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 ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
);
117 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
);
118 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
120 static int sock_close(struct inode
*inode
, struct file
*file
);
121 static unsigned int sock_poll(struct file
*file
,
122 struct poll_table_struct
*wait
);
123 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
125 static long compat_sock_ioctl(struct file
*file
,
126 unsigned int cmd
, unsigned long arg
);
128 static int sock_fasync(int fd
, struct file
*filp
, int on
);
129 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
130 int offset
, size_t size
, loff_t
*ppos
, int more
);
131 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
132 struct pipe_inode_info
*pipe
, size_t len
,
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
140 static const struct file_operations socket_file_ops
= {
141 .owner
= THIS_MODULE
,
143 .read
= new_sync_read
,
144 .write
= new_sync_write
,
145 .read_iter
= sock_read_iter
,
146 .write_iter
= sock_write_iter
,
148 .unlocked_ioctl
= sock_ioctl
,
150 .compat_ioctl
= compat_sock_ioctl
,
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
));
376 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
378 if (unlikely(IS_ERR(file
))) {
379 /* drop dentry, keep inode */
380 ihold(path
.dentry
->d_inode
);
386 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
387 file
->private_data
= sock
;
390 EXPORT_SYMBOL(sock_alloc_file
);
392 static int sock_map_fd(struct socket
*sock
, int flags
)
394 struct file
*newfile
;
395 int fd
= get_unused_fd_flags(flags
);
396 if (unlikely(fd
< 0))
399 newfile
= sock_alloc_file(sock
, flags
, NULL
);
400 if (likely(!IS_ERR(newfile
))) {
401 fd_install(fd
, newfile
);
406 return PTR_ERR(newfile
);
409 struct socket
*sock_from_file(struct file
*file
, int *err
)
411 if (file
->f_op
== &socket_file_ops
)
412 return file
->private_data
; /* set in sock_map_fd */
417 EXPORT_SYMBOL(sock_from_file
);
420 * sockfd_lookup - Go from a file number to its socket slot
422 * @err: pointer to an error code return
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
429 * On a success the socket object pointer is returned.
432 struct socket
*sockfd_lookup(int fd
, int *err
)
443 sock
= sock_from_file(file
, err
);
448 EXPORT_SYMBOL(sockfd_lookup
);
450 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
452 struct fd f
= fdget(fd
);
457 sock
= sock_from_file(f
.file
, err
);
459 *fput_needed
= f
.flags
;
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 * sock_release - close a socket
561 * @sock: socket to close
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
568 void sock_release(struct socket
*sock
)
571 struct module
*owner
= sock
->ops
->owner
;
573 sock
->ops
->release(sock
);
578 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
579 pr_err("%s: fasync list not empty!\n", __func__
);
581 if (test_bit(SOCK_EXTERNALLY_ALLOCATED
, &sock
->flags
))
584 this_cpu_sub(sockets_in_use
, 1);
586 iput(SOCK_INODE(sock
));
591 EXPORT_SYMBOL(sock_release
);
593 void __sock_tx_timestamp(const struct sock
*sk
, __u8
*tx_flags
)
595 u8 flags
= *tx_flags
;
597 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
598 flags
|= SKBTX_HW_TSTAMP
;
600 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
601 flags
|= SKBTX_SW_TSTAMP
;
603 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
604 flags
|= SKBTX_SCHED_TSTAMP
;
606 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_TX_ACK
)
607 flags
|= SKBTX_ACK_TSTAMP
;
611 EXPORT_SYMBOL(__sock_tx_timestamp
);
613 static inline int __sock_sendmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
614 struct msghdr
*msg
, size_t size
)
616 return sock
->ops
->sendmsg(iocb
, sock
, msg
, size
);
619 static inline int __sock_sendmsg(struct kiocb
*iocb
, struct socket
*sock
,
620 struct msghdr
*msg
, size_t size
)
622 int err
= security_socket_sendmsg(sock
, msg
, size
);
624 return err
?: __sock_sendmsg_nosec(iocb
, sock
, msg
, size
);
627 static int do_sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
628 size_t size
, bool nosec
)
633 init_sync_kiocb(&iocb
, NULL
);
634 ret
= nosec
? __sock_sendmsg_nosec(&iocb
, sock
, msg
, size
) :
635 __sock_sendmsg(&iocb
, sock
, msg
, size
);
636 if (-EIOCBQUEUED
== ret
)
637 ret
= wait_on_sync_kiocb(&iocb
);
641 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
643 return do_sock_sendmsg(sock
, msg
, size
, false);
645 EXPORT_SYMBOL(sock_sendmsg
);
647 static int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
, size_t size
)
649 return do_sock_sendmsg(sock
, msg
, size
, true);
652 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
653 struct kvec
*vec
, size_t num
, size_t size
)
655 mm_segment_t oldfs
= get_fs();
660 * the following is safe, since for compiler definitions of kvec and
661 * iovec are identical, yielding the same in-core layout and alignment
663 iov_iter_init(&msg
->msg_iter
, WRITE
, (struct iovec
*)vec
, num
, size
);
664 result
= sock_sendmsg(sock
, msg
, size
);
668 EXPORT_SYMBOL(kernel_sendmsg
);
671 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
673 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
676 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
677 struct scm_timestamping tss
;
679 struct skb_shared_hwtstamps
*shhwtstamps
=
682 /* Race occurred between timestamp enabling and packet
683 receiving. Fill in the current time for now. */
684 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
685 __net_timestamp(skb
);
687 if (need_software_tstamp
) {
688 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
690 skb_get_timestamp(skb
, &tv
);
691 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
695 skb_get_timestampns(skb
, &ts
);
696 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
701 memset(&tss
, 0, sizeof(tss
));
702 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
703 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
706 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
707 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2))
710 put_cmsg(msg
, SOL_SOCKET
,
711 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
713 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
715 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
720 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
722 if (!skb
->wifi_acked_valid
)
725 ack
= skb
->wifi_acked
;
727 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
729 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
731 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
734 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& skb
->dropcount
)
735 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
736 sizeof(__u32
), &skb
->dropcount
);
739 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
742 sock_recv_timestamp(msg
, sk
, skb
);
743 sock_recv_drops(msg
, sk
, skb
);
745 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
747 static inline int __sock_recvmsg_nosec(struct kiocb
*iocb
, struct socket
*sock
,
748 struct msghdr
*msg
, size_t size
, int flags
)
750 return sock
->ops
->recvmsg(iocb
, sock
, msg
, size
, flags
);
753 static inline int __sock_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
754 struct msghdr
*msg
, size_t size
, int flags
)
756 int err
= security_socket_recvmsg(sock
, msg
, size
, flags
);
758 return err
?: __sock_recvmsg_nosec(iocb
, sock
, msg
, size
, flags
);
761 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
762 size_t size
, int flags
)
767 init_sync_kiocb(&iocb
, NULL
);
768 ret
= __sock_recvmsg(&iocb
, sock
, msg
, size
, flags
);
769 if (-EIOCBQUEUED
== ret
)
770 ret
= wait_on_sync_kiocb(&iocb
);
773 EXPORT_SYMBOL(sock_recvmsg
);
775 static int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
776 size_t size
, int flags
)
781 init_sync_kiocb(&iocb
, NULL
);
782 ret
= __sock_recvmsg_nosec(&iocb
, sock
, msg
, size
, flags
);
783 if (-EIOCBQUEUED
== ret
)
784 ret
= wait_on_sync_kiocb(&iocb
);
789 * kernel_recvmsg - Receive a message from a socket (kernel space)
790 * @sock: The socket to receive the message from
791 * @msg: Received message
792 * @vec: Input s/g array for message data
793 * @num: Size of input s/g array
794 * @size: Number of bytes to read
795 * @flags: Message flags (MSG_DONTWAIT, etc...)
