fasync_list locking strategy.
1181 * 1. fasync_list is modified only under process context socket lock
1182 * i.e. under semaphore.
1183 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1184 * or under socket lock
1187 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1189 struct socket
*sock
= filp
->private_data
;
1190 struct sock
*sk
= sock
->sk
;
1191 struct socket_wq
*wq
;
1197 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1198 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1200 if (!wq
->fasync_list
)
1201 sock_reset_flag(sk
, SOCK_FASYNC
);
1203 sock_set_flag(sk
, SOCK_FASYNC
);
1209 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1211 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1213 struct socket_wq
*wq
;
1218 wq
= rcu_dereference(sock
->wq
);
1219 if (!wq
|| !wq
->fasync_list
) {
1224 case SOCK_WAKE_WAITD
:
1225 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1228 case SOCK_WAKE_SPACE
:
1229 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1234 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1237 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1242 EXPORT_SYMBOL(sock_wake_async
);
1244 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1245 struct socket
**res
, int kern
)
1248 struct socket
*sock
;
1249 const struct net_proto_family
*pf
;
1252 * Check protocol is in range
1254 if (family
< 0 || family
>= NPROTO
)
1255 return -EAFNOSUPPORT
;
1256 if (type
< 0 || type
>= SOCK_MAX
)
1261 This uglymoron is moved from INET layer to here to avoid
1262 deadlock in module load.
1264 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1268 printk(KERN_INFO
"%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1274 err
= security_socket_create(family
, type
, protocol
, kern
);
1279 * Allocate the socket and allow the family to set things up. if
1280 * the protocol is 0, the family is instructed to select an appropriate
1283 sock
= sock_alloc();
1285 net_warn_ratelimited("socket: no more sockets\n");
1286 return -ENFILE
; /* Not exactly a match, but its the
1287 closest posix thing */
1292 #ifdef CONFIG_MODULES
1293 /* Attempt to load a protocol module if the find failed.
1295 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1296 * requested real, full-featured networking support upon configuration.
1297 * Otherwise module support will break!
1299 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1300 request_module("net-pf-%d", family
);
1304 pf
= rcu_dereference(net_families
[family
]);
1305 err
= -EAFNOSUPPORT
;
1310 * We will call the ->create function, that possibly is in a loadable
1311 * module, so we have to bump that loadable module refcnt first.
1313 if (!try_module_get(pf
->owner
))
1316 /* Now protected by module ref count */
1319 err
= pf
->create(net
, sock
, protocol
, kern
);
1321 goto out_module_put
;
1324 * Now to bump the refcnt of the [loadable] module that owns this
1325 * socket at sock_release time we decrement its refcnt.
1327 if (!try_module_get(sock
->ops
->owner
))
1328 goto out_module_busy
;
1331 * Now that we're done with the ->create function, the [loadable]
1332 * module can have its refcnt decremented
1334 module_put(pf
->owner
);
1335 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1337 goto out_sock_release
;
1343 err
= -EAFNOSUPPORT
;
1346 module_put(pf
->owner
);
1353 goto out_sock_release
;
1355 EXPORT_SYMBOL(__sock_create
);
1357 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1359 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1361 EXPORT_SYMBOL(sock_create
);
1363 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1365 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1367 EXPORT_SYMBOL(sock_create_kern
);
1369 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1372 struct socket
*sock
;
1375 /* Check the SOCK_* constants for consistency. */
1376 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1377 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1378 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1379 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1381 flags
= type
& ~SOCK_TYPE_MASK
;
1382 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1384 type
&= SOCK_TYPE_MASK
;
1386 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1387 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1389 retval
= sock_create(family
, type
, protocol
, &sock
);
1393 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1398 /* It may be already another descriptor 8) Not kernel problem. */
1407 * Create a pair of connected sockets.
1410 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1411 int __user
*, usockvec
)
1413 struct socket
*sock1
, *sock2
;
1415 struct file
*newfile1
, *newfile2
;
1418 flags
= type
& ~SOCK_TYPE_MASK
;
1419 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1421 type
&= SOCK_TYPE_MASK
;
1423 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1424 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1427 * Obtain the first socket and check if the underlying protocol
1428 * supports the socketpair call.
1431 err
= sock_create(family
, type
, protocol
, &sock1
);
1435 err
= sock_create(family
, type
, protocol
, &sock2
);
1439 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1441 goto out_release_both
;
1443 fd1
= get_unused_fd_flags(flags
);
1444 if (unlikely(fd1
< 0)) {
1446 goto out_release_both
;
1448 fd2
= get_unused_fd_flags(flags
);
1449 if (unlikely(fd2
< 0)) {
1452 goto out_release_both
;
1455 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1456 if (unlikely(IS_ERR(newfile1
))) {
1457 err
= PTR_ERR(newfile1
);
1460 goto out_release_both
;
1463 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1464 if (IS_ERR(newfile2
)) {
1465 err
= PTR_ERR(newfile2
);
1469 sock_release(sock2
);
1473 audit_fd_pair(fd1
, fd2
);
1474 fd_install(fd1
, newfile1
);
1475 fd_install(fd2
, newfile2
);
1476 /* fd1 and fd2 may be already another descriptors.
