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_iter
= sock_read_iter
,
144 .write_iter
= sock_write_iter
,
146 .unlocked_ioctl
= sock_ioctl
,
148 .compat_ioctl
= compat_sock_ioctl
,
151 .release
= sock_close
,
152 .fasync
= sock_fasync
,
153 .sendpage
= sock_sendpage
,
154 .splice_write
= generic_splice_sendpage
,
155 .splice_read
= sock_splice_read
,
159 * The protocol list. Each protocol is registered in here.
162 static DEFINE_SPINLOCK(net_family_lock
);
163 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
166 * Statistics counters of the socket lists
169 static DEFINE_PER_CPU(int, sockets_in_use
);
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
190 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
194 if (copy_from_user(kaddr
, uaddr
, ulen
))
196 return audit_sockaddr(ulen
, kaddr
);
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
216 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
217 void __user
*uaddr
, int __user
*ulen
)
222 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
223 err
= get_user(len
, ulen
);
231 if (audit_sockaddr(klen
, kaddr
))
233 if (copy_to_user(uaddr
, kaddr
, len
))
237 * "fromlen shall refer to the value before truncation.."
240 return __put_user(klen
, ulen
);
243 static struct kmem_cache
*sock_inode_cachep __read_mostly
;
245 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
247 struct socket_alloc
*ei
;
248 struct socket_wq
*wq
;
250 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
253 wq
= kmalloc(sizeof(*wq
), GFP_KERNEL
);
255 kmem_cache_free(sock_inode_cachep
, ei
);
258 init_waitqueue_head(&wq
->wait
);
259 wq
->fasync_list
= NULL
;
261 RCU_INIT_POINTER(ei
->socket
.wq
, wq
);
263 ei
->socket
.state
= SS_UNCONNECTED
;
264 ei
->socket
.flags
= 0;
265 ei
->socket
.ops
= NULL
;
266 ei
->socket
.sk
= NULL
;
267 ei
->socket
.file
= NULL
;
269 return &ei
->vfs_inode
;
272 static void sock_destroy_inode(struct inode
*inode
)
274 struct socket_alloc
*ei
;
275 struct socket_wq
*wq
;
277 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
278 wq
= rcu_dereference_protected(ei
->socket
.wq
, 1);
280 kmem_cache_free(sock_inode_cachep
, ei
);
283 static void init_once(void *foo
)
285 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
287 inode_init_once(&ei
->vfs_inode
);
290 static int init_inodecache(void)
292 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
293 sizeof(struct socket_alloc
),
295 (SLAB_HWCACHE_ALIGN
|
296 SLAB_RECLAIM_ACCOUNT
|
297 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
299 if (sock_inode_cachep
== NULL
)
304 static const struct super_operations sockfs_ops
= {
305 .alloc_inode
= sock_alloc_inode
,
306 .destroy_inode
= sock_destroy_inode
,
307 .statfs
= simple_statfs
,
311 * sockfs_dname() is called from d_path().
313 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
315 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
316 d_inode(dentry
)->i_ino
);
319 static const struct dentry_operations sockfs_dentry_operations
= {
320 .d_dname
= sockfs_dname
,
323 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
324 int flags
, const char *dev_name
, void *data
)
326 return mount_pseudo(fs_type
, "socket:", &sockfs_ops
,
327 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
330 static struct vfsmount
*sock_mnt __read_mostly
;
332 static struct file_system_type sock_fs_type
= {
334 .mount
= sockfs_mount
,
335 .kill_sb
= kill_anon_super
,
339 * Obtains the first available file descriptor and sets it up for use.
341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
355 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
357 struct qstr name
= { .name
= "" };
363 name
.len
= strlen(name
.name
);
364 } else if (sock
->sk
) {
365 name
.name
= sock
->sk
->sk_prot_creator
->name
;
366 name
.len
= strlen(name
.name
);
368 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
369 if (unlikely(!path
.dentry
))
370 return ERR_PTR(-ENOMEM
);
371 path
.mnt
= mntget(sock_mnt
);
373 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
375 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
378 /* drop dentry, keep inode */
379 ihold(d_inode(path
.dentry
));
385 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
386 file
->private_data
= sock
;
389 EXPORT_SYMBOL(sock_alloc_file
);
391 static int sock_map_fd(struct socket
*sock
, int flags
)
393 struct file
*newfile
;
394 int fd
= get_unused_fd_flags(flags
);
395 if (unlikely(fd
< 0))
398 newfile
= sock_alloc_file(sock
, flags
, NULL
);
399 if (likely(!IS_ERR(newfile
))) {
400 fd_install(fd
, newfile
);
405 return PTR_ERR(newfile
);
408 struct socket
*sock_from_file(struct file
*file
, int *err
)
410 if (file
->f_op
== &socket_file_ops
)
411 return file
->private_data
; /* set in sock_map_fd */
416 EXPORT_SYMBOL(sock_from_file
);
419 * sockfd_lookup - Go from a file number to its socket slot
421 * @err: pointer to an error code return
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
428 * On a success the socket object pointer is returned.
