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/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
93 #include <asm/uaccess.h>
94 #include <asm/unistd.h>
96 #include <net/compat.h>
98 #include <net/cls_cgroup.h>
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly
;
113 unsigned int sysctl_net_busy_poll __read_mostly
;
116 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
);
117 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
118 unsigned long nr_segs
, loff_t pos
);
119 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
120 unsigned long nr_segs
, loff_t pos
);
121 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
123 static int sock_close(struct inode
*inode
, struct file
*file
);
124 static unsigned int sock_poll(struct file
*file
,
125 struct poll_table_struct
*wait
);
126 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
128 static long compat_sock_ioctl(struct file
*file
,
129 unsigned int cmd
, unsigned long arg
);
131 static int sock_fasync(int fd
, struct file
*filp
, int on
);
132 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
133 int offset
, size_t size
, loff_t
*ppos
, int more
);
134 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
135 struct pipe_inode_info
*pipe
, size_t len
,
139 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
140 * in the operation structures but are done directly via the socketcall() multiplexor.
143 static const struct file_operations socket_file_ops
= {
144 .owner
= THIS_MODULE
,
146 .aio_read
= sock_aio_read
,
147 .aio_write
= sock_aio_write
,
149 .unlocked_ioctl
= sock_ioctl
,
151 .compat_ioctl
= compat_sock_ioctl
,
154 .open
= sock_no_open
, /* special open code to disallow open via /proc */
155 .release
= sock_close
,
156 .fasync
= sock_fasync
,
157 .sendpage
= sock_sendpage
,
158 .splice_write
= generic_splice_sendpage
,
159 .splice_read
= sock_splice_read
,
163 * The protocol list. Each protocol is registered in here.
166 static DEFINE_SPINLOCK(net_family_lock
);
167 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
170 * Statistics counters of the socket lists
173 static DEFINE_PER_CPU(int, sockets_in_use
);
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
192 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
194 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
198 if (copy_from_user(kaddr
, uaddr
, ulen
))
200 return audit_sockaddr(ulen
, kaddr
);
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
220 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
221 void __user
*uaddr
, int __user
*ulen
)
226 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
227 err
= get_user(len
, ulen
);
235 if (audit_sockaddr(klen
, kaddr
))
237 if (copy_to_user(uaddr
, kaddr
, len
))
241 * "fromlen shall refer to the value before truncation.."
244 return __put_user(klen
, ulen
);
247 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
249 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
251 struct socket_alloc
*ei
;
252 struct socket_wq
*wq
;
254 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
257 wq
= kmalloc(sizeof(*wq
), GFP_KERNEL
);
259 kmem_cache_free(sock_inode_cachep
, ei
);
262 init_waitqueue_head(&wq
->wait
);
263 wq
->fasync_list
= NULL
;
264 RCU_INIT_POINTER(ei
->socket
.wq
, wq
);
266 ei
->socket
.state
= SS_UNCONNECTED
;
267 ei
->socket
.flags
= 0;
268 ei
->socket
.ops
= NULL
;
269 ei
->socket
.sk
= NULL
;
270 ei
->socket
.file
= NULL
;
272 return &ei
->vfs_inode
;
275 static void sock_destroy_inode(struct inode
*inode
)
277 struct socket_alloc
*ei
;
278 struct socket_wq
*wq
;
280 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
281 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
283 kmem_cache_free(sock_inode_cachep
, ei
);
286 static void init_once(void *foo
)
288 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
290 inode_init_once(&ei
->vfs_inode
);
293 static int init_inodecache(void)
295 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
296 sizeof(struct socket_alloc
),
298 (SLAB_HWCACHE_ALIGN
|
299 SLAB_RECLAIM_ACCOUNT
|
302 if (sock_inode_cachep
== NULL
)
307 static const struct super_operations sockfs_ops
= {
308 .alloc_inode
= sock_alloc_inode
,
309 .destroy_inode
= sock_destroy_inode
,
310 .statfs
= simple_statfs
,
314 * sockfs_dname() is called from d_path().
316 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
318 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
319 dentry
->d_inode
->i_ino
);
322 static const struct dentry_operations sockfs_dentry_operations
= {
323 .d_dname
= sockfs_dname
,
326 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
327 int flags
, const char *dev_name
, void *data
)
329 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
330 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
333 static struct vfsmount
*sock_mnt __read_mostly
;
335 static struct file_system_type sock_fs_type
= {
337 .mount
= sockfs_mount
,
338 .kill_sb
= kill_anon_super
,
342 * Obtains the first available file descriptor and sets it up for use.
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
358 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
360 struct qstr name
= { .name
= "" };
366 name
.len
= strlen(name
.name
);
367 } else if (sock
->sk
) {
368 name
.name
= sock
->sk
->sk_prot_creator
->name
;
369 name
.len
= strlen(name
.name
);
371 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
372 if (unlikely(!path
.dentry
))
373 return ERR_PTR(-ENOMEM
);
374 path
.mnt
= mntget(sock_mnt
);
376 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
377 SOCK_INODE(sock
)->i_fop
= &socket_file_ops
;
379 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
381 if (unlikely(IS_ERR(file
))) {
382 /* drop dentry, keep inode */
383 ihold(path
.dentry
->d_inode
);
389 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
390 file
->private_data
= sock
;
393 EXPORT_SYMBOL(sock_alloc_file
);
395 static int sock_map_fd(struct socket
*sock
, int flags
)
397 struct file
*newfile
;
398 int fd
= get_unused_fd_flags(flags
);
399 if (unlikely(fd
< 0))
402 newfile
= sock_alloc_file(sock
, flags
, NULL
);
403 if (likely(!IS_ERR(newfile
))) {
404 fd_install(fd
, newfile
);
409 return PTR_ERR(newfile
);
412 struct socket
*sock_from_file(struct file
*file
, int *err
)
414 if (file
->f_op
== &socket_file_ops
)
415 return file
->private_data
; /* set in sock_map_fd */
420 EXPORT_SYMBOL(sock_from_file
);
423 * sockfd_lookup - Go from a file number to its socket slot
425 * @err: pointer to an error code return
427 * The file handle passed in is locked and the socket it is bound
428 * too is returned. If an error occurs the err pointer is overwritten
429 * with a negative errno code and NULL is returned. The function checks
430 * for both invalid handles and passing a handle which is not a socket.
432 * On a success the socket object pointer is returned.
435 struct socket
*sockfd_lookup(int fd
, int *err
)
446 sock
= sock_from_file(file
, err
);
451 EXPORT_SYMBOL(sockfd_lookup
);
453 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
455 struct fd f
= fdget(fd
);
460 sock
= sock_from_file(f
.file
, err
);
462 *fput_needed
= f
.flags
;
470 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
471 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
472 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
473 static ssize_t
sockfs_getxattr(struct dentry
*dentry
,
474 const char *name
, void *value
, size_t size
)
476 const char *proto_name
;
481 if (!strncmp(name
, XATTR_NAME_SOCKPROTONAME
, XATTR_NAME_SOCKPROTONAME_LEN
)) {
482 proto_name
= dentry
->d_name
.name
;
483 proto_size
= strlen(proto_name
);
487 if (proto_size
+ 1 > size
)
490 strncpy(value
, proto_name
, proto_size
+ 1);
492 error
= proto_size
+ 1;
499 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
505 len
= security_inode_listsecurity(dentry
->d_inode
, buffer
, size
);
515 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
520 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
527 static const struct inode_operations sockfs_inode_ops
= {
528 .getxattr
= sockfs_getxattr
,
529 .listxattr
= sockfs_listxattr
,
533 * sock_alloc - allocate a socket
535 * Allocate a new inode and socket object. The two are bound together
536 * and initialised. The socket is then returned. If we are out of inodes
540 static struct socket
*sock_alloc(void)
545 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
549 sock
= SOCKET_I(inode
);
551 kmemcheck_annotate_bitfield(sock
, type
);
552 inode
->i_ino
= get_next_ino();
553 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
554 inode
->i_uid
= current_fsuid();
555 inode
->i_gid
= current_fsgid();
556 inode
->i_op
= &sockfs_inode_ops
;
558 this_cpu_add(sockets_in_use
, 1);
563 * In theory you can't get an open on this inode, but /proc provides
564 * a back door. Remember to keep it shut otherwise you'll let the
565 * creepy crawlies in.