797 * On return the msg structure contains the scatter/gather array passed in the
798 * vec argument. The array is modified so that it consists of the unfilled
799 * portion of the original array.
801 * The returned value is the total number of bytes received, or an error.
803 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
804 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
806 mm_segment_t oldfs
= get_fs();
811 * the following is safe, since for compiler definitions of kvec and
812 * iovec are identical, yielding the same in-core layout and alignment
814 iov_iter_init(&msg
->msg_iter
, READ
, (struct iovec
*)vec
, num
, size
);
815 result
= sock_recvmsg(sock
, msg
, size
, flags
);
819 EXPORT_SYMBOL(kernel_recvmsg
);
821 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
822 int offset
, size_t size
, loff_t
*ppos
, int more
)
827 sock
= file
->private_data
;
829 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
830 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
833 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
836 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
837 struct pipe_inode_info
*pipe
, size_t len
,
840 struct socket
*sock
= file
->private_data
;
842 if (unlikely(!sock
->ops
->splice_read
))
845 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
848 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
850 struct file
*file
= iocb
->ki_filp
;
851 struct socket
*sock
= file
->private_data
;
852 struct msghdr msg
= {.msg_iter
= *to
};
855 if (file
->f_flags
& O_NONBLOCK
)
856 msg
.msg_flags
= MSG_DONTWAIT
;
858 if (iocb
->ki_pos
!= 0)
861 if (iocb
->ki_nbytes
== 0) /* Match SYS5 behaviour */
864 res
= __sock_recvmsg(iocb
, sock
, &msg
,
865 iocb
->ki_nbytes
, msg
.msg_flags
);
870 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
872 struct file
*file
= iocb
->ki_filp
;
873 struct socket
*sock
= file
->private_data
;
874 struct msghdr msg
= {.msg_iter
= *from
};
877 if (iocb
->ki_pos
!= 0)
880 if (file
->f_flags
& O_NONBLOCK
)
881 msg
.msg_flags
= MSG_DONTWAIT
;
883 if (sock
->type
== SOCK_SEQPACKET
)
884 msg
.msg_flags
|= MSG_EOR
;
886 res
= __sock_sendmsg(iocb
, sock
, &msg
, iocb
->ki_nbytes
);
887 *from
= msg
.msg_iter
;
892 * Atomic setting of ioctl hooks to avoid race
893 * with module unload.
896 static DEFINE_MUTEX(br_ioctl_mutex
);
897 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
899 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
901 mutex_lock(&br_ioctl_mutex
);
902 br_ioctl_hook
= hook
;
903 mutex_unlock(&br_ioctl_mutex
);
905 EXPORT_SYMBOL(brioctl_set
);
907 static DEFINE_MUTEX(vlan_ioctl_mutex
);
908 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
910 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
912 mutex_lock(&vlan_ioctl_mutex
);
913 vlan_ioctl_hook
= hook
;
914 mutex_unlock(&vlan_ioctl_mutex
);
916 EXPORT_SYMBOL(vlan_ioctl_set
);
918 static DEFINE_MUTEX(dlci_ioctl_mutex
);
919 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
921 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
923 mutex_lock(&dlci_ioctl_mutex
);
924 dlci_ioctl_hook
= hook
;
925 mutex_unlock(&dlci_ioctl_mutex
);
927 EXPORT_SYMBOL(dlci_ioctl_set
);
929 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
930 unsigned int cmd
, unsigned long arg
)
933 void __user
*argp
= (void __user
*)arg
;
935 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
938 * If this ioctl is unknown try to hand it down
941 if (err
== -ENOIOCTLCMD
)
942 err
= dev_ioctl(net
, cmd
, argp
);
948 * With an ioctl, arg may well be a user mode pointer, but we don't know
949 * what to do with it - that's up to the protocol still.
952 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
956 void __user
*argp
= (void __user
*)arg
;
960 sock
= file
->private_data
;
963 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
964 err
= dev_ioctl(net
, cmd
, argp
);
966 #ifdef CONFIG_WEXT_CORE
967 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
968 err
= dev_ioctl(net
, cmd
, argp
);
975 if (get_user(pid
, (int __user
*)argp
))
977 f_setown(sock
->file
, pid
, 1);
982 err
= put_user(f_getown(sock
->file
),
991 request_module("bridge");
993 mutex_lock(&br_ioctl_mutex
);
995 err
= br_ioctl_hook(net
, cmd
, argp
);
996 mutex_unlock(&br_ioctl_mutex
);
1001 if (!vlan_ioctl_hook
)
1002 request_module("8021q");
1004 mutex_lock(&vlan_ioctl_mutex
);
1005 if (vlan_ioctl_hook
)
1006 err
= vlan_ioctl_hook(net
, argp
);
1007 mutex_unlock(&vlan_ioctl_mutex
);
1012 if (!dlci_ioctl_hook
)
1013 request_module("dlci");
1015 mutex_lock(&dlci_ioctl_mutex
);
1016 if (dlci_ioctl_hook
)
1017 err
= dlci_ioctl_hook(cmd
, argp
);
1018 mutex_unlock(&dlci_ioctl_mutex
);
1021 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1027 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1030 struct socket
*sock
= NULL
;
1032 err
= security_socket_create(family
, type
, protocol
, 1);
1036 sock
= sock_alloc();
1043 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1055 EXPORT_SYMBOL(sock_create_lite
);
1057 /* No kernel lock held - perfect */
1058 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1060 unsigned int busy_flag
= 0;
1061 struct socket
*sock
;
1064 * We can't return errors to poll, so it's either yes or no.