431 struct socket
*sockfd_lookup(int fd
, int *err
)
442 sock
= sock_from_file(file
, err
);
447 EXPORT_SYMBOL(sockfd_lookup
);
449 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
451 struct fd f
= fdget(fd
);
456 sock
= sock_from_file(f
.file
, err
);
458 *fput_needed
= f
.flags
;
466 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469 static ssize_t
sockfs_getxattr(struct dentry
*dentry
, struct inode
*inode
,
470 const char *name
, void *value
, size_t size
)
472 const char *proto_name
;
477 if (!strncmp(name
, XATTR_NAME_SOCKPROTONAME
, XATTR_NAME_SOCKPROTONAME_LEN
)) {
478 proto_name
= dentry
->d_name
.name
;
479 proto_size
= strlen(proto_name
);
483 if (proto_size
+ 1 > size
)
486 strncpy(value
, proto_name
, proto_size
+ 1);
488 error
= proto_size
+ 1;
495 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
501 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
511 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
516 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
523 static const struct inode_operations sockfs_inode_ops
= {
524 .getxattr
= sockfs_getxattr
,
525 .listxattr
= sockfs_listxattr
,
529 * sock_alloc - allocate a socket
531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
536 struct socket
*sock_alloc(void)
541 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
545 sock
= SOCKET_I(inode
);
547 kmemcheck_annotate_bitfield(sock
, type
);
548 inode
->i_ino
= get_next_ino();
549 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
550 inode
->i_uid
= current_fsuid();
551 inode
->i_gid
= current_fsgid();
552 inode
->i_op
= &sockfs_inode_ops
;
554 this_cpu_add(sockets_in_use
, 1);
557 EXPORT_SYMBOL(sock_alloc
);
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 this_cpu_sub(sockets_in_use
, 1);
583 iput(SOCK_INODE(sock
));
588 EXPORT_SYMBOL(sock_release
);
590 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
592 u8 flags
= *tx_flags
;
594 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
595 flags
|= SKBTX_HW_TSTAMP
;
597 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
598 flags
|= SKBTX_SW_TSTAMP
;
600 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
601 flags
|= SKBTX_SCHED_TSTAMP
;
605 EXPORT_SYMBOL(__sock_tx_timestamp
);
607 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
609 int ret
= sock
->ops
->sendmsg(sock
, msg
, msg_data_left(msg
));
610 BUG_ON(ret
== -EIOCBQUEUED
);
614 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
616 int err
= security_socket_sendmsg(sock
, msg
,
619 return err
?: sock_sendmsg_nosec(sock
, msg
);
621 EXPORT_SYMBOL(sock_sendmsg
);
623 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
624 struct kvec
*vec
, size_t num
, size_t size
)
626 iov_iter_kvec(&msg
->msg_iter
, WRITE
| ITER_KVEC
, vec
, num
, size
);
627 return sock_sendmsg(sock
, msg
);
629 EXPORT_SYMBOL(kernel_sendmsg
);
632 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
634 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
637 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
638 struct scm_timestamping tss
;
640 struct skb_shared_hwtstamps
*shhwtstamps
=
643 /* Race occurred between timestamp enabling and packet
644 receiving. Fill in the current time for now. */
645 if (need_software_tstamp
&& skb
->tstamp
.tv64
== 0)
646 __net_timestamp(skb
);
648 if (need_software_tstamp
) {
649 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
651 skb_get_timestamp(skb
, &tv
);
652 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
656 skb_get_timestampns(skb
, &ts
);
657 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
662 memset(&tss
, 0, sizeof(tss
));
663 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
664 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
667 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
668 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2))
671 put_cmsg(msg
, SOL_SOCKET
,
672 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
674 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
676 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
681 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
683 if (!skb
->wifi_acked_valid
)
686 ack
= skb
->wifi_acked
;
688 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
690 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
692 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
695 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
696 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
697 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
700 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
703 sock_recv_timestamp(msg
, sk
, skb
);
704 sock_recv_drops(msg
, sk
, skb
);
706 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
708 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
711 return sock
->ops
->recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
714 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
716 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
718 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
720 EXPORT_SYMBOL(sock_recvmsg
);
723 * kernel_recvmsg - Receive a message from a socket (kernel space)
724 * @sock: The socket to receive the message from
725 * @msg: Received message
726 * @vec: Input s/g array for message data
727 * @num: Size of input s/g array
728 * @size: Number of bytes to read
729 * @flags: Message flags (MSG_DONTWAIT, etc...)
731 * On return the msg structure contains the scatter/gather array passed in the
732 * vec argument. The array is modified so that it consists of the unfilled
733 * portion of the original array.
735 * The returned value is the total number of bytes received, or an error.
737 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
738 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
740 mm_segment_t oldfs
= get_fs();
743 iov_iter_kvec(&msg
->msg_iter
, READ
| ITER_KVEC
, vec
, num
, size
);
745 result
= sock_recvmsg(sock
, msg
, flags
);
749 EXPORT_SYMBOL(kernel_recvmsg
);
751 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
752 int offset
, size_t size
, loff_t
*ppos
, int more
)
757 sock
= file
->private_data
;
759 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
760 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
763 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
766 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
767 struct pipe_inode_info
*pipe
, size_t len
,
770 struct socket
*sock
= file
->private_data
;
772 if (unlikely(!sock
->ops
->splice_read
))
775 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
778 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
780 struct file
*file
= iocb
->ki_filp
;
781 struct socket
*sock
= file
->private_data
;
782 struct msghdr msg
= {.msg_iter
= *to
,
786 if (file
->f_flags
& O_NONBLOCK
)
787 msg
.msg_flags
= MSG_DONTWAIT
;
789 if (iocb
->ki_pos
!= 0)
792 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
795 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
800 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
802 struct file
*file
= iocb
->ki_filp
;
803 struct socket
*sock
= file
->private_data
;
804 struct msghdr msg
= {.msg_iter
= *from
,
808 if (iocb
->ki_pos
!= 0)
811 if (file
->f_flags
& O_NONBLOCK
)
812 msg
.msg_flags
= MSG_DONTWAIT
;
814 if (sock
->type
== SOCK_SEQPACKET
)
815 msg
.msg_flags
|= MSG_EOR
;
817 res
= sock_sendmsg(sock
, &msg
);
818 *from
= msg
.msg_iter
;
823 * Atomic setting of ioctl hooks to avoid race
824 * with module unload.