568 static int sock_no_open(struct inode
*irrelevant
, struct file
*dontcare
)
573 const struct file_operations bad_sock_fops
= {
574 .owner
= THIS_MODULE
,
575 .open
= sock_no_open
,
576 .llseek
= noop_llseek
,
580 * sock_release - close a socket
581 * @sock: socket to close
583 * The socket is released from the protocol stack if it has a release
584 * callback, and the inode is then released if the socket is bound to
585 * an inode not a file.
588 void sock_release(struct socket
*sock
)
591 struct module
*owner
= sock
->ops
->owner
;
593 sock
->ops
->release(sock
);
598 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
599 pr_err("%s: fasync list not empty!\n", __func__
);
601 if (test_bit(SOCK_EXTERNALLY_ALLOCATED
, &sock
->flags
))
604 this_cpu_sub(sockets_in_use
, 1);
606 iput(SOCK_INODE(sock
));
611 EXPORT_SYMBOL(sock_release
);
613 void __sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
)
615 u8 flags
= *tx_flags
;
617 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
618 flags
|= SKBTX_HW_TSTAMP
;
620 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
621 flags
|= SKBTX_SW_TSTAMP
;
623 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
624 flags
|= SKBTX_SCHED_TSTAMP
;
626 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_ACK
)
627 flags
|= SKBTX_ACK_TSTAMP
;
631 EXPORT_SYMBOL(__sock_tx_timestamp
);
633 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
634 struct msghdr
*msg
, size_t size
)
636 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
643 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
646 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
647 struct msghdr
*msg
, size_t size
)
649 int err
= security_socket_sendmsg(sock
, msg
, size
);
651 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
654 static int do_sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
655 size_t size
, bool nosec
)
658 struct sock_iocb siocb
;
661 init_sync_kiocb(&iocb
, NULL
);
662 iocb
.private = &siocb
;
663 ret
= nosec
? __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
) :
664 __sock_sendmsg(&iocb
, sock
, msg
, size
);
665 if (-EIOCBQUEUED
== ret
)
666 ret
= wait_on_sync_kiocb(&iocb
);
670 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
672 return do_sock_sendmsg(sock
, msg
, size
, false);
674 EXPORT_SYMBOL(sock_sendmsg
);
676 static int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
678 return do_sock_sendmsg(sock
, msg
, size
, true);
681 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
682 struct kvec
*vec
, size_t num
, size_t size
)
684 mm_segment_t oldfs
= get_fs();
689 * the following is safe, since for compiler definitions of kvec and
690 * iovec are identical, yielding the same in-core layout and alignment
692 msg
->msg_iov
= (struct iovec
*)vec
;
693 msg
->msg_iovlen
= num
;
694 result
= sock_sendmsg(sock
, msg
, size
);
698 EXPORT_SYMBOL(kernel_sendmsg
);
701 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
703 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
706 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
707 struct scm_timestamping tss
;
709 struct skb_shared_hwtstamps
*shhwtstamps
=
712 /* Race occurred between timestamp enabling and packet
713 receiving. Fill in the current time for now. */
714 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
715 __net_timestamp(skb
);
717 if (need_software_tstamp
) {
718 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
720 skb_get_timestamp(skb
, &tv
);
721 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
725 skb_get_timestampns(skb
, &ts
);
726 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
731 memset(&tss
, 0, sizeof(tss
));
732 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
733 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
736 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
737 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2))
740 put_cmsg(msg
, SOL_SOCKET
,
741 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
743 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
745 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
750 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
752 if (!skb
->wifi_acked_valid
)
755 ack
= skb
->wifi_acked
;
757 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
759 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
761 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
764 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
765 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
766 sizeof(__u32
), &skb
->dropcount
);
769 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
772 sock_recv_timestamp(msg
, sk
, skb
);
773 sock_recv_drops(msg
, sk
, skb
);
775 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
777 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
778 struct msghdr
*msg
, size_t size
, int flags
)
780 struct sock_iocb
*si
= kiocb_to_siocb(iocb
);
788 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
791 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
792 struct msghdr
*msg
, size_t size
, int flags
)
794 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
796 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
799 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
800 size_t size
, int flags
)
803 struct sock_iocb siocb
;
806 init_sync_kiocb(&iocb
, NULL
);
807 iocb
.private = &siocb
;
808 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
809 if (-EIOCBQUEUED
== ret
)
810 ret
= wait_on_sync_kiocb(&iocb
);
813 EXPORT_SYMBOL(sock_recvmsg
);
815 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
816 size_t size
, int flags
)
819 struct sock_iocb siocb
;
822 init_sync_kiocb(&iocb
, NULL
);
823 iocb
.private = &siocb
;
824 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
825 if (-EIOCBQUEUED
== ret
)
826 ret
= wait_on_sync_kiocb(&iocb
);
831 * kernel_recvmsg - Receive a message from a socket (kernel space)
832 * @sock: The socket to receive the message from
833 * @msg: Received message
834 * @vec: Input s/g array for message data
835 * @num: Size of input s/g array
836 * @size: Number of bytes to read
837 * @flags: Message flags (MSG_DONTWAIT, etc...)
839 * On return the msg structure contains the scatter/gather array passed in the
840 * vec argument. The array is modified so that it consists of the unfilled
841 * portion of the original array.
843 * The returned value is the total number of bytes received, or an error.