1477 * Not kernel problem.
1480 err
= put_user(fd1
, &usockvec
[0]);
1482 err
= put_user(fd2
, &usockvec
[1]);
1491 sock_release(sock2
);
1493 sock_release(sock1
);
1499 * Bind a name to a socket. Nothing much to do here since it's
1500 * the protocol's responsibility to handle the local address.
1502 * We move the socket address to kernel space before we call
1503 * the protocol layer (having also checked the address is ok).
1506 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1508 struct socket
*sock
;
1509 struct sockaddr_storage address
;
1510 int err
, fput_needed
;
1512 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1514 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1516 err
= security_socket_bind(sock
,
1517 (struct sockaddr
*)&address
,
1520 err
= sock
->ops
->bind(sock
,
1524 fput_light(sock
->file
, fput_needed
);
1530 * Perform a listen. Basically, we allow the protocol to do anything
1531 * necessary for a listen, and if that works, we mark the socket as
1532 * ready for listening.
1535 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1537 struct socket
*sock
;
1538 int err
, fput_needed
;
1541 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1543 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1544 if ((unsigned int)backlog
> somaxconn
)
1545 backlog
= somaxconn
;
1547 err
= security_socket_listen(sock
, backlog
);
1549 err
= sock
->ops
->listen(sock
, backlog
);
1551 fput_light(sock
->file
, fput_needed
);
1557 * For accept, we attempt to create a new socket, set up the link
1558 * with the client, wake up the client, then return the new
1559 * connected fd. We collect the address of the connector in kernel
1560 * space and move it to user at the very end. This is unclean because
1561 * we open the socket then return an error.
1563 * 1003.1g adds the ability to recvmsg() to query connection pending
1564 * status to recvmsg. We need to add that support in a way thats
1565 * clean when we restucture accept also.
1568 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1569 int __user
*, upeer_addrlen
, int, flags
)
1571 struct socket
*sock
, *newsock
;
1572 struct file
*newfile
;
1573 int err
, len
, newfd
, fput_needed
;
1574 struct sockaddr_storage address
;
1576 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1579 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1580 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1582 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1587 newsock
= sock_alloc();
1591 newsock
->type
= sock
->type
;
1592 newsock
->ops
= sock
->ops
;
1595 * We don't need try_module_get here, as the listening socket (sock)
1596 * has the protocol module (sock->ops->owner) held.
1598 __module_get(newsock
->ops
->owner
);
1600 newfd
= get_unused_fd_flags(flags
);
1601 if (unlikely(newfd
< 0)) {
1603 sock_release(newsock
);
1606 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1607 if (unlikely(IS_ERR(newfile
))) {
1608 err
= PTR_ERR(newfile
);
1609 put_unused_fd(newfd
);
1610 sock_release(newsock
);
1614 err
= security_socket_accept(sock
, newsock
);
1618 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1622 if (upeer_sockaddr
) {
1623 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1625 err
= -ECONNABORTED
;
1628 err
= move_addr_to_user(&address
,
1629 len
, upeer_sockaddr
, upeer_addrlen
);
1634 /* File flags are not inherited via accept() unlike another OSes. */
1636 fd_install(newfd
, newfile
);
1640 fput_light(sock
->file
, fput_needed
);
1645 put_unused_fd(newfd
);
1649 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1650 int __user
*, upeer_addrlen
)
1652 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1656 * Attempt to connect to a socket with the server address. The address
1657 * is in user space so we verify it is OK and move it to kernel space.
1659 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1662 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1663 * other SEQPACKET protocols that take time to connect() as it doesn't
1664 * include the -EINPROGRESS status for such sockets.
1667 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1670 struct socket
*sock
;
1671 struct sockaddr_storage address
;
1672 int err
, fput_needed
;
1674 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1677 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1682 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1686 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1687 sock
->file
->f_flags
);
1689 fput_light(sock
->file
, fput_needed
);
1695 * Get the local address ('name') of a socket object. Move the obtained
1696 * name to user space.
1699 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1700 int __user
*, usockaddr_len
)
1702 struct socket
*sock
;
1703 struct sockaddr_storage address
;
1704 int len
, err
, fput_needed
;
1706 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1710 err
= security_socket_getsockname(sock
);
1714 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1717 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1720 fput_light(sock
->file
, fput_needed
);
1726 * Get the remote address ('name') of a socket object. Move the obtained
1727 * name to user space.
1730 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1731 int __user
*, usockaddr_len
)
1733 struct socket
*sock
;
1734 struct sockaddr_storage address
;
1735 int len
, err
, fput_needed
;
1737 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1739 err
= security_socket_getpeername(sock
);
1741 fput_light(sock
->file
, fput_needed
);
1746 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1749 err
= move_addr_to_user(&address
, len
, usockaddr
,
1751 fput_light(sock
->file
, fput_needed
);
1757 * Send a datagram to a given address. We move the address into kernel
1758 * space and check the user space data area is readable before invoking
1762 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1763 unsigned int, flags
, struct sockaddr __user
*, addr
,
1766 struct socket
*sock
;
1767 struct sockaddr_storage address
;
1775 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1779 iov
.iov_base
= buff
;
1781 msg
.msg_name
= NULL
;
1784 msg
.msg_control
= NULL
;
1785 msg
.msg_controllen
= 0;
1786 msg
.msg_namelen
= 0;
1788 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1791 msg
.msg_name
= (struct sockaddr
*)&address
;
1792 msg
.msg_namelen
= addr_len
;
1794 if (sock
->file
->f_flags
& O_NONBLOCK
)
1795 flags
|= MSG_DONTWAIT
;
1796 msg
.msg_flags
= flags
;
1797 err
= sock_sendmsg(sock
, &msg
, len
);
1800 fput_light(sock
->file
, fput_needed
);
1806 * Send a datagram down a socket.