827 static DEFINE_MUTEX(br_ioctl_mutex
);
828 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
830 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
832 mutex_lock(&br_ioctl_mutex
);
833 br_ioctl_hook
= hook
;
834 mutex_unlock(&br_ioctl_mutex
);
836 EXPORT_SYMBOL(brioctl_set
);
838 static DEFINE_MUTEX(vlan_ioctl_mutex
);
839 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
841 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
843 mutex_lock(&vlan_ioctl_mutex
);
844 vlan_ioctl_hook
= hook
;
845 mutex_unlock(&vlan_ioctl_mutex
);
847 EXPORT_SYMBOL(vlan_ioctl_set
);
849 static DEFINE_MUTEX(dlci_ioctl_mutex
);
850 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
852 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
854 mutex_lock(&dlci_ioctl_mutex
);
855 dlci_ioctl_hook
= hook
;
856 mutex_unlock(&dlci_ioctl_mutex
);
858 EXPORT_SYMBOL(dlci_ioctl_set
);
860 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
861 unsigned int cmd
, unsigned long arg
)
864 void __user
*argp
= (void __user
*)arg
;
866 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
869 * If this ioctl is unknown try to hand it down
872 if (err
== -ENOIOCTLCMD
)
873 err
= dev_ioctl(net
, cmd
, argp
);
879 * With an ioctl, arg may well be a user mode pointer, but we don't know
880 * what to do with it - that's up to the protocol still.
883 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
887 void __user
*argp
= (void __user
*)arg
;
891 sock
= file
->private_data
;
894 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
895 err
= dev_ioctl(net
, cmd
, argp
);
897 #ifdef CONFIG_WEXT_CORE
898 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
899 err
= dev_ioctl(net
, cmd
, argp
);
906 if (get_user(pid
, (int __user
*)argp
))
908 f_setown(sock
->file
, pid
, 1);
913 err
= put_user(f_getown(sock
->file
),
922 request_module("bridge");
924 mutex_lock(&br_ioctl_mutex
);
926 err
= br_ioctl_hook(net
, cmd
, argp
);
927 mutex_unlock(&br_ioctl_mutex
);
932 if (!vlan_ioctl_hook
)
933 request_module("8021q");
935 mutex_lock(&vlan_ioctl_mutex
);
937 err
= vlan_ioctl_hook(net
, argp
);
938 mutex_unlock(&vlan_ioctl_mutex
);
943 if (!dlci_ioctl_hook
)
944 request_module("dlci");
946 mutex_lock(&dlci_ioctl_mutex
);
948 err
= dlci_ioctl_hook(cmd
, argp
);
949 mutex_unlock(&dlci_ioctl_mutex
);
952 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
958 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
961 struct socket
*sock
= NULL
;
963 err
= security_socket_create(family
, type
, protocol
, 1);
974 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
986 EXPORT_SYMBOL(sock_create_lite
);
988 /* No kernel lock held - perfect */
989 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
991 unsigned int busy_flag
= 0;
995 * We can't return errors to poll, so it's either yes or no.
997 sock
= file
->private_data
;
999 if (sk_can_busy_loop(sock
->sk
)) {
1000 /* this socket can poll_ll so tell the system call */
1001 busy_flag
= POLL_BUSY_LOOP
;
1003 /* once, only if requested by syscall */
1004 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1005 sk_busy_loop(sock
->sk
, 1);
1008 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1011 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1013 struct socket
*sock
= file
->private_data
;
1015 return sock
->ops
->mmap(file
, sock
, vma
);
1018 static int sock_close(struct inode
*inode
, struct file
*filp
)
1020 sock_release(SOCKET_I(inode
));
1025 * Update the socket async list
1027 * Fasync_list locking strategy.
1029 * 1. fasync_list is modified only under process context socket lock
1030 * i.e. under semaphore.
1031 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1032 * or under socket lock
1035 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1037 struct socket
*sock
= filp
->private_data
;
1038 struct sock
*sk
= sock
->sk
;
1039 struct socket_wq
*wq
;
1045 wq
= rcu_dereference_protected(sock
->wq
, lockdep_sock_is_held(sk
));
1046 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1048 if (!wq
->fasync_list
)
1049 sock_reset_flag(sk
, SOCK_FASYNC
);
1051 sock_set_flag(sk
, SOCK_FASYNC
);
1057 /* This function may be called only under rcu_lock */
1059 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1061 if (!wq
|| !wq
->fasync_list
)
1065 case SOCK_WAKE_WAITD
:
1066 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1069 case SOCK_WAKE_SPACE
:
1070 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1075 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1078 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1083 EXPORT_SYMBOL(sock_wake_async
);
1085 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1086 struct socket
**res
, int kern
)
1089 struct socket
*sock
;
1090 const struct net_proto_family
*pf
;
1093 * Check protocol is in range
1095 if (family
< 0 || family
>= NPROTO
)
1096 return -EAFNOSUPPORT
;
1097 if (type
< 0 || type
>= SOCK_MAX
)
1102 This uglymoron is moved from INET layer to here to avoid
1103 deadlock in module load.
1105 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1106 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1111 err
= security_socket_create(family
, type
, protocol
, kern
);
1116 * Allocate the socket and allow the family to set things up. if
1117 * the protocol is 0, the family is instructed to select an appropriate
1120 sock
= sock_alloc();
1122 net_warn_ratelimited("socket: no more sockets\n");
1123 return -ENFILE
; /* Not exactly a match, but its the
1124 closest posix thing */
1129 #ifdef CONFIG_MODULES
1130 /* Attempt to load a protocol module if the find failed.
1132 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1133 * requested real, full-featured networking support upon configuration.
1134 * Otherwise module support will break!
1136 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1137 request_module("net-pf-%d", family
);
1141 pf
= rcu_dereference(net_families
[family
]);
1142 err
= -EAFNOSUPPORT
;
1147 * We will call the ->create function, that possibly is in a loadable
1148 * module, so we have to bump that loadable module refcnt first.