1066 sock
= file
->private_data
;
1068 if (sk_can_busy_loop(sock
->sk
)) {
1069 /* this socket can poll_ll so tell the system call */
1070 busy_flag
= POLL_BUSY_LOOP
;
1072 /* once, only if requested by syscall */
1073 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1074 sk_busy_loop(sock
->sk
, 1);
1077 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1080 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1082 struct socket
*sock
= file
->private_data
;
1084 return sock
->ops
->mmap(file
, sock
, vma
);
1087 static int sock_close(struct inode
*inode
, struct file
*filp
)
1089 sock_release(SOCKET_I(inode
));
1094 * Update the socket async list
1096 * Fasync_list locking strategy.
1098 * 1. fasync_list is modified only under process context socket lock
1099 * i.e. under semaphore.
1100 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1101 * or under socket lock
1104 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1106 struct socket
*sock
= filp
->private_data
;
1107 struct sock
*sk
= sock
->sk
;
1108 struct socket_wq
*wq
;
1114 wq
= rcu_dereference_protected(sock
->wq
, sock_owned_by_user(sk
));
1115 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1117 if (!wq
->fasync_list
)
1118 sock_reset_flag(sk
, SOCK_FASYNC
);
1120 sock_set_flag(sk
, SOCK_FASYNC
);
1126 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1128 int sock_wake_async(struct socket
*sock
, int how
, int band
)
1130 struct socket_wq
*wq
;
1135 wq
= rcu_dereference(sock
->wq
);
1136 if (!wq
|| !wq
->fasync_list
) {
1141 case SOCK_WAKE_WAITD
:
1142 if (test_bit(SOCK_ASYNC_WAITDATA
, &sock
->flags
))
1145 case SOCK_WAKE_SPACE
:
1146 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE
, &sock
->flags
))
1151 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1154 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1159 EXPORT_SYMBOL(sock_wake_async
);
1161 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1162 struct socket
**res
, int kern
)
1165 struct socket
*sock
;
1166 const struct net_proto_family
*pf
;
1169 * Check protocol is in range
1171 if (family
< 0 || family
>= NPROTO
)
1172 return -EAFNOSUPPORT
;
1173 if (type
< 0 || type
>= SOCK_MAX
)
1178 This uglymoron is moved from INET layer to here to avoid
1179 deadlock in module load.
1181 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1185 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1191 err
= security_socket_create(family
, type
, protocol
, kern
);
1196 * Allocate the socket and allow the family to set things up. if
1197 * the protocol is 0, the family is instructed to select an appropriate
1200 sock
= sock_alloc();
1202 net_warn_ratelimited("socket: no more sockets\n");
1203 return -ENFILE
; /* Not exactly a match, but its the
1204 closest posix thing */
1209 #ifdef CONFIG_MODULES
1210 /* Attempt to load a protocol module if the find failed.
1212 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1213 * requested real, full-featured networking support upon configuration.
1214 * Otherwise module support will break!
1216 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1217 request_module("net-pf-%d", family
);
1221 pf
= rcu_dereference(net_families
[family
]);
1222 err
= -EAFNOSUPPORT
;
1227 * We will call the ->create function, that possibly is in a loadable
1228 * module, so we have to bump that loadable module refcnt first.
1230 if (!try_module_get(pf
->owner
))
1233 /* Now protected by module ref count */
1236 err
= pf
->create(net
, sock
, protocol
, kern
);
1238 goto out_module_put
;
1241 * Now to bump the refcnt of the [loadable] module that owns this
1242 * socket at sock_release time we decrement its refcnt.
1244 if (!try_module_get(sock
->ops
->owner
))
1245 goto out_module_busy
;
1248 * Now that we're done with the ->create function, the [loadable]
1249 * module can have its refcnt decremented
1251 module_put(pf
->owner
);
1252 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1254 goto out_sock_release
;
1260 err
= -EAFNOSUPPORT
;
1263 module_put(pf
->owner
);
1270 goto out_sock_release
;
1272 EXPORT_SYMBOL(__sock_create
);
1274 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1276 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1278 EXPORT_SYMBOL(sock_create
);
1280 int sock_create_kern(int family
, int type
, int protocol
, struct socket
**res
)
1282 return __sock_create(&init_net
, family
, type
, protocol
, res
, 1);
1284 EXPORT_SYMBOL(sock_create_kern
);
1286 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1289 struct socket
*sock
;
1292 /* Check the SOCK_* constants for consistency. */
1293 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1294 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1295 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1296 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1298 flags
= type
& ~SOCK_TYPE_MASK
;
1299 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1301 type
&= SOCK_TYPE_MASK
;
1303 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1304 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1306 retval
= sock_create(family
, type
, protocol
, &sock
);
1310 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1315 /* It may be already another descriptor 8) Not kernel problem. */
1324 * Create a pair of connected sockets.
1327 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1328 int __user
*, usockvec
)
1330 struct socket
*sock1
, *sock2
;
1332 struct file
*newfile1
, *newfile2
;
1335 flags
= type
& ~SOCK_TYPE_MASK
;
1336 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1338 type
&= SOCK_TYPE_MASK
;
1340 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1341 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1344 * Obtain the first socket and check if the underlying protocol
1345 * supports the socketpair call.
1348 err
= sock_create(family
, type
, protocol
, &sock1
);
1352 err
= sock_create(family
, type
, protocol
, &sock2
);
1356 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1358 goto out_release_both
;
1360 fd1
= get_unused_fd_flags(flags
);
1361 if (unlikely(fd1
< 0)) {
1363 goto out_release_both
;
1366 fd2
= get_unused_fd_flags(flags
);
1367 if (unlikely(fd2
< 0)) {
1369 goto out_put_unused_1
;
1372 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1373 if (unlikely(IS_ERR(newfile1
))) {
1374 err
= PTR_ERR(newfile1
);
1375 goto out_put_unused_both
;
1378 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1379 if (IS_ERR(newfile2
)) {
1380 err
= PTR_ERR(newfile2
);
1384 err
= put_user(fd1
, &usockvec
[0]);
1388 err
= put_user(fd2
, &usockvec
[1]);
1392 audit_fd_pair(fd1
, fd2
);
1394 fd_install(fd1
, newfile1
);
1395 fd_install(fd2
, newfile2
);
1396 /* fd1 and fd2 may be already another descriptors.
1397 * Not kernel problem.
1413 sock_release(sock2
);
1416 out_put_unused_both
:
1421 sock_release(sock2
);
1423 sock_release(sock1
);
1429 * Bind a name to a socket. Nothing much to do here since it's
1430 * the protocol's responsibility to handle the local address.
1432 * We move the socket address to kernel space before we call
1433 * the protocol layer (having also checked the address is ok).
1436 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1438 struct socket
*sock
;
1439 struct sockaddr_storage address
;
1440 int err
, fput_needed
;
1442 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1444 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1446 err
= security_socket_bind(sock
,
1447 (struct sockaddr
*)&address
,
1450 err
= sock
->ops
->bind(sock
,
1454 fput_light(sock
->file
, fput_needed
);
1460 * Perform a listen. Basically, we allow the protocol to do anything
1461 * necessary for a listen, and if that works, we mark the socket as
1462 * ready for listening.