1809 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1810 unsigned int, flags
)
1812 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1816 * Receive a frame from the socket and optionally record the address of the
1817 * sender. We verify the buffers are writable and if needed move the
1818 * sender address from kernel to user space.
1821 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1822 unsigned int, flags
, struct sockaddr __user
*, addr
,
1823 int __user
*, addr_len
)
1825 struct socket
*sock
;
1828 struct sockaddr_storage address
;
1834 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1838 msg
.msg_control
= NULL
;
1839 msg
.msg_controllen
= 0;
1843 iov
.iov_base
= ubuf
;
1844 /* Save some cycles and don't copy the address if not needed */
1845 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1846 /* We assume all kernel code knows the size of sockaddr_storage */
1847 msg
.msg_namelen
= 0;
1848 if (sock
->file
->f_flags
& O_NONBLOCK
)
1849 flags
|= MSG_DONTWAIT
;
1850 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1852 if (err
>= 0 && addr
!= NULL
) {
1853 err2
= move_addr_to_user(&address
,
1854 msg
.msg_namelen
, addr
, addr_len
);
1859 fput_light(sock
->file
, fput_needed
);
1865 * Receive a datagram from a socket.
1868 asmlinkage
long sys_recv(int fd
, void __user
*ubuf
, size_t size
,
1871 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1875 * Set a socket option. Because we don't know the option lengths we have
1876 * to pass the user mode parameter for the protocols to sort out.
1879 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1880 char __user
*, optval
, int, optlen
)
1882 int err
, fput_needed
;
1883 struct socket
*sock
;
1888 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1890 err
= security_socket_setsockopt(sock
, level
, optname
);
1894 if (level
== SOL_SOCKET
)
1896 sock_setsockopt(sock
, level
, optname
, optval
,
1900 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1903 fput_light(sock
->file
, fput_needed
);
1909 * Get a socket option. Because we don't know the option lengths we have
1910 * to pass a user mode parameter for the protocols to sort out.
1913 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1914 char __user
*, optval
, int __user
*, optlen
)
1916 int err
, fput_needed
;
1917 struct socket
*sock
;
1919 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1921 err
= security_socket_getsockopt(sock
, level
, optname
);
1925 if (level
== SOL_SOCKET
)
1927 sock_getsockopt(sock
, level
, optname
, optval
,
1931 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1934 fput_light(sock
->file
, fput_needed
);
1940 * Shutdown a socket.
1943 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1945 int err
, fput_needed
;
1946 struct socket
*sock
;
1948 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1950 err
= security_socket_shutdown(sock
, how
);
1952 err
= sock
->ops
->shutdown(sock
, how
);
1953 fput_light(sock
->file
, fput_needed
);
1958 /* A couple of helpful macros for getting the address of the 32/64 bit
1959 * fields which are the same type (int / unsigned) on our platforms.
1961 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1962 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1963 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1965 struct used_address
{
1966 struct sockaddr_storage name
;
1967 unsigned int name_len
;
1970 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
1971 struct msghdr __user
*umsg
)
1973 if (copy_from_user(kmsg
, umsg
, sizeof(struct msghdr
)))
1975 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1976 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1980 static int ___sys_sendmsg(struct socket
*sock
, struct msghdr __user
*msg
,
1981 struct msghdr
*msg_sys
, unsigned int flags
,
1982 struct used_address
*used_address
)
1984 struct compat_msghdr __user
*msg_compat
=
1985 (struct compat_msghdr __user
*)msg
;
1986 struct sockaddr_storage address
;
1987 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1988 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1989 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1990 /* 20 is size of ipv6_pktinfo */
1991 unsigned char *ctl_buf
= ctl
;
1992 int err
, ctl_len
, total_len
;
1995 if (MSG_CMSG_COMPAT
& flags
) {
1996 if (get_compat_msghdr(msg_sys
, msg_compat
))
1999 err
= copy_msghdr_from_user(msg_sys
, msg
);
2004 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2006 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2009 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2015 /* This will also move the address data into kernel space */
2016 if (MSG_CMSG_COMPAT
& flags
) {
2017 err
= verify_compat_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2019 err
= verify_iovec(msg_sys
, iov
, &address
, VERIFY_READ
);
2026 if (msg_sys
->msg_controllen
> INT_MAX
)
2028 ctl_len
= msg_sys
->msg_controllen
;
2029 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2031 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2035 ctl_buf
= msg_sys
->msg_control
;
2036 ctl_len
= msg_sys
->msg_controllen
;
2037 } else if (ctl_len
) {
2038 if (ctl_len
> sizeof(ctl
)) {
2039 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2040 if (ctl_buf
== NULL
)
2045 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2046 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2047 * checking falls down on this.