1150 if (!try_module_get(pf
->owner
))
1153 /* Now protected by module ref count */
1156 err
= pf
->create(net
, sock
, protocol
, kern
);
1158 goto out_module_put
;
1161 * Now to bump the refcnt of the [loadable] module that owns this
1162 * socket at sock_release time we decrement its refcnt.
1164 if (!try_module_get(sock
->ops
->owner
))
1165 goto out_module_busy
;
1168 * Now that we're done with the ->create function, the [loadable]
1169 * module can have its refcnt decremented
1171 module_put(pf
->owner
);
1172 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1174 goto out_sock_release
;
1180 err
= -EAFNOSUPPORT
;
1183 module_put(pf
->owner
);
1190 goto out_sock_release
;
1192 EXPORT_SYMBOL(__sock_create
);
1194 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1196 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1198 EXPORT_SYMBOL(sock_create
);
1200 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1202 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1204 EXPORT_SYMBOL(sock_create_kern
);
1206 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1209 struct socket
*sock
;
1212 /* Check the SOCK_* constants for consistency. */
1213 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1214 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1215 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1216 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1218 flags
= type
& ~SOCK_TYPE_MASK
;
1219 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1221 type
&= SOCK_TYPE_MASK
;
1223 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1224 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1226 retval
= sock_create(family
, type
, protocol
, &sock
);
1230 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1235 /* It may be already another descriptor 8) Not kernel problem. */
1244 * Create a pair of connected sockets.
1247 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1248 int __user
*, usockvec
)
1250 struct socket
*sock1
, *sock2
;
1252 struct file
*newfile1
, *newfile2
;
1255 flags
= type
& ~SOCK_TYPE_MASK
;
1256 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1258 type
&= SOCK_TYPE_MASK
;
1260 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1261 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1264 * Obtain the first socket and check if the underlying protocol
1265 * supports the socketpair call.
1268 err
= sock_create(family
, type
, protocol
, &sock1
);
1272 err
= sock_create(family
, type
, protocol
, &sock2
);
1276 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1278 goto out_release_both
;
1280 fd1
= get_unused_fd_flags(flags
);
1281 if (unlikely(fd1
< 0)) {
1283 goto out_release_both
;
1286 fd2
= get_unused_fd_flags(flags
);
1287 if (unlikely(fd2
< 0)) {
1289 goto out_put_unused_1
;
1292 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1293 if (IS_ERR(newfile1
)) {
1294 err
= PTR_ERR(newfile1
);
1295 goto out_put_unused_both
;
1298 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1299 if (IS_ERR(newfile2
)) {
1300 err
= PTR_ERR(newfile2
);
1304 err
= put_user(fd1
, &usockvec
[0]);
1308 err
= put_user(fd2
, &usockvec
[1]);
1312 audit_fd_pair(fd1
, fd2
);
1314 fd_install(fd1
, newfile1
);
1315 fd_install(fd2
, newfile2
);
1316 /* fd1 and fd2 may be already another descriptors.
1317 * Not kernel problem.
1333 sock_release(sock2
);
1336 out_put_unused_both
:
1341 sock_release(sock2
);
1343 sock_release(sock1
);
1349 * Bind a name to a socket. Nothing much to do here since it's
1350 * the protocol's responsibility to handle the local address.
1352 * We move the socket address to kernel space before we call
1353 * the protocol layer (having also checked the address is ok).
1356 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1358 struct socket
*sock
;
1359 struct sockaddr_storage address
;
1360 int err
, fput_needed
;
1362 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1364 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1366 err
= security_socket_bind(sock
,
1367 (struct sockaddr
*)&address
,
1370 err
= sock
->ops
->bind(sock
,
1374 fput_light(sock
->file
, fput_needed
);
1380 * Perform a listen. Basically, we allow the protocol to do anything
1381 * necessary for a listen, and if that works, we mark the socket as
1382 * ready for listening.
1385 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1387 struct socket
*sock
;
1388 int err
, fput_needed
;
1391 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1393 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1394 if ((unsigned int)backlog
> somaxconn
)
1395 backlog
= somaxconn
;
1397 err
= security_socket_listen(sock
, backlog
);
1399 err
= sock
->ops
->listen(sock
, backlog
);
1401 fput_light(sock
->file
, fput_needed
);
1407 * For accept, we attempt to create a new socket, set up the link
1408 * with the client, wake up the client, then return the new
1409 * connected fd. We collect the address of the connector in kernel
1410 * space and move it to user at the very end. This is unclean because
1411 * we open the socket then return an error.
1413 * 1003.1g adds the ability to recvmsg() to query connection pending
1414 * status to recvmsg. We need to add that support in a way thats
1415 * clean when we restucture accept also.
1418 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1419 int __user
*, upeer_addrlen
, int, flags
)
1421 struct socket
*sock
, *newsock
;
1422 struct file
*newfile
;
1423 int err
, len
, newfd
, fput_needed
;
1424 struct sockaddr_storage address
;
1426 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1429 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1430 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1432 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1437 newsock
= sock_alloc();
1441 newsock
->type
= sock
->type
;
1442 newsock
->ops
= sock
->ops
;
1445 * We don't need try_module_get here, as the listening socket (sock)
1446 * has the protocol module (sock->ops->owner) held.
1448 __module_get(newsock
->ops
->owner
);
1450 newfd
= get_unused_fd_flags(flags
);
1451 if (unlikely(newfd
< 0)) {
1453 sock_release(newsock
);
1456 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1457 if (IS_ERR(newfile
)) {
1458 err
= PTR_ERR(newfile
);
1459 put_unused_fd(newfd
);
1460 sock_release(newsock
);
1464 err
= security_socket_accept(sock
, newsock
);
1468 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1472 if (upeer_sockaddr
) {
1473 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1475 err
= -ECONNABORTED
;
1478 err
= move_addr_to_user(&address
,
1479 len
, upeer_sockaddr
, upeer_addrlen
);
1484 /* File flags are not inherited via accept() unlike another OSes. */
1486 fd_install(newfd
, newfile
);
1490 fput_light(sock
->file
, fput_needed
);
1495 put_unused_fd(newfd
);
1499 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1500 int __user
*, upeer_addrlen
)
1502 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1506 * Attempt to connect to a socket with the server address. The address
1507 * is in user space so we verify it is OK and move it to kernel space.