845 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
846 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
848 mm_segment_t oldfs
= get_fs();
853 * the following is safe, since for compiler definitions of kvec and
854 * iovec are identical, yielding the same in-core layout and alignment
856 msg
->msg_iov
= (struct iovec
*)vec
, msg
->msg_iovlen
= num
;
857 result
= sock_recvmsg(sock
, msg
, size
, flags
);
861 EXPORT_SYMBOL(kernel_recvmsg
);
863 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
864 int offset
, size_t size
, loff_t
*ppos
, int more
)
869 sock
= file
->private_data
;
871 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
872 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
875 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
878 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
879 struct pipe_inode_info
*pipe
, size_t len
,
882 struct socket
*sock
= file
->private_data
;
884 if (unlikely(!sock
->ops
->splice_read
))
887 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
890 static struct sock_iocb
*alloc_sock_iocb(struct kiocb
*iocb
,
891 struct sock_iocb
*siocb
)
893 if (!is_sync_kiocb(iocb
))
897 iocb
->private = siocb
;
901 static ssize_t
do_sock_read(struct msghdr
*msg
, struct kiocb
*iocb
,
902 struct file
*file
, const struct iovec
*iov
,
903 unsigned long nr_segs
)
905 struct socket
*sock
= file
->private_data
;
909 for (i
= 0; i
< nr_segs
; i
++)
910 size
+= iov
[i
].iov_len
;
912 msg
->msg_name
= NULL
;
913 msg
->msg_namelen
= 0;
914 msg
->msg_control
= NULL
;
915 msg
->msg_controllen
= 0;
916 msg
->msg_iov
= (struct iovec
*)iov
;
917 msg
->msg_iovlen
= nr_segs
;
918 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
920 return __sock_recvmsg(iocb
, sock
, msg
, size
, msg
->msg_flags
);
923 static ssize_t
sock_aio_read(struct kiocb
*iocb
, const struct iovec
*iov
,
924 unsigned long nr_segs
, loff_t pos
)
926 struct sock_iocb siocb
, *x
;
931 if (iocb
->ki_nbytes
== 0) /* Match SYS5 behaviour */
935 x
= alloc_sock_iocb(iocb
, &siocb
);
938 return do_sock_read(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
941 static ssize_t
do_sock_write(struct msghdr
*msg
, struct kiocb
*iocb
,
942 struct file
*file
, const struct iovec
*iov
,
943 unsigned long nr_segs
)
945 struct socket
*sock
= file
->private_data
;
949 for (i
= 0; i
< nr_segs
; i
++)
950 size
+= iov
[i
].iov_len
;
952 msg
->msg_name
= NULL
;
953 msg
->msg_namelen
= 0;
954 msg
->msg_control
= NULL
;
955 msg
->msg_controllen
= 0;
956 msg
->msg_iov
= (struct iovec
*)iov
;
957 msg
->msg_iovlen
= nr_segs
;
958 msg
->msg_flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
959 if (sock
->type
== SOCK_SEQPACKET
)
960 msg
->msg_flags
|= MSG_EOR
;
962 return __sock_sendmsg(iocb
, sock
, msg
, size
);
965 static ssize_t
sock_aio_write(struct kiocb
*iocb
, const struct iovec
*iov
,
966 unsigned long nr_segs
, loff_t pos
)
968 struct sock_iocb siocb
, *x
;
973 x
= alloc_sock_iocb(iocb
, &siocb
);
977 return do_sock_write(&x
->async_msg
, iocb
, iocb
->ki_filp
, iov
, nr_segs
);
981 * Atomic setting of ioctl hooks to avoid race
982 * with module unload.
985 static DEFINE_MUTEX(br_ioctl_mutex
);
986 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
988 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
990 mutex_lock(&br_ioctl_mutex
);
991 br_ioctl_hook
= hook
;
992 mutex_unlock(&br_ioctl_mutex
);
994 EXPORT_SYMBOL(brioctl_set
);
996 static DEFINE_MUTEX(vlan_ioctl_mutex
);
997 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
999 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1001 mutex_lock(&vlan_ioctl_mutex
);
1002 vlan_ioctl_hook
= hook
;
1003 mutex_unlock(&vlan_ioctl_mutex
);
1005 EXPORT_SYMBOL(vlan_ioctl_set
);
1007 static DEFINE_MUTEX(dlci_ioctl_mutex
);
1008 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
1010 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
1012 mutex_lock(&dlci_ioctl_mutex
);
1013 dlci_ioctl_hook
= hook
;
1014 mutex_unlock(&dlci_ioctl_mutex
);
1016 EXPORT_SYMBOL(dlci_ioctl_set
);
1018 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1019 unsigned int cmd
, unsigned long arg
)
1022 void __user
*argp
= (void __user
*)arg
;
1024 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1027 * If this ioctl is unknown try to hand it down
1028 * to the NIC driver.
1030 if (err
== -ENOIOCTLCMD
)
1031 err
= dev_ioctl(net
, cmd
, argp
);
1037 * With an ioctl, arg may well be a user mode pointer, but we don't know
1038 * what to do with it - that's up to the protocol still.
1041 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1043 struct socket
*sock
;
1045 void __user
*argp
= (void __user
*)arg
;
1049 sock
= file
->private_data
;
1052 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
1053 err
= dev_ioctl(net
, cmd
, argp
);
1055 #ifdef CONFIG_WEXT_CORE
1056 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1057 err
= dev_ioctl(net
, cmd
, argp
);
1064 if (get_user(pid
, (int __user
*)argp
))
1066 f_setown(sock
->file
, pid
, 1);
1071 err
= put_user(f_getown(sock
->file
),
1072 (int __user
*)argp
);
1080 request_module("bridge");
1082 mutex_lock(&br_ioctl_mutex
);
1084 err
= br_ioctl_hook(net
, cmd
, argp
);
1085 mutex_unlock(&br_ioctl_mutex
);
1090 if (!vlan_ioctl_hook
)
1091 request_module("8021q");
1093 mutex_lock(&vlan_ioctl_mutex
);
1094 if (vlan_ioctl_hook
)
1095 err
= vlan_ioctl_hook(net
, argp
);
1096 mutex_unlock(&vlan_ioctl_mutex
);
1101 if (!dlci_ioctl_hook
)
1102 request_module("dlci");
1104 mutex_lock(&dlci_ioctl_mutex
);
1105 if (dlci_ioctl_hook
)
1106 err
= dlci_ioctl_hook(cmd
, argp
);
1107 mutex_unlock(&dlci_ioctl_mutex
);
1110 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1116 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1119 struct socket
*sock
= NULL
;
1121 err
= security_socket_create(family
, type
, protocol
, 1);
1125 sock
= sock_alloc();
1132 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1144 EXPORT_SYMBOL(sock_create_lite
);
1146 /* No kernel lock held - perfect */
1147 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1149 unsigned int busy_flag
= 0;
1150 struct socket
*sock
;
1153 * We can't return errors to poll, so it's either yes or no.
1155 sock
= file
->private_data
;
1157 if (sk_can_busy_loop(sock
->sk
)) {
1158 /* this socket can poll_ll so tell the system call */
1159 busy_flag
= POLL_BUSY_LOOP
;
1161 /* once, only if requested by syscall */
1162 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1163 sk_busy_loop(sock
->sk
, 1);
1166 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1169 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1171 struct socket
*sock
= file
->private_data
;
1173 return sock
->ops
->mmap(file
, sock
, vma
);
1176 static int sock_close(struct inode
*inode
, struct file
*filp
)
1178 sock_release(SOCKET_I(inode
));
1183 * Update the socket async list
1185 * Fasync_list locking strategy.
1187 * 1. fasync_list is modified only under process context socket lock
1188 * i.e. under semaphore.
1189 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1190 * or under socket lock
1193 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1195 struct socket
*sock
= filp
->private_data
;
1196 struct sock
*sk
= sock
->sk
;
1197 struct socket_wq
*wq
;
1203 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1204 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1206 if (!wq
->fasync_list
)
1207 sock_reset_flag(sk
, SOCK_FASYNC
);
1209 sock_set_flag(sk
, SOCK_FASYNC
);
1215 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1217 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1219 struct socket_wq
*wq
;
1224 wq
= rcu_dereference(sock
->wq
);
1225 if (!wq
|| !wq
->fasync_list
) {
1230 case SOCK_WAKE_WAITD
:
1231 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1234 case SOCK_WAKE_SPACE
:
1235 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1240 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1243 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1248 EXPORT_SYMBOL(sock_wake_async
);
1250 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1251 struct socket
**res
, int kern
)
1254 struct socket
*sock
;
1255 const struct net_proto_family
*pf
;
1258 * Check protocol is in range
1260 if (family
< 0 || family
>= NPROTO
)
1261 return -EAFNOSUPPORT
;
1262 if (type
< 0 || type
>= SOCK_MAX
)
1267 This uglymoron is moved from INET layer to here to avoid
1268 deadlock in module load.