2049 if (copy_from_user(ctl_buf
,
2050 (void __user __force
*)msg_sys
->msg_control
,
2053 msg_sys
->msg_control
= ctl_buf
;
2055 msg_sys
->msg_flags
= flags
;
2057 if (sock
->file
->f_flags
& O_NONBLOCK
)
2058 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2060 * If this is sendmmsg() and current destination address is same as
2061 * previously succeeded address, omit asking LSM's decision.
2062 * used_address->name_len is initialized to UINT_MAX so that the first
2063 * destination address never matches.
2065 if (used_address
&& msg_sys
->msg_name
&&
2066 used_address
->name_len
== msg_sys
->msg_namelen
&&
2067 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2068 used_address
->name_len
)) {
2069 err
= sock_sendmsg_nosec(sock
, msg_sys
, total_len
);
2072 err
= sock_sendmsg(sock
, msg_sys
, total_len
);
2074 * If this is sendmmsg() and sending to current destination address was
2075 * successful, remember it.
2077 if (used_address
&& err
>= 0) {
2078 used_address
->name_len
= msg_sys
->msg_namelen
;
2079 if (msg_sys
->msg_name
)
2080 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2081 used_address
->name_len
);
2086 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2088 if (iov
!= iovstack
)
2095 * BSD sendmsg interface
2098 long __sys_sendmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
2100 int fput_needed
, err
;
2101 struct msghdr msg_sys
;
2102 struct socket
*sock
;
2104 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2108 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
);
2110 fput_light(sock
->file
, fput_needed
);
2115 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct msghdr __user
*, msg
, unsigned int, flags
)
2117 if (flags
& MSG_CMSG_COMPAT
)
2119 return __sys_sendmsg(fd
, msg
, flags
);
2123 * Linux sendmmsg interface
2126 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2129 int fput_needed
, err
, datagrams
;
2130 struct socket
*sock
;
2131 struct mmsghdr __user
*entry
;
2132 struct compat_mmsghdr __user
*compat_entry
;
2133 struct msghdr msg_sys
;
2134 struct used_address used_address
;
2136 if (vlen
> UIO_MAXIOV
)
2141 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2145 used_address
.name_len
= UINT_MAX
;
2147 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2150 while (datagrams
< vlen
) {
2151 if (MSG_CMSG_COMPAT
& flags
) {
2152 err
= ___sys_sendmsg(sock
, (struct msghdr __user
*)compat_entry
,
2153 &msg_sys
, flags
, &used_address
);
2156 err
= __put_user(err
, &compat_entry
->msg_len
);
2159 err
= ___sys_sendmsg(sock
,
2160 (struct msghdr __user
*)entry
,
2161 &msg_sys
, flags
, &used_address
);
2164 err
= put_user(err
, &entry
->msg_len
);
2173 fput_light(sock
->file
, fput_needed
);
2175 /* We only return an error if no datagrams were able to be sent */
2182 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2183 unsigned int, vlen
, unsigned int, flags
)
2185 if (flags
& MSG_CMSG_COMPAT
)
2187 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2190 static int ___sys_recvmsg(struct socket
*sock
, struct msghdr __user
*msg
,
2191 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2193 struct compat_msghdr __user
*msg_compat
=
2194 (struct compat_msghdr __user
*)msg
;
2195 struct iovec iovstack
[UIO_FASTIOV
];
2196 struct iovec
*iov
= iovstack
;
2197 unsigned long cmsg_ptr
;
2198 int err
, total_len
, len
;
2200 /* kernel mode address */
2201 struct sockaddr_storage addr
;
2203 /* user mode address pointers */
2204 struct sockaddr __user
*uaddr
;
2205 int __user
*uaddr_len
;
2207 if (MSG_CMSG_COMPAT
& flags
) {
2208 if (get_compat_msghdr(msg_sys
, msg_compat
))
2211 err
= copy_msghdr_from_user(msg_sys
, msg
);
2216 if (msg_sys
->msg_iovlen
> UIO_FASTIOV
) {
2218 if (msg_sys
->msg_iovlen
> UIO_MAXIOV
)
2221 iov
= kmalloc(msg_sys
->msg_iovlen
* sizeof(struct iovec
),
2227 /* Save the user-mode address (verify_iovec will change the
2228 * kernel msghdr to use the kernel address space)
2230 uaddr
= (__force
void __user
*)msg_sys
->msg_name
;
2231 uaddr_len
= COMPAT_NAMELEN(msg
);
2232 if (MSG_CMSG_COMPAT
& flags
)
2233 err
= verify_compat_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2235 err
= verify_iovec(msg_sys
, iov
, &addr
, VERIFY_WRITE
);
2240 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2241 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2243 /* We assume all kernel code knows the size of sockaddr_storage */
2244 msg_sys
->msg_namelen
= 0;
2246 if (sock
->file
->f_flags
& O_NONBLOCK
)
2247 flags
|= MSG_DONTWAIT
;
2248 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2254 if (uaddr
!= NULL
) {
2255 err
= move_addr_to_user(&addr
,
2256 msg_sys
->msg_namelen
, uaddr
,
2261 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2265 if (MSG_CMSG_COMPAT
& flags
)
2266 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2267 &msg_compat
->msg_controllen
);
2269 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2270 &msg
->msg_controllen
);
2276 if (iov
!= iovstack
)
2283 * BSD recvmsg interface
2286 long __sys_recvmsg(int fd
, struct msghdr __user
*msg
, unsigned flags
)
2288 int fput_needed
, err
;
2289 struct msghdr msg_sys
;
2290 struct socket
*sock
;
2292 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2296 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2298 fput_light(sock
->file
, fput_needed
);
2303 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct msghdr __user
*, msg
,
2304 unsigned int, flags
)
2306 if (flags
& MSG_CMSG_COMPAT
)
2308 return __sys_recvmsg(fd
, msg
, flags
);
2312 * Linux recvmmsg interface
2315 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2316 unsigned int flags
, struct timespec
*timeout
)
2318 int fput_needed
, err
, datagrams
;
2319 struct socket
*sock
;
2320 struct mmsghdr __user
*entry
;
2321 struct compat_mmsghdr __user
*compat_entry
;
2322 struct msghdr msg_sys
;
2323 struct timespec end_time
;
2326 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2332 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2336 err
= sock_error(sock
->sk
);
2341 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2343 while (datagrams
< vlen
) {
2345 * No need to ask LSM for more than the first datagram.