1509 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1512 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1513 * other SEQPACKET protocols that take time to connect() as it doesn't
1514 * include the -EINPROGRESS status for such sockets.
1517 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1520 struct socket
*sock
;
1521 struct sockaddr_storage address
;
1522 int err
, fput_needed
;
1524 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1527 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1532 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1536 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1537 sock
->file
->f_flags
);
1539 fput_light(sock
->file
, fput_needed
);
1545 * Get the local address ('name') of a socket object. Move the obtained
1546 * name to user space.
1549 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1550 int __user
*, usockaddr_len
)
1552 struct socket
*sock
;
1553 struct sockaddr_storage address
;
1554 int len
, err
, fput_needed
;
1556 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1560 err
= security_socket_getsockname(sock
);
1564 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1567 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1570 fput_light(sock
->file
, fput_needed
);
1576 * Get the remote address ('name') of a socket object. Move the obtained
1577 * name to user space.
1580 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1581 int __user
*, usockaddr_len
)
1583 struct socket
*sock
;
1584 struct sockaddr_storage address
;
1585 int len
, err
, fput_needed
;
1587 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1589 err
= security_socket_getpeername(sock
);
1591 fput_light(sock
->file
, fput_needed
);
1596 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1599 err
= move_addr_to_user(&address
, len
, usockaddr
,
1601 fput_light(sock
->file
, fput_needed
);
1607 * Send a datagram to a given address. We move the address into kernel
1608 * space and check the user space data area is readable before invoking
1612 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1613 unsigned int, flags
, struct sockaddr __user
*, addr
,
1616 struct socket
*sock
;
1617 struct sockaddr_storage address
;
1623 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
1626 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1630 msg
.msg_name
= NULL
;
1631 msg
.msg_control
= NULL
;
1632 msg
.msg_controllen
= 0;
1633 msg
.msg_namelen
= 0;
1635 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1638 msg
.msg_name
= (struct sockaddr
*)&address
;
1639 msg
.msg_namelen
= addr_len
;
1641 if (sock
->file
->f_flags
& O_NONBLOCK
)
1642 flags
|= MSG_DONTWAIT
;
1643 msg
.msg_flags
= flags
;
1644 err
= sock_sendmsg(sock
, &msg
);
1647 fput_light(sock
->file
, fput_needed
);
1653 * Send a datagram down a socket.
1656 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1657 unsigned int, flags
)
1659 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1663 * Receive a frame from the socket and optionally record the address of the
1664 * sender. We verify the buffers are writable and if needed move the
1665 * sender address from kernel to user space.
1668 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1669 unsigned int, flags
, struct sockaddr __user
*, addr
,
1670 int __user
*, addr_len
)
1672 struct socket
*sock
;
1675 struct sockaddr_storage address
;
1679 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
1682 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1686 msg
.msg_control
= NULL
;
1687 msg
.msg_controllen
= 0;
1688 /* Save some cycles and don't copy the address if not needed */
1689 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1690 /* We assume all kernel code knows the size of sockaddr_storage */
1691 msg
.msg_namelen
= 0;
1692 msg
.msg_iocb
= NULL
;
1693 if (sock
->file
->f_flags
& O_NONBLOCK
)
1694 flags
|= MSG_DONTWAIT
;
1695 err
= sock_recvmsg(sock
, &msg
, flags
);
1697 if (err
>= 0 && addr
!= NULL
) {
1698 err2
= move_addr_to_user(&address
,
1699 msg
.msg_namelen
, addr
, addr_len
);
1704 fput_light(sock
->file
, fput_needed
);
1710 * Receive a datagram from a socket.
1713 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1714 unsigned int, flags
)
1716 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1720 * Set a socket option. Because we don't know the option lengths we have
1721 * to pass the user mode parameter for the protocols to sort out.
1724 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1725 char __user
*, optval
, int, optlen
)
1727 int err
, fput_needed
;
1728 struct socket
*sock
;
1733 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1735 err
= security_socket_setsockopt(sock
, level
, optname
);
1739 if (level
== SOL_SOCKET
)
1741 sock_setsockopt(sock
, level
, optname
, optval
,
1745 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1748 fput_light(sock
->file
, fput_needed
);
1754 * Get a socket option. Because we don't know the option lengths we have
1755 * to pass a user mode parameter for the protocols to sort out.
1758 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1759 char __user
*, optval
, int __user
*, optlen
)
1761 int err
, fput_needed
;
1762 struct socket
*sock
;
1764 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1766 err
= security_socket_getsockopt(sock
, level
, optname
);
1770 if (level
== SOL_SOCKET
)
1772 sock_getsockopt(sock
, level
, optname
, optval
,
1776 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1779 fput_light(sock
->file
, fput_needed
);
1785 * Shutdown a socket.
1788 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1790 int err
, fput_needed
;
1791 struct socket
*sock
;
1793 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1795 err
= security_socket_shutdown(sock
, how
);
1797 err
= sock
->ops
->shutdown(sock
, how
);
1798 fput_light(sock
->file
, fput_needed
);
1803 /* A couple of helpful macros for getting the address of the 32/64 bit
1804 * fields which are the same type (int / unsigned) on our platforms.