1465 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1467 struct socket
*sock
;
1468 int err
, fput_needed
;
1471 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1473 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1474 if ((unsigned int)backlog
> somaxconn
)
1475 backlog
= somaxconn
;
1477 err
= security_socket_listen(sock
, backlog
);
1479 err
= sock
->ops
->listen(sock
, backlog
);
1481 fput_light(sock
->file
, fput_needed
);
1487 * For accept, we attempt to create a new socket, set up the link
1488 * with the client, wake up the client, then return the new
1489 * connected fd. We collect the address of the connector in kernel
1490 * space and move it to user at the very end. This is unclean because
1491 * we open the socket then return an error.
1493 * 1003.1g adds the ability to recvmsg() to query connection pending
1494 * status to recvmsg. We need to add that support in a way thats
1495 * clean when we restucture accept also.
1498 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1499 int __user
*, upeer_addrlen
, int, flags
)
1501 struct socket
*sock
, *newsock
;
1502 struct file
*newfile
;
1503 int err
, len
, newfd
, fput_needed
;
1504 struct sockaddr_storage address
;
1506 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1509 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1510 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1512 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1517 newsock
= sock_alloc();
1521 newsock
->type
= sock
->type
;
1522 newsock
->ops
= sock
->ops
;
1525 * We don't need try_module_get here, as the listening socket (sock)
1526 * has the protocol module (sock->ops->owner) held.
1528 __module_get(newsock
->ops
->owner
);
1530 newfd
= get_unused_fd_flags(flags
);
1531 if (unlikely(newfd
< 0)) {
1533 sock_release(newsock
);
1536 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1537 if (unlikely(IS_ERR(newfile
))) {
1538 err
= PTR_ERR(newfile
);
1539 put_unused_fd(newfd
);
1540 sock_release(newsock
);
1544 err
= security_socket_accept(sock
, newsock
);
1548 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1552 if (upeer_sockaddr
) {
1553 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1555 err
= -ECONNABORTED
;
1558 err
= move_addr_to_user(&address
,
1559 len
, upeer_sockaddr
, upeer_addrlen
);
1564 /* File flags are not inherited via accept() unlike another OSes. */
1566 fd_install(newfd
, newfile
);
1570 fput_light(sock
->file
, fput_needed
);
1575 put_unused_fd(newfd
);
1579 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1580 int __user
*, upeer_addrlen
)
1582 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1586 * Attempt to connect to a socket with the server address. The address
1587 * is in user space so we verify it is OK and move it to kernel space.
1589 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1592 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1593 * other SEQPACKET protocols that take time to connect() as it doesn't
1594 * include the -EINPROGRESS status for such sockets.
1597 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1600 struct socket
*sock
;
1601 struct sockaddr_storage address
;
1602 int err
, fput_needed
;
1604 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1607 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1612 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1616 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1617 sock
->file
->f_flags
);
1619 fput_light(sock
->file
, fput_needed
);
1625 * Get the local address ('name') of a socket object. Move the obtained
1626 * name to user space.
1629 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1630 int __user
*, usockaddr_len
)
1632 struct socket
*sock
;
1633 struct sockaddr_storage address
;
1634 int len
, err
, fput_needed
;
1636 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1640 err
= security_socket_getsockname(sock
);
1644 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1647 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1650 fput_light(sock
->file
, fput_needed
);
1656 * Get the remote address ('name') of a socket object. Move the obtained
1657 * name to user space.
1660 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1661 int __user
*, usockaddr_len
)
1663 struct socket
*sock
;
1664 struct sockaddr_storage address
;
1665 int len
, err
, fput_needed
;
1667 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1669 err
= security_socket_getpeername(sock
);
1671 fput_light(sock
->file
, fput_needed
);
1676 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1679 err
= move_addr_to_user(&address
, len
, usockaddr
,
1681 fput_light(sock
->file
, fput_needed
);
1687 * Send a datagram to a given address. We move the address into kernel
1688 * space and check the user space data area is readable before invoking
1692 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1693 unsigned int, flags
, struct sockaddr __user
*, addr
,
1696 struct socket
*sock
;
1697 struct sockaddr_storage address
;
1705 if (unlikely(!access_ok(VERIFY_READ
, buff
, len
)))
1707 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1711 iov
.iov_base
= buff
;
1713 msg
.msg_name
= NULL
;
1714 iov_iter_init(&msg
.msg_iter
, WRITE
, &iov
, 1, len
);
1715 msg
.msg_control
= NULL
;
1716 msg
.msg_controllen
= 0;
1717 msg
.msg_namelen
= 0;
1719 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1722 msg
.msg_name
= (struct sockaddr
*)&address
;
1723 msg
.msg_namelen
= addr_len
;
1725 if (sock
->file
->f_flags
& O_NONBLOCK
)
1726 flags
|= MSG_DONTWAIT
;
1727 msg
.msg_flags
= flags
;
1728 err
= sock_sendmsg(sock
, &msg
, len
);
1731 fput_light(sock
->file
, fput_needed
);
1737 * Send a datagram down a socket.
1740 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1741 unsigned int, flags
)
1743 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1747 * Receive a frame from the socket and optionally record the address of the
1748 * sender. We verify the buffers are writable and if needed move the
1749 * sender address from kernel to user space.
1752 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1753 unsigned int, flags
, struct sockaddr __user
*, addr
,
1754 int __user
*, addr_len
)
1756 struct socket
*sock
;
1759 struct sockaddr_storage address
;
1765 if (unlikely(!access_ok(VERIFY_WRITE
, ubuf
, size
)))
1767 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1771 msg
.msg_control
= NULL
;
1772 msg
.msg_controllen
= 0;
1774 iov
.iov_base
= ubuf
;
1775 iov_iter_init(&msg
.msg_iter
, READ
, &iov
, 1, size
);
1776 /* Save some cycles and don't copy the address if not needed */
1777 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1778 /* We assume all kernel code knows the size of sockaddr_storage */
1779 msg
.msg_namelen
= 0;
1780 if (sock
->file
->f_flags
& O_NONBLOCK
)
1781 flags
|= MSG_DONTWAIT
;
1782 err
= sock_recvmsg(sock
, &msg
, size
, flags
);
1784 if (err
>= 0 && addr
!= NULL
) {
1785 err2
= move_addr_to_user(&address
,
1786 msg
.msg_namelen
, addr
, addr_len
);
1791 fput_light(sock
->file
, fput_needed
);
1797 * Receive a datagram from a socket.
1800 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1801 unsigned int, flags
)
1803 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1807 * Set a socket option. Because we don't know the option lengths we have
1808 * to pass the user mode parameter for the protocols to sort out.