2347 if (MSG_CMSG_COMPAT
& flags
) {
2348 err
= ___sys_recvmsg(sock
, (struct msghdr __user
*)compat_entry
,
2349 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2353 err
= __put_user(err
, &compat_entry
->msg_len
);
2356 err
= ___sys_recvmsg(sock
,
2357 (struct msghdr __user
*)entry
,
2358 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2362 err
= put_user(err
, &entry
->msg_len
);
2370 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2371 if (flags
& MSG_WAITFORONE
)
2372 flags
|= MSG_DONTWAIT
;
2375 ktime_get_ts(timeout
);
2376 *timeout
= timespec_sub(end_time
, *timeout
);
2377 if (timeout
->tv_sec
< 0) {
2378 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2382 /* Timeout, return less than vlen datagrams */
2383 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2387 /* Out of band data, return right away */
2388 if (msg_sys
.msg_flags
& MSG_OOB
)
2393 fput_light(sock
->file
, fput_needed
);
2398 if (datagrams
!= 0) {
2400 * We may return less entries than requested (vlen) if the
2401 * sock is non block and there aren't enough datagrams...
2403 if (err
!= -EAGAIN
) {
2405 * ... or if recvmsg returns an error after we
2406 * received some datagrams, where we record the
2407 * error to return on the next call or if the
2408 * app asks about it using getsockopt(SO_ERROR).
2410 sock
->sk
->sk_err
= -err
;
2419 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2420 unsigned int, vlen
, unsigned int, flags
,
2421 struct timespec __user
*, timeout
)
2424 struct timespec timeout_sys
;
2426 if (flags
& MSG_CMSG_COMPAT
)
2430 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2432 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2435 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2437 if (datagrams
> 0 &&
2438 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2439 datagrams
= -EFAULT
;
2444 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2445 /* Argument list sizes for sys_socketcall */
2446 #define AL(x) ((x) * sizeof(unsigned long))
2447 static const unsigned char nargs
[21] = {
2448 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2449 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2450 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2457 * System call vectors.
2459 * Argument checking cleaned up. Saved 20% in size.
2460 * This function doesn't need to set the kernel lock because
2461 * it is set by the callees.
2464 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2466 unsigned long a
[AUDITSC_ARGS
];
2467 unsigned long a0
, a1
;
2471 if (call
< 1 || call
> SYS_SENDMMSG
)
2475 if (len
> sizeof(a
))
2478 /* copy_from_user should be SMP safe. */
2479 if (copy_from_user(a
, args
, len
))
2482 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2491 err
= sys_socket(a0
, a1
, a
[2]);
2494 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2497 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2500 err
= sys_listen(a0
, a1
);
2503 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2504 (int __user
*)a
[2], 0);
2506 case SYS_GETSOCKNAME
:
2508 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2509 (int __user
*)a
[2]);
2511 case SYS_GETPEERNAME
:
2513 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2514 (int __user
*)a
[2]);
2516 case SYS_SOCKETPAIR
:
2517 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2520 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2523 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2524 (struct sockaddr __user
*)a
[4], a
[5]);
2527 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2530 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2531 (struct sockaddr __user
*)a
[4],
2532 (int __user
*)a
[5]);
2535 err
= sys_shutdown(a0
, a1
);
2537 case SYS_SETSOCKOPT
:
2538 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2540 case SYS_GETSOCKOPT
:
2542 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2543 (int __user
*)a
[4]);
2546 err
= sys_sendmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2549 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2552 err
= sys_recvmsg(a0
, (struct msghdr __user
*)a1
, a
[2]);
2555 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2556 (struct timespec __user
*)a
[4]);
2559 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2560 (int __user
*)a
[2], a
[3]);
2569 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2572 * sock_register - add a socket protocol handler
2573 * @ops: description of protocol
2575 * This function is called by a protocol handler that wants to
2576 * advertise its address family, and have it linked into the
2577 * socket interface. The value ops->family coresponds to the
2578 * socket system call protocol family.