1806 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1807 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1808 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1810 struct used_address
{
1811 struct sockaddr_storage name
;
1812 unsigned int name_len
;
1815 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
1816 struct user_msghdr __user
*umsg
,
1817 struct sockaddr __user
**save_addr
,
1820 struct sockaddr __user
*uaddr
;
1821 struct iovec __user
*uiov
;
1825 if (!access_ok(VERIFY_READ
, umsg
, sizeof(*umsg
)) ||
1826 __get_user(uaddr
, &umsg
->msg_name
) ||
1827 __get_user(kmsg
->msg_namelen
, &umsg
->msg_namelen
) ||
1828 __get_user(uiov
, &umsg
->msg_iov
) ||
1829 __get_user(nr_segs
, &umsg
->msg_iovlen
) ||
1830 __get_user(kmsg
->msg_control
, &umsg
->msg_control
) ||
1831 __get_user(kmsg
->msg_controllen
, &umsg
->msg_controllen
) ||
1832 __get_user(kmsg
->msg_flags
, &umsg
->msg_flags
))
1836 kmsg
->msg_namelen
= 0;
1838 if (kmsg
->msg_namelen
< 0)
1841 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1842 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1847 if (uaddr
&& kmsg
->msg_namelen
) {
1849 err
= move_addr_to_kernel(uaddr
, kmsg
->msg_namelen
,
1855 kmsg
->msg_name
= NULL
;
1856 kmsg
->msg_namelen
= 0;
1859 if (nr_segs
> UIO_MAXIOV
)
1862 kmsg
->msg_iocb
= NULL
;
1864 return import_iovec(save_addr
? READ
: WRITE
, uiov
, nr_segs
,
1865 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
1868 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
1869 struct msghdr
*msg_sys
, unsigned int flags
,
1870 struct used_address
*used_address
,
1871 unsigned int allowed_msghdr_flags
)
1873 struct compat_msghdr __user
*msg_compat
=
1874 (struct compat_msghdr __user
*)msg
;
1875 struct sockaddr_storage address
;
1876 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1877 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1878 __attribute__ ((aligned(sizeof(__kernel_size_t
))));
1879 /* 20 is size of ipv6_pktinfo */
1880 unsigned char *ctl_buf
= ctl
;
1884 msg_sys
->msg_name
= &address
;
1886 if (MSG_CMSG_COMPAT
& flags
)
1887 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
1889 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
1895 if (msg_sys
->msg_controllen
> INT_MAX
)
1897 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
1898 ctl_len
= msg_sys
->msg_controllen
;
1899 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1901 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
1905 ctl_buf
= msg_sys
->msg_control
;
1906 ctl_len
= msg_sys
->msg_controllen
;
1907 } else if (ctl_len
) {
1908 if (ctl_len
> sizeof(ctl
)) {
1909 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1910 if (ctl_buf
== NULL
)
1915 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1916 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1917 * checking falls down on this.
1919 if (copy_from_user(ctl_buf
,
1920 (void __user __force
*)msg_sys
->msg_control
,
1923 msg_sys
->msg_control
= ctl_buf
;
1925 msg_sys
->msg_flags
= flags
;
1927 if (sock
->file
->f_flags
& O_NONBLOCK
)
1928 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
1930 * If this is sendmmsg() and current destination address is same as
1931 * previously succeeded address, omit asking LSM's decision.
1932 * used_address->name_len is initialized to UINT_MAX so that the first
1933 * destination address never matches.
1935 if (used_address
&& msg_sys
->msg_name
&&
1936 used_address
->name_len
== msg_sys
->msg_namelen
&&
1937 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
1938 used_address
->name_len
)) {
1939 err
= sock_sendmsg_nosec(sock
, msg_sys
);
1942 err
= sock_sendmsg(sock
, msg_sys
);
1944 * If this is sendmmsg() and sending to current destination address was
1945 * successful, remember it.
1947 if (used_address
&& err
>= 0) {
1948 used_address
->name_len
= msg_sys
->msg_namelen
;
1949 if (msg_sys
->msg_name
)
1950 memcpy(&used_address
->name
, msg_sys
->msg_name
,
1951 used_address
->name_len
);
1956 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
1963 * BSD sendmsg interface
1966 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
1968 int fput_needed
, err
;
1969 struct msghdr msg_sys
;
1970 struct socket
*sock
;
1972 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1976 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
1978 fput_light(sock
->file
, fput_needed
);
1983 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
1985 if (flags
& MSG_CMSG_COMPAT
)
1987 return __sys_sendmsg(fd
, msg
, flags
);
1991 * Linux sendmmsg interface
1994 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
1997 int fput_needed
, err
, datagrams
;
1998 struct socket
*sock
;
1999 struct mmsghdr __user
*entry
;
2000 struct compat_mmsghdr __user
*compat_entry
;
2001 struct msghdr msg_sys
;
2002 struct used_address used_address
;
2003 unsigned int oflags
= flags
;
2005 if (vlen
> UIO_MAXIOV
)
2010 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2014 used_address
.name_len
= UINT_MAX
;
2016 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2020 while (datagrams
< vlen
) {
2021 if (datagrams
== vlen
- 1)
2024 if (MSG_CMSG_COMPAT
& flags
) {
2025 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2026 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2029 err
= __put_user(err
, &compat_entry
->msg_len
);
2032 err
= ___sys_sendmsg(sock
,
2033 (struct user_msghdr __user
*)entry
,
2034 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2037 err
= put_user(err
, &entry
->msg_len
);
2047 fput_light(sock
->file
, fput_needed
);
2049 /* We only return an error if no datagrams were able to be sent */
2056 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2057 unsigned int, vlen
, unsigned int, flags
)
2059 if (flags
& MSG_CMSG_COMPAT
)
2061 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2064 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2065 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2067 struct compat_msghdr __user
*msg_compat
=
2068 (struct compat_msghdr __user
*)msg
;
2069 struct iovec iovstack
[UIO_FASTIOV
];
2070 struct iovec
*iov
= iovstack
;
2071 unsigned long cmsg_ptr
;
2075 /* kernel mode address */
2076 struct sockaddr_storage addr
;
2078 /* user mode address pointers */
2079 struct sockaddr __user
*uaddr
;
2080 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2082 msg_sys
->msg_name
= &addr
;
2084 if (MSG_CMSG_COMPAT
& flags
)
2085 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2087 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2091 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2092 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2094 /* We assume all kernel code knows the size of sockaddr_storage */