1811 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1812 char __user
*, optval
, int, optlen
)
1814 int err
, fput_needed
;
1815 struct socket
*sock
;
1820 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1822 err
= security_socket_setsockopt(sock
, level
, optname
);
1826 if (level
== SOL_SOCKET
)
1828 sock_setsockopt(sock
, level
, optname
, optval
,
1832 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1835 fput_light(sock
->file
, fput_needed
);
1841 * Get a socket option. Because we don't know the option lengths we have
1842 * to pass a user mode parameter for the protocols to sort out.
1845 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1846 char __user
*, optval
, int __user
*, optlen
)
1848 int err
, fput_needed
;
1849 struct socket
*sock
;
1851 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1853 err
= security_socket_getsockopt(sock
, level
, optname
);
1857 if (level
== SOL_SOCKET
)
1859 sock_getsockopt(sock
, level
, optname
, optval
,
1863 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1866 fput_light(sock
->file
, fput_needed
);
1872 * Shutdown a socket.
1875 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1877 int err
, fput_needed
;
1878 struct socket
*sock
;
1880 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1882 err
= security_socket_shutdown(sock
, how
);
1884 err
= sock
->ops
->shutdown(sock
, how
);
1885 fput_light(sock
->file
, fput_needed
);
1890 /* A couple of helpful macros for getting the address of the 32/64 bit
1891 * fields which are the same type (int / unsigned) on our platforms.
1893 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1894 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1895 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1897 struct used_address
{
1898 struct sockaddr_storage name
;
1899 unsigned int name_len
;
1902 static ssize_t
copy_msghdr_from_user(struct msghdr
*kmsg
,
1903 struct user_msghdr __user
*umsg
,
1904 struct sockaddr __user
**save_addr
,
1907 struct sockaddr __user
*uaddr
;
1908 struct iovec __user
*uiov
;
1912 if (!access_ok(VERIFY_READ
, umsg
, sizeof(*umsg
)) ||
1913 __get_user(uaddr
, &umsg
->msg_name
) ||
1914 __get_user(kmsg
->msg_namelen
, &umsg
->msg_namelen
) ||
1915 __get_user(uiov
, &umsg
->msg_iov
) ||
1916 __get_user(nr_segs
, &umsg
->msg_iovlen
) ||
1917 __get_user(kmsg
->msg_control
, &umsg
->msg_control
) ||
1918 __get_user(kmsg
->msg_controllen
, &umsg
->msg_controllen
) ||
1919 __get_user(kmsg
->msg_flags
, &umsg
->msg_flags
))
1923 kmsg
->msg_namelen
= 0;
1925 if (kmsg
->msg_namelen
< 0)
1928 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1929 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1934 if (uaddr
&& kmsg
->msg_namelen
) {
1936 err
= move_addr_to_kernel(uaddr
, kmsg
->msg_namelen
,
1942 kmsg
->msg_name
= NULL
;
1943 kmsg
->msg_namelen
= 0;
1946 if (nr_segs
> UIO_MAXIOV
)
1949 err
= rw_copy_check_uvector(save_addr
? READ
: WRITE
,
1951 UIO_FASTIOV
, *iov
, iov
);
1953 iov_iter_init(&kmsg
->msg_iter
, save_addr
? READ
: WRITE
,
1954 *iov
, nr_segs
, err
);
1958 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
1959 struct msghdr
*msg_sys
, unsigned int flags
,
1960 struct used_address
*used_address
)
1962 struct compat_msghdr __user
*msg_compat
=
1963 (struct compat_msghdr __user
*)msg
;
1964 struct sockaddr_storage address
;
1965 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1966 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1967 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1968 /* 20 is size of ipv6_pktinfo */
1969 unsigned char *ctl_buf
= ctl
;
1970 int ctl_len
, total_len
;
1973 msg_sys
->msg_name
= &address
;
1975 if (MSG_CMSG_COMPAT
& flags
)
1976 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
1978 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
1985 if (msg_sys
->msg_controllen
> INT_MAX
)
1987 ctl_len
= msg_sys
->msg_controllen
;
1988 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1990 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
1994 ctl_buf
= msg_sys
->msg_control
;
1995 ctl_len
= msg_sys
->msg_controllen
;
1996 } else if (ctl_len
) {
1997 if (ctl_len
> sizeof(ctl
)) {
1998 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1999 if (ctl_buf
== NULL
)
2004 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2005 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2006 * checking falls down on this.
2008 if (copy_from_user(ctl_buf
,
2009 (void __user __force
*)msg_sys
->msg_control
,
2012 msg_sys
->msg_control
= ctl_buf
;
2014 msg_sys
->msg_flags
= flags
;
2016 if (sock
->file
->f_flags
& O_NONBLOCK
)
2017 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2019 * If this is sendmmsg() and current destination address is same as
2020 * previously succeeded address, omit asking LSM's decision.
2021 * used_address->name_len is initialized to UINT_MAX so that the first
2022 * destination address never matches.
2024 if (used_address
&& msg_sys
->msg_name
&&
2025 used_address
->name_len
== msg_sys
->msg_namelen
&&
2026 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2027 used_address
->name_len
)) {
2028 err
= sock_sendmsg_nosec(sock
, msg_sys
, total_len
);
2031 err
= sock_sendmsg(sock
, msg_sys
, total_len
);
2033 * If this is sendmmsg() and sending to current destination address was
2034 * successful, remember it.