2580 int sock_register(const struct net_proto_family
*ops
)
2584 if (ops
->family
>= NPROTO
) {
2585 printk(KERN_CRIT
"protocol %d >= NPROTO(%d)\n", ops
->family
,
2590 spin_lock(&net_family_lock
);
2591 if (rcu_dereference_protected(net_families
[ops
->family
],
2592 lockdep_is_held(&net_family_lock
)))
2595 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2598 spin_unlock(&net_family_lock
);
2600 printk(KERN_INFO
"NET: Registered protocol family %d\n", ops
->family
);
2603 EXPORT_SYMBOL(sock_register
);
2606 * sock_unregister - remove a protocol handler
2607 * @family: protocol family to remove
2609 * This function is called by a protocol handler that wants to
2610 * remove its address family, and have it unlinked from the
2611 * new socket creation.
2613 * If protocol handler is a module, then it can use module reference
2614 * counts to protect against new references. If protocol handler is not
2615 * a module then it needs to provide its own protection in
2616 * the ops->create routine.
2618 void sock_unregister(int family
)
2620 BUG_ON(family
< 0 || family
>= NPROTO
);
2622 spin_lock(&net_family_lock
);
2623 RCU_INIT_POINTER(net_families
[family
], NULL
);
2624 spin_unlock(&net_family_lock
);
2628 printk(KERN_INFO
"NET: Unregistered protocol family %d\n", family
);
2630 EXPORT_SYMBOL(sock_unregister
);
2632 static int __init
sock_init(void)
2636 * Initialize the network sysctl infrastructure.
2638 err
= net_sysctl_init();
2643 * Initialize skbuff SLAB cache
2648 * Initialize the protocols module.
2653 err
= register_filesystem(&sock_fs_type
);
2656 sock_mnt
= kern_mount(&sock_fs_type
);
2657 if (IS_ERR(sock_mnt
)) {
2658 err
= PTR_ERR(sock_mnt
);
2662 /* The real protocol initialization is performed in later initcalls.
2665 #ifdef CONFIG_NETFILTER
2666 err
= netfilter_init();
2671 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2672 skb_timestamping_init();
2679 unregister_filesystem(&sock_fs_type
);
2684 core_initcall(sock_init
); /* early initcall */
2686 #ifdef CONFIG_PROC_FS
2687 void socket_seq_show(struct seq_file
*seq
)
2692 for_each_possible_cpu(cpu
)
2693 counter
+= per_cpu(sockets_in_use
, cpu
);
2695 /* It can be negative, by the way. 8) */
2699 seq_printf(seq
, "sockets: used %d\n", counter
);
2701 #endif /* CONFIG_PROC_FS */
2703 #ifdef CONFIG_COMPAT
2704 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2705 unsigned int cmd
, void __user
*up
)
2707 mm_segment_t old_fs
= get_fs();
2712 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2715 err
= compat_put_timeval(&ktv
, up
);
2720 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2721 unsigned int cmd
, void __user
*up
)
2723 mm_segment_t old_fs
= get_fs();
2724 struct timespec kts
;
2728 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2731 err
= compat_put_timespec(&kts
, up
);
2736 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2738 struct ifreq __user
*uifr
;
2741 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2742 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2745 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2749 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2755 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2757 struct compat_ifconf ifc32
;
2759 struct ifconf __user
*uifc
;
2760 struct compat_ifreq __user
*ifr32
;
2761 struct ifreq __user
*ifr
;
2765 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2768 memset(&ifc
, 0, sizeof(ifc
));
2769 if (ifc32
.ifcbuf
== 0) {
2773 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2775 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2776 sizeof(struct ifreq
);
2777 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2779 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2780 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2781 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2782 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2788 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2791 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2795 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2799 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2801 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2802 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2803 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2809 if (ifc32
.ifcbuf
== 0) {
2810 /* Translate from 64-bit structure multiple to
2814 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2819 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2825 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2827 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2828 bool convert_in
= false, convert_out
= false;
2829 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2830 struct ethtool_rxnfc __user
*rxnfc
;
2831 struct ifreq __user
*ifr
;
2832 u32 rule_cnt
= 0, actual_rule_cnt
;
2837 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2840 compat_rxnfc
= compat_ptr(data
);
2842 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2845 /* Most ethtool structures are defined without padding.
2846 * Unfortunately struct ethtool_rxnfc is an exception.
2851 case ETHTOOL_GRXCLSRLALL
:
2852 /* Buffer size is variable */
2853 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2855 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2857 buf_size
+= rule_cnt
* sizeof(u32
);
2859 case ETHTOOL_GRXRINGS
:
2860 case ETHTOOL_GRXCLSRLCNT
:
2861 case ETHTOOL_GRXCLSRULE
:
2862 case ETHTOOL_SRXCLSRLINS
:
2865 case ETHTOOL_SRXCLSRLDEL
:
2866 buf_size
+= sizeof(struct ethtool_rxnfc
);
2871 ifr
= compat_alloc_user_space(buf_size
);
2872 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2874 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2877 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2878 &ifr
->ifr_ifru
.ifru_data
))
2882 /* We expect there to be holes between fs.m_ext and
2883 * fs.ring_cookie and at the end of fs, but nowhere else.