2095 msg_sys
->msg_namelen
= 0;
2097 if (sock
->file
->f_flags
& O_NONBLOCK
)
2098 flags
|= MSG_DONTWAIT
;
2099 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
, flags
);
2104 if (uaddr
!= NULL
) {
2105 err
= move_addr_to_user(&addr
,
2106 msg_sys
->msg_namelen
, uaddr
,
2111 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2115 if (MSG_CMSG_COMPAT
& flags
)
2116 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2117 &msg_compat
->msg_controllen
);
2119 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2120 &msg
->msg_controllen
);
2131 * BSD recvmsg interface
2134 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2136 int fput_needed
, err
;
2137 struct msghdr msg_sys
;
2138 struct socket
*sock
;
2140 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2144 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2146 fput_light(sock
->file
, fput_needed
);
2151 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2152 unsigned int, flags
)
2154 if (flags
& MSG_CMSG_COMPAT
)
2156 return __sys_recvmsg(fd
, msg
, flags
);
2160 * Linux recvmmsg interface
2163 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2164 unsigned int flags
, struct timespec
*timeout
)
2166 int fput_needed
, err
, datagrams
;
2167 struct socket
*sock
;
2168 struct mmsghdr __user
*entry
;
2169 struct compat_mmsghdr __user
*compat_entry
;
2170 struct msghdr msg_sys
;
2171 struct timespec64 end_time
;
2172 struct timespec64 timeout64
;
2175 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2181 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2185 err
= sock_error(sock
->sk
);
2190 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2192 while (datagrams
< vlen
) {
2194 * No need to ask LSM for more than the first datagram.
2196 if (MSG_CMSG_COMPAT
& flags
) {
2197 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2198 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2202 err
= __put_user(err
, &compat_entry
->msg_len
);
2205 err
= ___sys_recvmsg(sock
,
2206 (struct user_msghdr __user
*)entry
,
2207 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2211 err
= put_user(err
, &entry
->msg_len
);
2219 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2220 if (flags
& MSG_WAITFORONE
)
2221 flags
|= MSG_DONTWAIT
;
2224 ktime_get_ts64(&timeout64
);
2225 *timeout
= timespec64_to_timespec(
2226 timespec64_sub(end_time
, timeout64
));
2227 if (timeout
->tv_sec
< 0) {
2228 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2232 /* Timeout, return less than vlen datagrams */
2233 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2237 /* Out of band data, return right away */
2238 if (msg_sys
.msg_flags
& MSG_OOB
)
2246 if (datagrams
== 0) {
2252 * We may return less entries than requested (vlen) if the
2253 * sock is non block and there aren't enough datagrams...
2255 if (err
!= -EAGAIN
) {
2257 * ... or if recvmsg returns an error after we
2258 * received some datagrams, where we record the
2259 * error to return on the next call or if the
2260 * app asks about it using getsockopt(SO_ERROR).
2262 sock
->sk
->sk_err
= -err
;
2265 fput_light(sock
->file
, fput_needed
);
2270 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2271 unsigned int, vlen
, unsigned int, flags
,
2272 struct timespec __user
*, timeout
)
2275 struct timespec timeout_sys
;
2277 if (flags
& MSG_CMSG_COMPAT
)
2281 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2283 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2286 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2288 if (datagrams
> 0 &&
2289 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2290 datagrams
= -EFAULT
;
2295 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2296 /* Argument list sizes for sys_socketcall */
2297 #define AL(x) ((x) * sizeof(unsigned long))
2298 static const unsigned char nargs
[21] = {
2299 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2300 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2301 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2308 * System call vectors.
2310 * Argument checking cleaned up. Saved 20% in size.
2311 * This function doesn't need to set the kernel lock because
2312 * it is set by the callees.
2315 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2317 unsigned long a
[AUDITSC_ARGS
];
2318 unsigned long a0
, a1
;
2322 if (call
< 1 || call
> SYS_SENDMMSG
)
2326 if (len
> sizeof(a
))
2329 /* copy_from_user should be SMP safe. */
2330 if (copy_from_user(a
, args
, len
))
2333 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2342 err
= sys_socket(a0
, a1
, a
[2]);
2345 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2348 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2351 err
= sys_listen(a0
, a1
);
2354 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2355 (int __user
*)a
[2], 0);
2357 case SYS_GETSOCKNAME
:
2359 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2360 (int __user
*)a
[2]);
2362 case SYS_GETPEERNAME
:
2364 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2365 (int __user
*)a
[2]);
2367 case SYS_SOCKETPAIR
:
2368 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2371 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2374 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2375 (struct sockaddr __user
*)a
[4], a
[5]);
2378 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2381 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2382 (struct sockaddr __user
*)a
[4],
2383 (int __user
*)a
[5]);
2386 err
= sys_shutdown(a0
, a1
);
2388 case SYS_SETSOCKOPT
:
2389 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2391 case SYS_GETSOCKOPT
:
2393 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2394 (int __user
*)a
[4]);
2397 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2400 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2403 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2406 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2407 (struct timespec __user
*)a
[4]);
2410 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2411 (int __user
*)a
[2], a
[3]);
2420 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2423 * sock_register - add a socket protocol handler
2424 * @ops: description of protocol
2426 * This function is called by a protocol handler that wants to
2427 * advertise its address family, and have it linked into the
2428 * socket interface. The value ops->family corresponds to the
2429 * socket system call protocol family.