2036 if (used_address
&& err
>= 0) {
2037 used_address
->name_len
= msg_sys
->msg_namelen
;
2038 if (msg_sys
->msg_name
)
2039 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2040 used_address
->name_len
);
2045 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2047 if (iov
!= iovstack
)
2053 * BSD sendmsg interface
2056 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2058 int fput_needed
, err
;
2059 struct msghdr msg_sys
;
2060 struct socket
*sock
;
2062 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2066 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
);
2068 fput_light(sock
->file
, fput_needed
);
2073 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2075 if (flags
& MSG_CMSG_COMPAT
)
2077 return __sys_sendmsg(fd
, msg
, flags
);
2081 * Linux sendmmsg interface
2084 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2087 int fput_needed
, err
, datagrams
;
2088 struct socket
*sock
;
2089 struct mmsghdr __user
*entry
;
2090 struct compat_mmsghdr __user
*compat_entry
;
2091 struct msghdr msg_sys
;
2092 struct used_address used_address
;
2094 if (vlen
> UIO_MAXIOV
)
2099 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2103 used_address
.name_len
= UINT_MAX
;
2105 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2108 while (datagrams
< vlen
) {
2109 if (MSG_CMSG_COMPAT
& flags
) {
2110 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2111 &msg_sys
, flags
, &used_address
);
2114 err
= __put_user(err
, &compat_entry
->msg_len
);
2117 err
= ___sys_sendmsg(sock
,
2118 (struct user_msghdr __user
*)entry
,
2119 &msg_sys
, flags
, &used_address
);
2122 err
= put_user(err
, &entry
->msg_len
);
2131 fput_light(sock
->file
, fput_needed
);
2133 /* We only return an error if no datagrams were able to be sent */
2140 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2141 unsigned int, vlen
, unsigned int, flags
)
2143 if (flags
& MSG_CMSG_COMPAT
)
2145 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2148 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2149 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2151 struct compat_msghdr __user
*msg_compat
=
2152 (struct compat_msghdr __user
*)msg
;
2153 struct iovec iovstack
[UIO_FASTIOV
];
2154 struct iovec
*iov
= iovstack
;
2155 unsigned long cmsg_ptr
;
2159 /* kernel mode address */
2160 struct sockaddr_storage addr
;
2162 /* user mode address pointers */
2163 struct sockaddr __user
*uaddr
;
2164 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2166 msg_sys
->msg_name
= &addr
;
2168 if (MSG_CMSG_COMPAT
& flags
)
2169 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2171 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2176 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2177 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2179 /* We assume all kernel code knows the size of sockaddr_storage */
2180 msg_sys
->msg_namelen
= 0;
2182 if (sock
->file
->f_flags
& O_NONBLOCK
)
2183 flags
|= MSG_DONTWAIT
;
2184 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
,
2190 if (uaddr
!= NULL
) {
2191 err
= move_addr_to_user(&addr
,
2192 msg_sys
->msg_namelen
, uaddr
,
2197 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2201 if (MSG_CMSG_COMPAT
& flags
)
2202 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2203 &msg_compat
->msg_controllen
);
2205 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2206 &msg
->msg_controllen
);
2212 if (iov
!= iovstack
)
2218 * BSD recvmsg interface
2221 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2223 int fput_needed
, err
;
2224 struct msghdr msg_sys
;
2225 struct socket
*sock
;
2227 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2231 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2233 fput_light(sock
->file
, fput_needed
);
2238 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2239 unsigned int, flags
)
2241 if (flags
& MSG_CMSG_COMPAT
)
2243 return __sys_recvmsg(fd
, msg
, flags
);
2247 * Linux recvmmsg interface
2250 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2251 unsigned int flags
, struct timespec
*timeout
)
2253 int fput_needed
, err
, datagrams
;
2254 struct socket
*sock
;
2255 struct mmsghdr __user
*entry
;
2256 struct compat_mmsghdr __user
*compat_entry
;
2257 struct msghdr msg_sys
;
2258 struct timespec end_time
;
2261 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2267 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2271 err
= sock_error(sock
->sk
);
2276 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2278 while (datagrams
< vlen
) {
2280 * No need to ask LSM for more than the first datagram.
2282 if (MSG_CMSG_COMPAT
& flags
) {
2283 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2284 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2288 err
= __put_user(err
, &compat_entry
->msg_len
);
2291 err
= ___sys_recvmsg(sock
,
2292 (struct user_msghdr __user
*)entry
,
2293 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2297 err
= put_user(err
, &entry
->msg_len
);
2305 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2306 if (flags
& MSG_WAITFORONE
)
2307 flags
|= MSG_DONTWAIT
;
2310 ktime_get_ts(timeout
);
2311 *timeout
= timespec_sub(end_time
, *timeout
);
2312 if (timeout
->tv_sec
< 0) {
2313 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2317 /* Timeout, return less than vlen datagrams */
2318 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2322 /* Out of band data, return right away */
2323 if (msg_sys
.msg_flags
& MSG_OOB
)
2328 fput_light(sock
->file
, fput_needed
);
2333 if (datagrams
!= 0) {
2335 * We may return less entries than requested (vlen) if the
2336 * sock is non block and there aren't enough datagrams...
2338 if (err
!= -EAGAIN
) {
2340 * ... or if recvmsg returns an error after we
2341 * received some datagrams, where we record the
2342 * error to return on the next call or if the
2343 * app asks about it using getsockopt(SO_ERROR).
2345 sock
->sk
->sk_err
= -err
;
2354 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2355 unsigned int, vlen
, unsigned int, flags
,
2356 struct timespec __user
*, timeout
)
2359 struct timespec timeout_sys
;
2361 if (flags
& MSG_CMSG_COMPAT
)
2365 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2367 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2370 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2372 if (datagrams
> 0 &&
2373 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2374 datagrams
= -EFAULT
;
2379 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2380 /* Argument list sizes for sys_socketcall */
2381 #define AL(x) ((x) * sizeof(unsigned long))
2382 static const unsigned char nargs
[21] = {
2383 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2384 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2385 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2392 * System call vectors.
2394 * Argument checking cleaned up. Saved 20% in size.
2395 * This function doesn't need to set the kernel lock because
2396 * it is set by the callees.
2399 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2401 unsigned long a
[AUDITSC_ARGS
];
2402 unsigned long a0
, a1
;
2406 if (call
< 1 || call
> SYS_SENDMMSG
)
2410 if (len
> sizeof(a
))
2413 /* copy_from_user should be SMP safe. */
2414 if (copy_from_user(a
, args
, len
))
2417 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2426 err
= sys_socket(a0
, a1
, a
[2]);
2429 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2432 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2435 err
= sys_listen(a0
, a1
);
2438 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2439 (int __user
*)a
[2], 0);
2441 case SYS_GETSOCKNAME
:
2443 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2444 (int __user
*)a
[2]);
2446 case SYS_GETPEERNAME
:
2448 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2449 (int __user
*)a
[2]);
2451 case SYS_SOCKETPAIR
:
2452 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2455 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2458 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2459 (struct sockaddr __user
*)a
[4], a
[5]);
2462 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2465 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2466 (struct sockaddr __user
*)a
[4],
2467 (int __user
*)a
[5]);
2470 err
= sys_shutdown(a0
, a1
);
2472 case SYS_SETSOCKOPT
:
2473 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2475 case SYS_GETSOCKOPT
:
2477 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2478 (int __user
*)a
[4]);
2481 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2484 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2487 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2490 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2491 (struct timespec __user
*)a
[4]);
2494 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2495 (int __user
*)a
[2], a
[3]);
2504 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2507 * sock_register - add a socket protocol handler
2508 * @ops: description of protocol
2510 * This function is called by a protocol handler that wants to
2511 * advertise its address family, and have it linked into the
2512 * socket interface. The value ops->family corresponds to the
2513 * socket system call protocol family.
2515 int sock_register(const struct net_proto_family
*ops
)
2519 if (ops
->family
>= NPROTO
) {
2520 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2524 spin_lock(&net_family_lock
);
2525 if (rcu_dereference_protected(net_families
[ops
->family
],
2526 lockdep_is_held(&net_family_lock
)))
2529 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2532 spin_unlock(&net_family_lock
);
2534 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2537 EXPORT_SYMBOL(sock_register
);
2540 * sock_unregister - remove a protocol handler
2541 * @family: protocol family to remove
2543 * This function is called by a protocol handler that wants to
2544 * remove its address family, and have it unlinked from the
2545 * new socket creation.