2885 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2886 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2887 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2888 sizeof(rxnfc
->fs
.m_ext
));
2890 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2891 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2892 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2893 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2895 if (copy_in_user(rxnfc
, compat_rxnfc
,
2896 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2897 (void __user
*)rxnfc
) ||
2898 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2899 &compat_rxnfc
->fs
.ring_cookie
,
2900 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2901 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2902 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2903 sizeof(rxnfc
->rule_cnt
)))
2907 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2912 if (copy_in_user(compat_rxnfc
, rxnfc
,
2913 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2914 (const void __user
*)rxnfc
) ||
2915 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2916 &rxnfc
->fs
.ring_cookie
,
2917 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2918 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2919 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2920 sizeof(rxnfc
->rule_cnt
)))
2923 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2924 /* As an optimisation, we only copy the actual
2925 * number of rules that the underlying
2926 * function returned. Since Mallory might
2927 * change the rule count in user memory, we
2928 * check that it is less than the rule count
2929 * originally given (as the user buffer size),
2930 * which has been range-checked.
2932 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2934 if (actual_rule_cnt
< rule_cnt
)
2935 rule_cnt
= actual_rule_cnt
;
2936 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2937 &rxnfc
->rule_locs
[0],
2938 rule_cnt
* sizeof(u32
)))
2946 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2949 compat_uptr_t uptr32
;
2950 struct ifreq __user
*uifr
;
2952 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2953 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2956 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2959 uptr
= compat_ptr(uptr32
);
2961 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2964 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2967 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2968 struct compat_ifreq __user
*ifr32
)
2971 struct ifreq __user
*uifr
;
2972 mm_segment_t old_fs
;
2978 case SIOCBONDENSLAVE
:
2979 case SIOCBONDRELEASE
:
2980 case SIOCBONDSETHWADDR
:
2981 case SIOCBONDCHANGEACTIVE
:
2982 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2987 err
= dev_ioctl(net
, cmd
,
2988 (struct ifreq __user __force
*) &kifr
);
2992 case SIOCBONDSLAVEINFOQUERY
:
2993 case SIOCBONDINFOQUERY
:
2994 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2995 if (copy_in_user(&uifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2998 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
3001 datap
= compat_ptr(data
);
3002 if (put_user(datap
, &uifr
->ifr_ifru
.ifru_data
))
3005 return dev_ioctl(net
, cmd
, uifr
);
3007 return -ENOIOCTLCMD
;
3011 static int siocdevprivate_ioctl(struct net
*net
, unsigned int cmd
,
3012 struct compat_ifreq __user
*u_ifreq32
)
3014 struct ifreq __user
*u_ifreq64
;
3015 char tmp_buf
[IFNAMSIZ
];
3016 void __user
*data64
;
3019 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
3022 if (__get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
3024 data64
= compat_ptr(data32
);
3026 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
3028 /* Don't check these user accesses, just let that get trapped
3029 * in the ioctl handler instead.
3031 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
3034 if (__put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
3037 return dev_ioctl(net
, cmd
, u_ifreq64
);
3040 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
3041 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
3043 struct ifreq __user
*uifr
;
3046 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3047 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
3050 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3061 case SIOCGIFBRDADDR
:
3062 case SIOCGIFDSTADDR
:
3063 case SIOCGIFNETMASK
:
3068 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3076 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3077 struct compat_ifreq __user
*uifr32
)
3080 struct compat_ifmap __user
*uifmap32
;
3081 mm_segment_t old_fs
;
3084 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3085 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3086 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3087 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3088 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3089 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3090 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3091 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3097 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3100 if (cmd
== SIOCGIFMAP
&& !err
) {
3101 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3102 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3103 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3104 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3105 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3106 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3107 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3114 static int compat_siocshwtstamp(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3117 compat_uptr_t uptr32
;
3118 struct ifreq __user
*uifr
;
3120 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3121 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
3124 if (get_user(uptr32
, &uifr32
->ifr_data
))
3127 uptr
= compat_ptr(uptr32
);
3129 if (put_user(uptr
, &uifr
->ifr_data
))
3132 return dev_ioctl(net
, SIOCSHWTSTAMP
, uifr
);
3137 struct sockaddr rt_dst
; /* target address */
3138 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3139 struct sockaddr rt_genmask
; /* target network mask (IP) */
3140 unsigned short rt_flags
;
3143 unsigned char rt_tos
;
3144 unsigned char rt_class
;
3146 short rt_metric
; /* +1 for binary compatibility! */
3147 /* char * */ u32 rt_dev
; /* forcing the device at add */
3148 u32 rt_mtu