1270 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1274 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1280 err
= security_socket_create(family
, type
, protocol
, kern
);
1285 * Allocate the socket and allow the family to set things up. if
1286 * the protocol is 0, the family is instructed to select an appropriate
1289 sock
= sock_alloc();
1291 net_warn_ratelimited("socket: no more sockets\n");
1292 return -ENFILE
; /* Not exactly a match, but its the
1293 closest posix thing */
1298 #ifdef CONFIG_MODULES
1299 /* Attempt to load a protocol module if the find failed.
1301 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1302 * requested real, full-featured networking support upon configuration.
1303 * Otherwise module support will break!
1305 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1306 request_module("net-pf-%d", family
);
1310 pf
= rcu_dereference(net_families
[family
]);
1311 err
= -EAFNOSUPPORT
;
1316 * We will call the ->create function, that possibly is in a loadable
1317 * module, so we have to bump that loadable module refcnt first.
1319 if (!try_module_get(pf
->owner
))
1322 /* Now protected by module ref count */
1325 err
= pf
->create(net
, sock
, protocol
, kern
);
1327 goto out_module_put
;
1330 * Now to bump the refcnt of the [loadable] module that owns this
1331 * socket at sock_release time we decrement its refcnt.
1333 if (!try_module_get(sock
->ops
->owner
))
1334 goto out_module_busy
;
1337 * Now that we're done with the ->create function, the [loadable]
1338 * module can have its refcnt decremented
1340 module_put(pf
->owner
);
1341 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1343 goto out_sock_release
;
1349 err
= -EAFNOSUPPORT
;
1352 module_put(pf
->owner
);
1359 goto out_sock_release
;
1361 EXPORT_SYMBOL(__sock_create
);
1363 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1365 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1367 EXPORT_SYMBOL(sock_create
);
1369 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1371 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1373 EXPORT_SYMBOL(sock_create_kern
);
1375 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1378 struct socket
*sock
;
1381 /* Check the SOCK_* constants for consistency. */
1382 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1383 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1384 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1385 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1387 flags
= type
& ~SOCK_TYPE_MASK
;
1388 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1390 type
&= SOCK_TYPE_MASK
;
1392 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1393 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1395 retval
= sock_create(family
, type
, protocol
, &sock
);
1399 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1404 /* It may be already another descriptor 8) Not kernel problem. */
1413 * Create a pair of connected sockets.
1416 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1417 int __user
*, usockvec
)
1419 struct socket
*sock1
, *sock2
;
1421 struct file
*newfile1
, *newfile2
;
1424 flags
= type
& ~SOCK_TYPE_MASK
;
1425 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1427 type
&= SOCK_TYPE_MASK
;
1429 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1430 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1433 * Obtain the first socket and check if the underlying protocol
1434 * supports the socketpair call.
1437 err
= sock_create(family
, type
, protocol
, &sock1
);
1441 err
= sock_create(family
, type
, protocol
, &sock2
);
1445 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1447 goto out_release_both
;
1449 fd1
= get_unused_fd_flags(flags
);
1450 if (unlikely(fd1
< 0)) {
1452 goto out_release_both
;
1455 fd2
= get_unused_fd_flags(flags
);
1456 if (unlikely(fd2
< 0)) {
1458 goto out_put_unused_1
;
1461 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1462 if (unlikely(IS_ERR(newfile1
))) {
1463 err
= PTR_ERR(newfile1
);
1464 goto out_put_unused_both
;
1467 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1468 if (IS_ERR(newfile2
)) {
1469 err
= PTR_ERR(newfile2
);
1473 err
= put_user(fd1
, &usockvec
[0]);
1477 err
= put_user(fd2
, &usockvec
[1]);
1481 audit_fd_pair(fd1
, fd2
);
1483 fd_install(fd1
, newfile1
);
1484 fd_install(fd2
, newfile2
);
1485 /* fd1 and fd2 may be already another descriptors.
1486 * Not kernel problem.
1502 sock_release(sock2
);
1505 out_put_unused_both
:
1510 sock_release(sock2
);
1512 sock_release(sock1
);
1518 * Bind a name to a socket. Nothing much to do here since it's
1519 * the protocol's responsibility to handle the local address.
1521 * We move the socket address to kernel space before we call
1522 * the protocol layer (having also checked the address is ok).
1525 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1527 struct socket
*sock
;
1528 struct sockaddr_storage address
;
1529 int err
, fput_needed
;
1531 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1533 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1535 err
= security_socket_bind(sock
,
1536 (struct sockaddr
*)&address
,
1539 err
= sock
->ops
->bind(sock
,
1543 fput_light(sock
->file
, fput_needed
);
1549 * Perform a listen. Basically, we allow the protocol to do anything
1550 * necessary for a listen, and if that works, we mark the socket as
1551 * ready for listening.
1554 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1556 struct socket
*sock
;
1557 int err
, fput_needed
;
1560 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1562 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1563 if ((unsigned int)backlog
> somaxconn
)
1564 backlog
= somaxconn
;
1566 err
= security_socket_listen(sock
, backlog
);
1568 err
= sock
->ops
->listen(sock
, backlog
);
1570 fput_light(sock
->file
, fput_needed
);
1576 * For accept, we attempt to create a new socket, set up the link
1577 * with the client, wake up the client, then return the new
1578 * connected fd. We collect the address of the connector in kernel
1579 * space and move it to user at the very end. This is unclean because
1580 * we open the socket then return an error.
1582 * 1003.1g adds the ability to recvmsg() to query connection pending
1583 * status to recvmsg. We need to add that support in a way thats
1584 * clean when we restucture accept also.
1587 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1588 int __user
*, upeer_addrlen
, int, flags
)
1590 struct socket
*sock
, *newsock
;
1591 struct file
*newfile
;
1592 int err
, len
, newfd
, fput_needed
;
1593 struct sockaddr_storage address
;
1595 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1598 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1599 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1601 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1606 newsock
= sock_alloc();
1610 newsock
->type
= sock
->type
;
1611 newsock
->ops
= sock
->ops
;
1614 * We don't need try_module_get here, as the listening socket (sock)
1615 * has the protocol module (sock->ops->owner) held.
1617 __module_get(newsock
->ops
->owner
);
1619 newfd
= get_unused_fd_flags(flags
);
1620 if (unlikely(newfd
< 0)) {
1622 sock_release(newsock
);
1625 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1626 if (unlikely(IS_ERR(newfile
))) {
1627 err
= PTR_ERR(newfile
);
1628 put_unused_fd(newfd
);
1629 sock_release(newsock
);
1633 err
= security_socket_accept(sock
, newsock
);
1637 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1641 if (upeer_sockaddr
) {
1642 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1644 err
= -ECONNABORTED
;
1647 err
= move_addr_to_user(&address
,
1648 len
, upeer_sockaddr
, upeer_addrlen
);
1653 /* File flags are not inherited via accept() unlike another OSes. */
1655 fd_install(newfd
, newfile
);
1659 fput_light(sock
->file
, fput_needed
);
1664 put_unused_fd(newfd
);
1668 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1669 int __user
*, upeer_addrlen
)
1671 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1675 * Attempt to connect to a socket with the server address. The address
1676 * is in user space so we verify it is OK and move it to kernel space.
1678 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1681 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1682 * other SEQPACKET protocols that take time to connect() as it doesn't
1683 * include the -EINPROGRESS status for such sockets.
1686 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1689 struct socket
*sock
;
1690 struct sockaddr_storage address
;
1691 int err
, fput_needed
;
1693 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1696 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1701 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1705 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1706 sock
->file
->f_flags
);
1708 fput_light(sock
->file
, fput_needed
);
1714 * Get the local address ('name') of a socket object. Move the obtained
1715 * name to user space.