2547 * If protocol handler is a module, then it can use module reference
2548 * counts to protect against new references. If protocol handler is not
2549 * a module then it needs to provide its own protection in
2550 * the ops->create routine.
2552 void sock_unregister(int family
)
2554 BUG_ON(family
< 0 || family
>= NPROTO
);
2556 spin_lock(&net_family_lock
);
2557 RCU_INIT_POINTER(net_families
[family
], NULL
);
2558 spin_unlock(&net_family_lock
);
2562 pr_info("NET: Unregistered protocol family %d\n", family
);
2564 EXPORT_SYMBOL(sock_unregister
);
2566 static int __init
sock_init(void)
2570 * Initialize the network sysctl infrastructure.
2572 err
= net_sysctl_init();
2577 * Initialize skbuff SLAB cache
2582 * Initialize the protocols module.
2587 err
= register_filesystem(&sock_fs_type
);
2590 sock_mnt
= kern_mount(&sock_fs_type
);
2591 if (IS_ERR(sock_mnt
)) {
2592 err
= PTR_ERR(sock_mnt
);
2596 /* The real protocol initialization is performed in later initcalls.
2599 #ifdef CONFIG_NETFILTER
2600 err
= netfilter_init();
2605 ptp_classifier_init();
2611 unregister_filesystem(&sock_fs_type
);
2616 core_initcall(sock_init
); /* early initcall */
2618 #ifdef CONFIG_PROC_FS
2619 void socket_seq_show(struct seq_file
*seq
)
2624 for_each_possible_cpu(cpu
)
2625 counter
+= per_cpu(sockets_in_use
, cpu
);
2627 /* It can be negative, by the way. 8) */
2631 seq_printf(seq
, "sockets: used %d\n", counter
);
2633 #endif /* CONFIG_PROC_FS */
2635 #ifdef CONFIG_COMPAT
2636 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2637 unsigned int cmd
, void __user
*up
)
2639 mm_segment_t old_fs
= get_fs();
2644 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2647 err
= compat_put_timeval(&ktv
, up
);
2652 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2653 unsigned int cmd
, void __user
*up
)
2655 mm_segment_t old_fs
= get_fs();
2656 struct timespec kts
;
2660 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2663 err
= compat_put_timespec(&kts
, up
);
2668 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2670 struct ifreq __user
*uifr
;
2673 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2674 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2677 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2681 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2687 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2689 struct compat_ifconf ifc32
;
2691 struct ifconf __user
*uifc
;
2692 struct compat_ifreq __user
*ifr32
;
2693 struct ifreq __user
*ifr
;
2697 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2700 memset(&ifc
, 0, sizeof(ifc
));
2701 if (ifc32
.ifcbuf
== 0) {
2705 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2707 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2708 sizeof(struct ifreq
);
2709 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2711 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2712 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2713 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2714 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2720 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2723 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2727 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2731 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2733 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2734 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2735 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2741 if (ifc32
.ifcbuf
== 0) {
2742 /* Translate from 64-bit structure multiple to
2746 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2751 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2757 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2759 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2760 bool convert_in
= false, convert_out
= false;
2761 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2762 struct ethtool_rxnfc __user
*rxnfc
;
2763 struct ifreq __user
*ifr
;
2764 u32 rule_cnt
= 0, actual_rule_cnt
;
2769 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2772 compat_rxnfc
= compat_ptr(data
);
2774 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2777 /* Most ethtool structures are defined without padding.
2778 * Unfortunately struct ethtool_rxnfc is an exception.
2783 case ETHTOOL_GRXCLSRLALL
:
2784 /* Buffer size is variable */
2785 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2787 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2789 buf_size
+= rule_cnt
* sizeof(u32
);
2791 case ETHTOOL_GRXRINGS
:
2792 case ETHTOOL_GRXCLSRLCNT
:
2793 case ETHTOOL_GRXCLSRULE
:
2794 case ETHTOOL_SRXCLSRLINS
:
2797 case ETHTOOL_SRXCLSRLDEL
:
2798 buf_size
+= sizeof(struct ethtool_rxnfc
);
2803 ifr
= compat_alloc_user_space(buf_size
);
2804 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2806 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2809 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2810 &ifr
->ifr_ifru
.ifru_data
))
2814 /* We expect there to be holes between fs.m_ext and
2815 * fs.ring_cookie and at the end of fs, but nowhere else.
2817 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2818 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2819 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2820 sizeof(rxnfc
->fs
.m_ext
));
2822 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2823 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2824 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2825 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2827 if (copy_in_user(rxnfc
, compat_rxnfc
,
2828 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2829 (void __user
*)rxnfc
) ||
2830 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2831 &compat_rxnfc
->fs
.ring_cookie
,
2832 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2833 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2834 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2835 sizeof(rxnfc
->rule_cnt
)))
2839 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2844 if (copy_in_user(compat_rxnfc
, rxnfc
,
2845 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2846 (const void __user
*)rxnfc
) ||
2847 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2848 &rxnfc
->fs
.ring_cookie
,
2849 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2850 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2851 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2852 sizeof(rxnfc
->rule_cnt
)))
2855 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2856 /* As an optimisation, we only copy the actual
2857 * number of rules that the underlying
2858 * function returned. Since Mallory might
2859 * change the rule count in user memory, we
2860 * check that it is less than the rule count
2861 * originally given (as the user buffer size),
2862 * which has been range-checked.