; /* per route MTU/Window */
3149 u32 rt_window
; /* Window clamping */
3150 unsigned short rt_irtt
; /* Initial RTT */
3153 struct in6_rtmsg32
{
3154 struct in6_addr rtmsg_dst
;
3155 struct in6_addr rtmsg_src
;
3156 struct in6_addr rtmsg_gateway
;
3166 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3167 unsigned int cmd
, void __user
*argp
)
3171 struct in6_rtmsg r6
;
3175 mm_segment_t old_fs
= get_fs();
3177 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3178 struct in6_rtmsg32 __user
*ur6
= argp
;
3179 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3180 3 * sizeof(struct in6_addr
));
3181 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3182 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3183 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3184 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3185 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3186 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3187 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3191 struct rtentry32 __user
*ur4
= argp
;
3192 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3193 3 * sizeof(struct sockaddr
));
3194 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3195 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3196 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3197 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3198 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3199 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3201 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3202 r4
.rt_dev
= (char __user __force
*)devname
;
3216 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3223 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3224 * for some operations; this forces use of the newer bridge-utils that
3225 * use compatible ioctls
3227 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3231 if (get_user(tmp
, argp
))
3233 if (tmp
== BRCTL_GET_VERSION
)
3234 return BRCTL_VERSION
+ 1;
3238 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3239 unsigned int cmd
, unsigned long arg
)
3241 void __user
*argp
= compat_ptr(arg
);
3242 struct sock
*sk
= sock
->sk
;
3243 struct net
*net
= sock_net(sk
);
3245 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3246 return siocdevprivate_ioctl(net
, cmd
, argp
);
3251 return old_bridge_ioctl(argp
);
3253 return dev_ifname32(net
, argp
);
3255 return dev_ifconf(net
, argp
);
3257 return ethtool_ioctl(net
, argp
);
3259 return compat_siocwandev(net
, argp
);
3262 return compat_sioc_ifmap(net
, cmd
, argp
);
3263 case SIOCBONDENSLAVE
:
3264 case SIOCBONDRELEASE
:
3265 case SIOCBONDSETHWADDR
:
3266 case SIOCBONDSLAVEINFOQUERY
:
3267 case SIOCBONDINFOQUERY
:
3268 case SIOCBONDCHANGEACTIVE
:
3269 return bond_ioctl(net
, cmd
, argp
);
3272 return routing_ioctl(net
, sock
, cmd
, argp
);
3274 return do_siocgstamp(net
, sock
, cmd
, argp
);
3276 return do_siocgstampns(net
, sock
, cmd
, argp
);
3278 return compat_siocshwtstamp(net
, argp
);
3290 return sock_ioctl(file
, cmd
, arg
);
3307 case SIOCSIFHWBROADCAST
:
3309 case SIOCGIFBRDADDR
:
3310 case SIOCSIFBRDADDR
:
3311 case SIOCGIFDSTADDR
:
3312 case SIOCSIFDSTADDR
:
3313 case SIOCGIFNETMASK
:
3314 case SIOCSIFNETMASK
:
3325 return dev_ifsioc(net
, sock
, cmd
, argp
);
3331 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3334 return -ENOIOCTLCMD
;
3337 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3340 struct socket
*sock
= file
->private_data
;
3341 int ret
= -ENOIOCTLCMD
;
3348 if (sock
->ops
->compat_ioctl
)
3349 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3351 if (ret
== -ENOIOCTLCMD
&&
3352 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3353 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3355 if (ret
== -ENOIOCTLCMD
)
3356 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3362 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3364 return sock
->ops
->bind(sock
, addr
, addrlen
);
3366 EXPORT_SYMBOL(kernel_bind
);
3368 int kernel_listen(struct socket
*sock
, int backlog
)
3370 return sock
->ops
->listen(sock
, backlog
);
3372 EXPORT_SYMBOL(kernel_listen
);
3374 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3376 struct sock
*sk
= sock
->sk
;
3379 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3384 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3386 sock_release(*newsock
);
3391 (*newsock
)->ops
= sock
->ops
;
3392 __module_get((*newsock
)->ops
->owner
);
3397 EXPORT_SYMBOL(kernel_accept
);
3399 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3402 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3404 EXPORT_SYMBOL(kernel_connect
);
3406 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3409 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3411 EXPORT_SYMBOL(kernel_getsockname
);
3413 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3416 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3418 EXPORT_SYMBOL(kernel_getpeername
);
3420 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3421 char *optval
, int *optlen
)
3423 mm_segment_t oldfs
= get_fs();
3424 char __user
*uoptval
;
3425 int __user
*uoptlen
;
3428 uoptval
= (char __user __force
*) optval
;
3429 uoptlen
= (int __user __force
*) optlen
;
3432 if (level
== SOL_SOCKET
)
3433 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3435 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3440 EXPORT_SYMBOL(kernel_getsockopt
);
3442 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3443 char *optval
, unsigned int optlen
)
3445 mm_segment_t oldfs
= get_fs();
3446 char __user
*uoptval
;
3449 uoptval
= (char __user __force
*) optval
;
3452 if (level
== SOL_SOCKET
)
3453 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3455 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3460 EXPORT_SYMBOL(kernel_setsockopt
);
3462 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3463 size_t size
, int flags
)
3465 if (sock
->ops
->sendpage
)
3466 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3468 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3470 EXPORT_SYMBOL(kernel_sendpage
);
3472 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3474 mm_segment_t oldfs
= get_fs();
3478 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3483 EXPORT_SYMBOL(kernel_sock_ioctl
);
3485 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3487 return sock
->ops
->shutdown(sock
, how
);
3489 EXPORT_SYMBOL(kernel_sock_shutdown
);