1718 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1719 int __user
*, usockaddr_len
)
1721 struct socket
*sock
;
1722 struct sockaddr_storage address
;
1723 int len
, err
, fput_needed
;
1725 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1729 err
= security_socket_getsockname(sock
);
1733 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1736 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1739 fput_light(sock
->file
, fput_needed
);
1745 * Get the remote address ('name') of a socket object. Move the obtained
1746 * name to user space.
1749 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1750 int __user
*, usockaddr_len
)
1752 struct socket
*sock
;
1753 struct sockaddr_storage address
;
1754 int len
, err
, fput_needed
;
1756 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1758 err
= security_socket_getpeername(sock
);
1760 fput_light(sock
->file
, fput_needed
);
1765 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1768 err
= move_addr_to_user(&address
, len
, usockaddr
,
1770 fput_light(sock
->file
, fput_needed
);
1776 * Send a datagram to a given address. We move the address into kernel
1777 * space and check the user space data area is readable before invoking
1781 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1782 unsigned int, flags
, struct sockaddr __user
*, addr
,
1785 struct socket
*sock
;
1786 struct sockaddr_storage address
;
1794 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1798 iov
.iov_base
= buff
;
1800 msg
.msg_name
= NULL
;
1803 msg
.msg_control
= NULL
;
1804 msg
.msg_controllen
= 0;
1805 msg
.msg_namelen
= 0;
1807 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1810 msg
.msg_name
= (struct sockaddr
*)&address
;
1811 msg
.msg_namelen
= addr_len
;
1813 if (sock
->file
->f_flags
& O_NONBLOCK
)
1814 flags
|= MSG_DONTWAIT
;
1815 msg
.msg_flags
= flags
;
1816 err
= sock_sendmsg(sock
, &msg
, len
);
1819 fput_light(sock
->file
, fput_needed
);
1825 * Send a datagram down a socket.
1828 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1829 unsigned int, flags
)
1831 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1835 * Receive a frame from the socket and optionally record the address of the
1836 * sender. We verify the buffers are writable and if needed move the
1837 * sender address from kernel to user space.
1840 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1841 unsigned int, flags
, struct sockaddr __user
*, addr
,
1842 int __user
*, addr_len
)
1844 struct socket
*sock
;
1847 struct sockaddr_storage address
;
1853 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1857 msg
.msg_control
= NULL
;
1858 msg
.msg_controllen
= 0;
1862 iov
.iov_base
= ubuf
;
1863 /* Save some cycles and don't copy the address if not needed */
1864 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1865 /* We assume all kernel code knows the size of sockaddr_storage */
1866 msg
.msg_namelen
= 0;
1867 if (sock
->file
->f_flags
& O_NONBLOCK
)
1868 flags
|= MSG_DONTWAIT
;
1869 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1871 if (err
>= 0 && addr
!= NULL
) {
1872 err2
= move_addr_to_user(&address
,
1873 msg
.msg_namelen
, addr
, addr_len
);
1878 fput_light(sock
->file
, fput_needed
);
1884 * Receive a datagram from a socket.
1887 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1888 unsigned int, flags
)
1890 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1894 * Set a socket option. Because we don't know the option lengths we have
1895 * to pass the user mode parameter for the protocols to sort out.
1898 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1899 char __user
*, optval
, int, optlen
)
1901 int err
, fput_needed
;
1902 struct socket
*sock
;
1907 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1909 err
= security_socket_setsockopt(sock
, level
, optname
);
1913 if (level
== SOL_SOCKET
)
1915 sock_setsockopt(sock
, level
, optname
, optval
,
1919 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1922 fput_light(sock
->file
, fput_needed
);
1928 * Get a socket option. Because we don't know the option lengths we have
1929 * to pass a user mode parameter for the protocols to sort out.
1932 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1933 char __user
*, optval
, int __user
*, optlen
)
1935 int err
, fput_needed
;
1936 struct socket
*sock
;
1938 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1940 err
= security_socket_getsockopt(sock
, level
, optname
);
1944 if (level
== SOL_SOCKET
)
1946 sock_getsockopt(sock
, level
, optname
, optval
,
1950 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1953 fput_light(sock
->file
, fput_needed
);
1959 * Shutdown a socket.
1962 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1964 int err
, fput_needed
;
1965 struct socket
*sock
;
1967 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1969 err
= security_socket_shutdown(sock
, how
);
1971 err
= sock
->ops
->shutdown(sock
, how
);
1972 fput_light(sock
->file
, fput_needed
);
1977 /* A couple of helpful macros for getting the address of the 32/64 bit
1978 * fields which are the same type (int / unsigned) on our platforms.
1980 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1981 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1982 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1984 struct used_address
{
1985 struct sockaddr_storage name
;
1986 unsigned int name_len
;
1989 static ssize_t
copy_msghdr_from_user(struct msghdr
*kmsg
,
1990 struct user_msghdr __user
*umsg
,
1991 struct sockaddr __user
**save_addr
,
1994 struct sockaddr __user
*uaddr
;
1995 struct iovec __user
*uiov
;
1998 if (!access_ok(VERIFY_READ
, umsg
, sizeof(*umsg
)) ||
1999 __get_user(uaddr
, &umsg
->msg_name
) ||
2000 __get_user(kmsg
->msg_namelen
, &umsg
->msg_namelen
) ||
2001 __get_user(uiov
, &umsg
->msg_iov
) ||
2002 __get_user(kmsg
->msg_iovlen
, &umsg
->msg_iovlen
) ||
2003 __get_user(kmsg
->msg_control
, &umsg
->msg_control
) ||
2004 __get_user(kmsg
->msg_controllen
, &umsg
->msg_controllen
) ||
2005 __get_user(kmsg
->msg_flags
, &umsg
->msg_flags
))
2009 kmsg
->msg_namelen
= 0;
2011 if (kmsg
->msg_namelen
< 0)
2014 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2015 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2020 if (uaddr
&& kmsg
->msg_namelen
) {
2022 err
= move_addr_to_kernel(uaddr
, kmsg
->msg_namelen
,
2028 kmsg
->msg_name
= NULL
;
2029 kmsg
->msg_namelen
= 0;
2032 if (kmsg
->msg_iovlen
> UIO_MAXIOV
)
2035 err
= rw_copy_check_uvector(save_addr
? READ
: WRITE
,
2036 uiov
, kmsg
->msg_iovlen
,
2037 UIO_FASTIOV
, *iov
, iov
);
2039 kmsg
->msg_iov
= *iov
;
2043 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2044 struct msghdr
*msg_sys
, unsigned int flags
,
2045 struct used_address
*used_address
)
2047 struct compat_msghdr __user
*msg_compat
=
2048 (struct compat_msghdr __user
*)msg
;
2049 struct sockaddr_storage address
;
2050 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2051 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2052 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
2053 /* 20 is size of ipv6_pktinfo */
2054 unsigned char *ctl_buf
= ctl
;
2055 int ctl_len
, total_len
;
2058 msg_sys
->msg_name
= &address
;
2060 if (MSG_CMSG_COMPAT
& flags
)
2061 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
2063 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
2070 if (msg_sys
->msg_controllen
> INT_MAX
)
2072 ctl_len
= msg_sys
->msg_controllen
;
2073 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2075 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2079 ctl_buf
= msg_sys
->msg_control
;
2080 ctl_len
= msg_sys
->msg_controllen
;
2081 } else if (ctl_len
) {
2082 if (ctl_len
> sizeof(ctl
)) {
2083 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2084 if (ctl_buf
== NULL
)
2089 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2090 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2091 * checking falls down on this.