2864 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2866 if (actual_rule_cnt
< rule_cnt
)
2867 rule_cnt
= actual_rule_cnt
;
2868 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2869 &rxnfc
->rule_locs
[0],
2870 rule_cnt
* sizeof(u32
)))
2878 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2881 compat_uptr_t uptr32
;
2882 struct ifreq __user
*uifr
;
2884 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2885 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2888 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2891 uptr
= compat_ptr(uptr32
);
2893 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2896 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2899 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2900 struct compat_ifreq __user
*ifr32
)
2903 mm_segment_t old_fs
;
2907 case SIOCBONDENSLAVE
:
2908 case SIOCBONDRELEASE
:
2909 case SIOCBONDSETHWADDR
:
2910 case SIOCBONDCHANGEACTIVE
:
2911 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2916 err
= dev_ioctl(net
, cmd
,
2917 (struct ifreq __user __force
*) &kifr
);
2922 return -ENOIOCTLCMD
;
2926 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2927 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
2928 struct compat_ifreq __user
*u_ifreq32
)
2930 struct ifreq __user
*u_ifreq64
;
2931 char tmp_buf
[IFNAMSIZ
];
2932 void __user
*data64
;
2935 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2938 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2940 data64
= compat_ptr(data32
);
2942 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2944 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2947 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2950 return dev_ioctl(net
, cmd
, u_ifreq64
);
2953 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2954 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2956 struct ifreq __user
*uifr
;
2959 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2960 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2963 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2974 case SIOCGIFBRDADDR
:
2975 case SIOCGIFDSTADDR
:
2976 case SIOCGIFNETMASK
:
2981 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2989 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2990 struct compat_ifreq __user
*uifr32
)
2993 struct compat_ifmap __user
*uifmap32
;
2994 mm_segment_t old_fs
;
2997 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2998 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2999 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3000 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3001 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3002 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3003 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3004 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3010 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3013 if (cmd
== SIOCGIFMAP
&& !err
) {
3014 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3015 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3016 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3017 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3018 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3019 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3020 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3029 struct sockaddr rt_dst
; /* target address */
3030 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3031 struct sockaddr rt_genmask
; /* target network mask (IP) */
3032 unsigned short rt_flags
;
3035 unsigned char rt_tos
;
3036 unsigned char rt_class
;
3038 short rt_metric
; /* +1 for binary compatibility! */
3039 /* char * */ u32 rt_dev
; /* forcing the device at add */
3040 u32 rt_mtu
; /* per route MTU/Window */
3041 u32 rt_window
; /* Window clamping */
3042 unsigned short rt_irtt
; /* Initial RTT */
3045 struct in6_rtmsg32
{
3046 struct in6_addr rtmsg_dst
;
3047 struct in6_addr rtmsg_src
;
3048 struct in6_addr rtmsg_gateway
;
3058 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3059 unsigned int cmd
, void __user
*argp
)
3063 struct in6_rtmsg r6
;
3067 mm_segment_t old_fs
= get_fs();
3069 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3070 struct in6_rtmsg32 __user
*ur6
= argp
;
3071 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3072 3 * sizeof(struct in6_addr
));
3073 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3074 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3075 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3076 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3077 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3078 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3079 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3083 struct rtentry32 __user
*ur4
= argp
;
3084 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3085 3 * sizeof(struct sockaddr
));
3086 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3087 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3088 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3089 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3090 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3091 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3093 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3094 r4
.rt_dev
= (char __user __force
*)devname
;
3108 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3115 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3116 * for some operations; this forces use of the newer bridge-utils that
3117 * use compatible ioctls
3119 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3123 if (get_user(tmp
, argp
))
3125 if (tmp
== BRCTL_GET_VERSION
)
3126 return BRCTL_VERSION
+ 1;
3130 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3131 unsigned int cmd
, unsigned long arg
)
3133 void __user
*argp
= compat_ptr(arg
);
3134 struct sock
*sk
= sock
->sk
;
3135 struct net
*net
= sock_net(sk
);
3137 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3138 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3143 return old_bridge_ioctl(argp
);
3145 return dev_ifname32(net
, argp
);
3147 return dev_ifconf(net
, argp
);
3149 return ethtool_ioctl(net
, argp
);
3151 return compat_siocwandev(net
, argp
);
3154 return compat_sioc_ifmap(net
, cmd
, argp
);
3155 case SIOCBONDENSLAVE
:
3156 case SIOCBONDRELEASE
:
3157 case SIOCBONDSETHWADDR
:
3158 case SIOCBONDCHANGEACTIVE
:
3159 return bond_ioctl(net
, cmd
, argp
);
3162 return routing_ioctl(net
, sock
, cmd
, argp
);
3164 return do_siocgstamp(net
, sock
, cmd
, argp
);
3166 return do_siocgstampns(net
, sock
, cmd
, argp
);
3167 case SIOCBONDSLAVEINFOQUERY
:
3168 case SIOCBONDINFOQUERY
:
3171 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3183 return sock_ioctl(file
, cmd
, arg
);
3200 case SIOCSIFHWBROADCAST
:
3202 case SIOCGIFBRDADDR
:
3203 case SIOCSIFBRDADDR
:
3204 case SIOCGIFDSTADDR
:
3205 case SIOCSIFDSTADDR
:
3206 case SIOCGIFNETMASK
:
3207 case SIOCSIFNETMASK
:
3218 return dev_ifsioc(net
, sock
, cmd
, argp
);
3224 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3227 return -ENOIOCTLCMD
;
3230 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3233 struct socket
*sock
= file
->private_data
;
3234 int ret
= -ENOIOCTLCMD
;
3241 if (sock
->ops
->compat_ioctl
)
3242 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3244 if (ret
== -ENOIOCTLCMD
&&
3245 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3246 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3248 if (ret
== -ENOIOCTLCMD
)
3249 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3255 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3257 return sock
->ops
->bind(sock
, addr
, addrlen
);
3259 EXPORT_SYMBOL(kernel_bind
);
3261 int kernel_listen(struct socket
*sock
, int backlog
)
3263 return sock
->ops
->listen(sock
, backlog
);
3265 EXPORT_SYMBOL(kernel_listen
);
3267 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3269 struct sock
*sk
= sock
->sk
;
3272 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3277 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3279 sock_release(*newsock
);
3284 (*newsock
)->ops
= sock
->ops
;
3285 __module_get((*newsock
)->ops
->owner
);
3290 EXPORT_SYMBOL(kernel_accept
);
3292 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3295 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3297 EXPORT_SYMBOL(kernel_connect
);
3299 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3302 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3304 EXPORT_SYMBOL(kernel_getsockname
);
3306 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3309 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3311 EXPORT_SYMBOL(kernel_getpeername
);
3313 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3314 char *optval
, int *optlen
)
3316 mm_segment_t oldfs
= get_fs();
3317 char __user
*uoptval
;
3318 int __user
*uoptlen
;
3321 uoptval
= (char __user __force
*) optval
;
3322 uoptlen
= (int __user __force
*) optlen
;
3325 if (level
== SOL_SOCKET
)
3326 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3328 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3333 EXPORT_SYMBOL(kernel_getsockopt
);
3335 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3336 char *optval
, unsigned int optlen
)
3338 mm_segment_t oldfs
= get_fs();
3339 char __user
*uoptval
;
3342 uoptval
= (char __user __force
*) optval
;
3345 if (level
== SOL_SOCKET
)
3346 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3348 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3353 EXPORT_SYMBOL(kernel_setsockopt
);
3355 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3356 size_t size
, int flags
)
3358 if (sock
->ops
->sendpage
)
3359 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3361 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3363 EXPORT_SYMBOL(kernel_sendpage
);
3365 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3367 mm_segment_t oldfs
= get_fs();
3371 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3376 EXPORT_SYMBOL(kernel_sock_ioctl
);
3378 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3380 return sock
->ops
->shutdown(sock
, how
);
3382 EXPORT_SYMBOL(kernel_sock_shutdown
);