2093 if (copy_from_user(ctl_buf
,
2094 (void __user __force
*)msg_sys
->msg_control
,
2097 msg_sys
->msg_control
= ctl_buf
;
2099 msg_sys
->msg_flags
= flags
;
2101 if (sock
->file
->f_flags
& O_NONBLOCK
)
2102 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2104 * If this is sendmmsg() and current destination address is same as
2105 * previously succeeded address, omit asking LSM's decision.
2106 * used_address->name_len is initialized to UINT_MAX so that the first
2107 * destination address never matches.
2109 if (used_address
&& msg_sys
->msg_name
&&
2110 used_address
->name_len
== msg_sys
->msg_namelen
&&
2111 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2112 used_address
->name_len
)) {
2113 err
= sock_sendmsg_nosec(sock
, msg_sys
, total_len
);
2116 err
= sock_sendmsg(sock
, msg_sys
, total_len
);
2118 * If this is sendmmsg() and sending to current destination address was
2119 * successful, remember it.
2121 if (used_address
&& err
>= 0) {
2122 used_address
->name_len
= msg_sys
->msg_namelen
;
2123 if (msg_sys
->msg_name
)
2124 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2125 used_address
->name_len
);
2130 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2132 if (iov
!= iovstack
)
2138 * BSD sendmsg interface
2141 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2143 int fput_needed
, err
;
2144 struct msghdr msg_sys
;
2145 struct socket
*sock
;
2147 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2151 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
);
2153 fput_light(sock
->file
, fput_needed
);
2158 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2160 if (flags
& MSG_CMSG_COMPAT
)
2162 return __sys_sendmsg(fd
, msg
, flags
);
2166 * Linux sendmmsg interface
2169 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2172 int fput_needed
, err
, datagrams
;
2173 struct socket
*sock
;
2174 struct mmsghdr __user
*entry
;
2175 struct compat_mmsghdr __user
*compat_entry
;
2176 struct msghdr msg_sys
;
2177 struct used_address used_address
;
2179 if (vlen
> UIO_MAXIOV
)
2184 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2188 used_address
.name_len
= UINT_MAX
;
2190 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2193 while (datagrams
< vlen
) {
2194 if (MSG_CMSG_COMPAT
& flags
) {
2195 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2196 &msg_sys
, flags
, &used_address
);
2199 err
= __put_user(err
, &compat_entry
->msg_len
);
2202 err
= ___sys_sendmsg(sock
,
2203 (struct user_msghdr __user
*)entry
,
2204 &msg_sys
, flags
, &used_address
);
2207 err
= put_user(err
, &entry
->msg_len
);
2216 fput_light(sock
->file
, fput_needed
);
2218 /* We only return an error if no datagrams were able to be sent */
2225 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2226 unsigned int, vlen
, unsigned int, flags
)
2228 if (flags
& MSG_CMSG_COMPAT
)
2230 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2233 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2234 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2236 struct compat_msghdr __user
*msg_compat
=
2237 (struct compat_msghdr __user
*)msg
;
2238 struct iovec iovstack
[UIO_FASTIOV
];
2239 struct iovec
*iov
= iovstack
;
2240 unsigned long cmsg_ptr
;
2244 /* kernel mode address */
2245 struct sockaddr_storage addr
;
2247 /* user mode address pointers */
2248 struct sockaddr __user
*uaddr
;
2249 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2251 msg_sys
->msg_name
= &addr
;
2253 if (MSG_CMSG_COMPAT
& flags
)
2254 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2256 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2261 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2262 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2264 /* We assume all kernel code knows the size of sockaddr_storage */
2265 msg_sys
->msg_namelen
= 0;
2267 if (sock
->file
->f_flags
& O_NONBLOCK
)
2268 flags
|= MSG_DONTWAIT
;
2269 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2275 if (uaddr
!= NULL
) {
2276 err
= move_addr_to_user(&addr
,
2277 msg_sys
->msg_namelen
, uaddr
,
2282 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2286 if (MSG_CMSG_COMPAT
& flags
)
2287 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2288 &msg_compat
->msg_controllen
);
2290 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2291 &msg
->msg_controllen
);
2297 if (iov
!= iovstack
)
2303 * BSD recvmsg interface
2306 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2308 int fput_needed
, err
;
2309 struct msghdr msg_sys
;
2310 struct socket
*sock
;
2312 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2316 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2318 fput_light(sock
->file
, fput_needed
);
2323 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2324 unsigned int, flags
)
2326 if (flags
& MSG_CMSG_COMPAT
)
2328 return __sys_recvmsg(fd
, msg
, flags
);
2332 * Linux recvmmsg interface
2335 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2336 unsigned int flags
, struct timespec
*timeout
)
2338 int fput_needed
, err
, datagrams
;
2339 struct socket
*sock
;
2340 struct mmsghdr __user
*entry
;
2341 struct compat_mmsghdr __user
*compat_entry
;
2342 struct msghdr msg_sys
;
2343 struct timespec end_time
;
2346 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2352 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2356 err
= sock_error(sock
->sk
);
2361 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2363 while (datagrams
< vlen
) {
2365 * No need to ask LSM for more than the first datagram.
2367 if (MSG_CMSG_COMPAT
& flags
) {
2368 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2369 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2373 err
= __put_user(err
, &compat_entry
->msg_len
);
2376 err
= ___sys_recvmsg(sock
,
2377 (struct user_msghdr __user
*)entry
,
2378 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2382 err
= put_user(err
, &entry
->msg_len
);
2390 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2391 if (flags
& MSG_WAITFORONE
)
2392 flags
|= MSG_DONTWAIT
;
2395 ktime_get_ts(timeout
);
2396 *timeout
= timespec_sub(end_time
, *timeout
);
2397 if (timeout
->tv_sec
< 0) {
2398 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2402 /* Timeout, return less than vlen datagrams */
2403 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2407 /* Out of band data, return right away */
2408 if (msg_sys
.msg_flags
& MSG_OOB
)
2413 fput_light(sock
->file
, fput_needed
);
2418 if (datagrams
!= 0) {
2420 * We may return less entries than requested (vlen) if the
2421 * sock is non block and there aren't enough datagrams...
2423 if (err
!= -EAGAIN
) {
2425 * ... or if recvmsg returns an error after we
2426 * received some datagrams, where we record the
2427 * error to return on the next call or if the
2428 * app asks about it using getsockopt(SO_ERROR).
2430 sock
->sk
->sk_err
= -err
;
2439 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2440 unsigned int, vlen
, unsigned int, flags
,
2441 struct timespec __user
*, timeout
)
2444 struct timespec timeout_sys
;
2446 if (flags
& MSG_CMSG_COMPAT
)
2450 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2452 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2455 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2457 if (datagrams
> 0 &&
2458 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2459 datagrams
= -EFAULT
;
2464 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2465 /* Argument list sizes for sys_socketcall */
2466 #define AL(x) ((x) * sizeof(unsigned long))
2467 static const unsigned char nargs
[21] = {
2468 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2469 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2470 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2477 * System call vectors.
2479 * Argument checking cleaned up. Saved 20% in size.
2480 * This function doesn't need to set the kernel lock because
2481 * it is set by the callees.
2484 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2486 unsigned long a
[AUDITSC_ARGS
];
2487 unsigned long a0
, a1
;
2491 if (call
< 1 || call
> SYS_SENDMMSG
)
2495 if (len
> sizeof(a
))
2498 /* copy_from_user should be SMP safe. */
2499 if (copy_from_user(a
, args
, len
))
2502 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2511 err
= sys_socket(a0
, a1
, a
[2]);
2514 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2517 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2520 err
= sys_listen(a0
, a1
);
2523 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2524 (int __user
*)a
[2], 0);
2526 case SYS_GETSOCKNAME
:
2528 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2529 (int __user
*)a
[2]);
2531 case SYS_GETPEERNAME
:
2533 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2534 (int __user
*)a
[2]);
2536 case SYS_SOCKETPAIR
:
2537 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2540 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2543 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2544 (struct sockaddr __user
*)a
[4], a
[5]);
2547 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2550 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2551 (struct sockaddr __user
*)a
[4],
2552 (int __user
*)a
[5]);
2555 err
= sys_shutdown(a0
, a1
);
2557 case SYS_SETSOCKOPT
:
2558 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2560 case SYS_GETSOCKOPT
:
2562 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2563 (int __user
*)a
[4]);
2566 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2569 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2572 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2575 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2576 (struct timespec __user
*)a
[4]);
2579 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2580 (int __user
*)a
[2], a
[3]);
2589 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2592 * sock_register - add a socket protocol handler
2593 * @ops: description of protocol
2595 * This function is called by a protocol handler that wants to
2596 * advertise its address family, and have it linked into the
2597 * socket interface. The value ops->family corresponds to the
2598 * socket system call protocol family.
2600 int sock_register(const struct net_proto_family
*ops
)
2604 if (ops
->family
>= NPROTO
) {
2605 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2609 spin_lock(&net_family_lock
);
2610 if (rcu_dereference_protected(net_families
[ops
->family
],
2611 lockdep_is_held(&net_family_lock
)))
2614 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2617 spin_unlock(&net_family_lock
);
2619 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2622 EXPORT_SYMBOL(sock_register
);
2625 * sock_unregister - remove a protocol handler
2626 * @family: protocol family to remove
2628 * This function is called by a protocol handler that wants to
2629 * remove its address family, and have it unlinked from the
2630 * new socket creation.
2632 * If protocol handler is a module, then it can use module reference
2633 * counts to protect against new references. If protocol handler is not
2634 * a module then it needs to provide its own protection in
2635 * the ops->create routine.
2637 void sock_unregister(int family
)
2639 BUG_ON(family
< 0 || family
>= NPROTO
);
2641 spin_lock(&net_family_lock
);
2642 RCU_INIT_POINTER(net_families
[family
], NULL
);
2643 spin_unlock(&net_family_lock
);
2647 pr_info("NET: Unregistered protocol family %d\n", family
);
2649 EXPORT_SYMBOL(sock_unregister
);
2651 static int __init
sock_init(void)
2655 * Initialize the network sysctl infrastructure.
2657 err
= net_sysctl_init();
2662 * Initialize skbuff SLAB cache
2667 * Initialize the protocols module.
2672 err
= register_filesystem(&sock_fs_type
);
2675 sock_mnt
= kern_mount(&sock_fs_type
);
2676 if (IS_ERR(sock_mnt
)) {
2677 err
= PTR_ERR(sock_mnt
);
2681 /* The real protocol initialization is performed in later initcalls.
2684 #ifdef CONFIG_NETFILTER
2685 err
= netfilter_init();
2690 ptp_classifier_init();
2696 unregister_filesystem(&sock_fs_type
);
2701 core_initcall(sock_init
); /* early initcall */
2703 #ifdef CONFIG_PROC_FS
2704 void socket_seq_show(struct seq_file
*seq
)
2709 for_each_possible_cpu(cpu
)
2710 counter
+= per_cpu(sockets_in_use
, cpu
);
2712 /* It can be negative, by the way. 8) */
2716 seq_printf(seq
, "sockets: used %d\n", counter
);
2718 #endif /* CONFIG_PROC_FS */
2720 #ifdef CONFIG_COMPAT
2721 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2722 unsigned int cmd
, void __user
*up
)
2724 mm_segment_t old_fs
= get_fs();
2729 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2732 err
= compat_put_timeval(&ktv
, up
);
2737 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2738 unsigned int cmd
, void __user
*up
)
2740 mm_segment_t old_fs
= get_fs();
2741 struct timespec kts
;
2745 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2748 err
= compat_put_timespec(&kts
, up
);
2753 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2755 struct ifreq __user
*uifr
;
2758 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2759 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2762 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2766 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2772 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2774 struct compat_ifconf ifc32
;
2776 struct ifconf __user
*uifc
;
2777 struct compat_ifreq __user
*ifr32
;
2778 struct ifreq __user
*ifr
;
2782 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2785 memset(&ifc
, 0, sizeof(ifc
));
2786 if (ifc32
.ifcbuf
== 0) {
2790 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2792 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2793 sizeof(struct ifreq
);
2794 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2796 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2797 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2798 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2799 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2805 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2808 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2812 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2816 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2818 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2819 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2820 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2826 if (ifc32
.ifcbuf
== 0) {
2827 /* Translate from 64-bit structure multiple to
2831 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2836 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2842 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2844 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2845 bool convert_in
= false, convert_out
= false;
2846 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2847 struct ethtool_rxnfc __user
*rxnfc
;
2848 struct ifreq __user
*ifr
;
2849 u32 rule_cnt
= 0, actual_rule_cnt
;
2854 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2857 compat_rxnfc
= compat_ptr(data
);
2859 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2862 /* Most ethtool structures are defined without padding.
2863 * Unfortunately struct ethtool_rxnfc is an exception.
2868 case ETHTOOL_GRXCLSRLALL
:
2869 /* Buffer size is variable */
2870 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2872 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2874 buf_size
+= rule_cnt
* sizeof(u32
);
2876 case ETHTOOL_GRXRINGS
:
2877 case ETHTOOL_GRXCLSRLCNT
:
2878 case ETHTOOL_GRXCLSRULE
:
2879 case ETHTOOL_SRXCLSRLINS
:
2882 case ETHTOOL_SRXCLSRLDEL
:
2883 buf_size
+= sizeof(struct ethtool_rxnfc
);
2888 ifr
= compat_alloc_user_space(buf_size
);
2889 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2891 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2894 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2895 &ifr
->ifr_ifru
.ifru_data
))
2899 /* We expect there to be holes between fs.m_ext and
2900 * fs.ring_cookie and at the end of fs, but nowhere else.
2902 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2903 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2904 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2905 sizeof(rxnfc
->fs
.m_ext
));
2907 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2908 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2909 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2910 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2912 if (copy_in_user(rxnfc
, compat_rxnfc
,
2913 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2914 (void __user
*)rxnfc
) ||
2915 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2916 &compat_rxnfc
->fs
.ring_cookie
,
2917 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2918 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2919 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2920 sizeof(rxnfc
->rule_cnt
)))
2924 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2929 if (copy_in_user(compat_rxnfc
, rxnfc
,
2930 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2931 (const void __user
*)rxnfc
) ||
2932 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2933 &rxnfc
->fs
.ring_cookie
,
2934 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2935 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2936 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2937 sizeof(rxnfc
->rule_cnt
)))
2940 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2941 /* As an optimisation, we only copy the actual
2942 * number of rules that the underlying
2943 * function returned. Since Mallory might
2944 * change the rule count in user memory, we
2945 * check that it is less than the rule count
2946 * originally given (as the user buffer size),
2947 * which has been range-checked.
2949 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2951 if (actual_rule_cnt
< rule_cnt
)
2952 rule_cnt
= actual_rule_cnt
;
2953 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2954 &rxnfc
->rule_locs
[0],
2955 rule_cnt
* sizeof(u32
)))
2963 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2966 compat_uptr_t uptr32
;
2967 struct ifreq __user
*uifr
;
2969 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2970 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2973 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2976 uptr
= compat_ptr(uptr32
);
2978 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2981 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2984 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2985 struct compat_ifreq __user
*ifr32
)
2988 mm_segment_t old_fs
;
2992 case SIOCBONDENSLAVE
:
2993 case SIOCBONDRELEASE
:
2994 case SIOCBONDSETHWADDR
:
2995 case SIOCBONDCHANGEACTIVE
:
2996 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
3001 err
= dev_ioctl(net
, cmd
,
3002 (struct ifreq __user __force
*) &kifr
);
3007 return -ENOIOCTLCMD
;
3011 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3012 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3013 struct compat_ifreq __user
*u_ifreq32
)
3015 struct ifreq __user
*u_ifreq64
;
3016 char tmp_buf
[IFNAMSIZ
];
3017 void __user
*data64
;
3020 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
3023 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
3025 data64
= compat_ptr(data32
);
3027 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
3029 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
3032 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
3035 return dev_ioctl(net
, cmd
, u_ifreq64
);
3038 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
3039 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
3041 struct ifreq __user
*uifr
;
3044 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3045 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
3048 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3059 case SIOCGIFBRDADDR
:
3060 case SIOCGIFDSTADDR
:
3061 case SIOCGIFNETMASK
:
3066 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3074 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3075 struct compat_ifreq __user
*uifr32
)
3078 struct compat_ifmap __user
*uifmap32
;
3079 mm_segment_t old_fs
;
3082 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3083 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3084 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3085 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3086 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3087 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3088 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3089 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3095 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3098 if (cmd
== SIOCGIFMAP
&& !err
) {
3099 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3100 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3101 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3102 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3103 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3104 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3105 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3114 struct sockaddr rt_dst
; /* target address */
3115 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3116 struct sockaddr rt_genmask
; /* target network mask (IP) */
3117 unsigned short rt_flags
;
3120 unsigned char rt_tos
;
3121 unsigned char rt_class
;
3123 short rt_metric
; /* +1 for binary compatibility! */
3124 /* char * */ u32 rt_dev
; /* forcing the device at add */
3125 u32 rt_mtu
; /* per route MTU/Window */
3126 u32 rt_window
; /* Window clamping */
3127 unsigned short rt_irtt
; /* Initial RTT */
3130 struct in6_rtmsg32
{
3131 struct in6_addr rtmsg_dst
;
3132 struct in6_addr rtmsg_src
;
3133 struct in6_addr rtmsg_gateway
;
3143 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3144 unsigned int cmd
, void __user
*argp
)
3148 struct in6_rtmsg r6
;
3152 mm_segment_t old_fs
= get_fs();
3154 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3155 struct in6_rtmsg32 __user
*ur6
= argp
;
3156 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3157 3 * sizeof(struct in6_addr
));
3158 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3159 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3160 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3161 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3162 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3163 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3164 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3168 struct rtentry32 __user
*ur4
= argp
;
3169 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3170 3 * sizeof(struct sockaddr
));
3171 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3172 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3173 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3174 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3175 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3176 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3178 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3179 r4
.rt_dev
= (char __user __force
*)devname
;
3193 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3200 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3201 * for some operations; this forces use of the newer bridge-utils that
3202 * use compatible ioctls
3204 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3208 if (get_user(tmp
, argp
))
3210 if (tmp
== BRCTL_GET_VERSION
)
3211 return BRCTL_VERSION
+ 1;
3215 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3216 unsigned int cmd
, unsigned long arg
)
3218 void __user
*argp
= compat_ptr(arg
);
3219 struct sock
*sk
= sock
->sk
;
3220 struct net
*net
= sock_net(sk
);
3222 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3223 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3228 return old_bridge_ioctl(argp
);
3230 return dev_ifname32(net
, argp
);
3232 return dev_ifconf(net
, argp
);
3234 return ethtool_ioctl(net
, argp
);
3236 return compat_siocwandev(net
, argp
);
3239 return compat_sioc_ifmap(net
, cmd
, argp
);
3240 case SIOCBONDENSLAVE
:
3241 case SIOCBONDRELEASE
:
3242 case SIOCBONDSETHWADDR
:
3243 case SIOCBONDCHANGEACTIVE
:
3244 return bond_ioctl(net
, cmd
, argp
);
3247 return routing_ioctl(net
, sock
, cmd
, argp
);
3249 return do_siocgstamp(net
, sock
, cmd
, argp
);
3251 return do_siocgstampns(net
, sock
, cmd
, argp
);
3252 case SIOCBONDSLAVEINFOQUERY
:
3253 case SIOCBONDINFOQUERY
:
3256 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3268 return sock_ioctl(file
, cmd
, arg
);
3285 case SIOCSIFHWBROADCAST
:
3287 case SIOCGIFBRDADDR
:
3288 case SIOCSIFBRDADDR
:
3289 case SIOCGIFDSTADDR
:
3290 case SIOCSIFDSTADDR
:
3291 case SIOCGIFNETMASK
:
3292 case SIOCSIFNETMASK
:
3303 return dev_ifsioc(net
, sock
, cmd
, argp
);
3309 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3312 return -ENOIOCTLCMD
;
3315 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3318 struct socket
*sock
= file
->private_data
;
3319 int ret
= -ENOIOCTLCMD
;
3326 if (sock
->ops
->compat_ioctl
)
3327 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3329 if (ret
== -ENOIOCTLCMD
&&
3330 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3331 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3333 if (ret
== -ENOIOCTLCMD
)
3334 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3340 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3342 return sock
->ops
->bind(sock
, addr
, addrlen
);
3344 EXPORT_SYMBOL(kernel_bind
);
3346 int kernel_listen(struct socket
*sock
, int backlog
)
3348 return sock
->ops
->listen(sock
, backlog
);
3350 EXPORT_SYMBOL(kernel_listen
);
3352 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3354 struct sock
*sk
= sock
->sk
;
3357 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3362 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3364 sock_release(*newsock
);
3369 (*newsock
)->ops
= sock
->ops
;
3370 __module_get((*newsock
)->ops
->owner
);
3375 EXPORT_SYMBOL(kernel_accept
);
3377 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3380 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3382 EXPORT_SYMBOL(kernel_connect
);
3384 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3387 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3389 EXPORT_SYMBOL(kernel_getsockname
);
3391 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3394 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3396 EXPORT_SYMBOL(kernel_getpeername
);
3398 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3399 char *optval
, int *optlen
)
3401 mm_segment_t oldfs
= get_fs();
3402 char __user
*uoptval
;
3403 int __user
*uoptlen
;
3406 uoptval
= (char __user __force
*) optval
;
3407 uoptlen
= (int __user __force
*) optlen
;
3410 if (level
== SOL_SOCKET
)
3411 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3413 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3418 EXPORT_SYMBOL(kernel_getsockopt
);
3420 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3421 char *optval
, unsigned int optlen
)
3423 mm_segment_t oldfs
= get_fs();
3424 char __user
*uoptval
;
3427 uoptval
= (char __user __force
*) optval
;
3430 if (level
== SOL_SOCKET
)
3431 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3433 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3438 EXPORT_SYMBOL(kernel_setsockopt
);
3440 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3441 size_t size
, int flags
)
3443 if (sock
->ops
->sendpage
)
3444 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3446 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3448 EXPORT_SYMBOL(kernel_sendpage
);
3450 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3452 mm_segment_t oldfs
= get_fs();
3456 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3461 EXPORT_SYMBOL(kernel_sock_ioctl
);
3463 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3465 return sock
->ops
->shutdown(sock
, how
);
3467 EXPORT_SYMBOL(kernel_sock_shutdown
);