2431 int sock_register(const struct net_proto_family
*ops
)
2435 if (ops
->family
>= NPROTO
) {
2436 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2440 spin_lock(&net_family_lock
);
2441 if (rcu_dereference_protected(net_families
[ops
->family
],
2442 lockdep_is_held(&net_family_lock
)))
2445 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2448 spin_unlock(&net_family_lock
);
2450 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2453 EXPORT_SYMBOL(sock_register
);
2456 * sock_unregister - remove a protocol handler
2457 * @family: protocol family to remove
2459 * This function is called by a protocol handler that wants to
2460 * remove its address family, and have it unlinked from the
2461 * new socket creation.
2463 * If protocol handler is a module, then it can use module reference
2464 * counts to protect against new references. If protocol handler is not
2465 * a module then it needs to provide its own protection in
2466 * the ops->create routine.
2468 void sock_unregister(int family
)
2470 BUG_ON(family
< 0 || family
>= NPROTO
);
2472 spin_lock(&net_family_lock
);
2473 RCU_INIT_POINTER(net_families
[family
], NULL
);
2474 spin_unlock(&net_family_lock
);
2478 pr_info("NET: Unregistered protocol family %d\n", family
);
2480 EXPORT_SYMBOL(sock_unregister
);
2482 static int __init
sock_init(void)
2486 * Initialize the network sysctl infrastructure.
2488 err
= net_sysctl_init();
2493 * Initialize skbuff SLAB cache
2498 * Initialize the protocols module.
2503 err
= register_filesystem(&sock_fs_type
);
2506 sock_mnt
= kern_mount(&sock_fs_type
);
2507 if (IS_ERR(sock_mnt
)) {
2508 err
= PTR_ERR(sock_mnt
);
2512 /* The real protocol initialization is performed in later initcalls.
2515 #ifdef CONFIG_NETFILTER
2516 err
= netfilter_init();
2521 ptp_classifier_init();
2527 unregister_filesystem(&sock_fs_type
);
2532 core_initcall(sock_init
); /* early initcall */
2534 #ifdef CONFIG_PROC_FS
2535 void socket_seq_show(struct seq_file
*seq
)
2540 for_each_possible_cpu(cpu
)
2541 counter
+= per_cpu(sockets_in_use
, cpu
);
2543 /* It can be negative, by the way. 8) */
2547 seq_printf(seq
, "sockets: used %d\n", counter
);
2549 #endif /* CONFIG_PROC_FS */
2551 #ifdef CONFIG_COMPAT
2552 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2553 unsigned int cmd
, void __user
*up
)
2555 mm_segment_t old_fs
= get_fs();
2560 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2563 err
= compat_put_timeval(&ktv
, up
);
2568 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2569 unsigned int cmd
, void __user
*up
)
2571 mm_segment_t old_fs
= get_fs();
2572 struct timespec kts
;
2576 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2579 err
= compat_put_timespec(&kts
, up
);
2584 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2586 struct ifreq __user
*uifr
;
2589 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2590 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2593 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2597 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2603 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2605 struct compat_ifconf ifc32
;
2607 struct ifconf __user
*uifc
;
2608 struct compat_ifreq __user
*ifr32
;
2609 struct ifreq __user
*ifr
;
2613 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2616 memset(&ifc
, 0, sizeof(ifc
));
2617 if (ifc32
.ifcbuf
== 0) {
2621 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2623 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2624 sizeof(struct ifreq
);
2625 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2627 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2628 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2629 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2630 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2636 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2639 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2643 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2647 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2649 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2650 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2651 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2657 if (ifc32
.ifcbuf
== 0) {
2658 /* Translate from 64-bit structure multiple to
2662 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2667 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2673 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2675 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2676 bool convert_in
= false, convert_out
= false;
2677 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2678 struct ethtool_rxnfc __user
*rxnfc
;
2679 struct ifreq __user
*ifr
;
2680 u32 rule_cnt
= 0, actual_rule_cnt
;
2685 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2688 compat_rxnfc
= compat_ptr(data
);
2690 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2693 /* Most ethtool structures are defined without padding.
2694 * Unfortunately struct ethtool_rxnfc is an exception.
2699 case ETHTOOL_GRXCLSRLALL
:
2700 /* Buffer size is variable */
2701 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2703 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2705 buf_size
+= rule_cnt
* sizeof(u32
);
2707 case ETHTOOL_GRXRINGS
:
2708 case ETHTOOL_GRXCLSRLCNT
:
2709 case ETHTOOL_GRXCLSRULE
:
2710 case ETHTOOL_SRXCLSRLINS
:
2713 case ETHTOOL_SRXCLSRLDEL
:
2714 buf_size
+= sizeof(struct ethtool_rxnfc
);
2719 ifr
= compat_alloc_user_space(buf_size
);
2720 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2722 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2725 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2726 &ifr
->ifr_ifru
.ifru_data
))
2730 /* We expect there to be holes between fs.m_ext and
2731 * fs.ring_cookie and at the end of fs, but nowhere else.
2733 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2734 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2735 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2736 sizeof(rxnfc
->fs
.m_ext
));
2738 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2739 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2740 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2741 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2743 if (copy_in_user(rxnfc
, compat_rxnfc
,
2744 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2745 (void __user
*)rxnfc
) ||
2746 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2747 &compat_rxnfc
->fs
.ring_cookie
,
2748 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2749 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2750 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2751 sizeof(rxnfc
->rule_cnt
)))
2755 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2760 if (copy_in_user(compat_rxnfc
, rxnfc
,
2761 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2762 (const void __user
*)rxnfc
) ||
2763 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2764 &rxnfc
->fs
.ring_cookie
,
2765 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2766 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2767 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2768 sizeof(rxnfc
->rule_cnt
)))
2771 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2772 /* As an optimisation, we only copy the actual
2773 * number of rules that the underlying
2774 * function returned. Since Mallory might
2775 * change the rule count in user memory, we
2776 * check that it is less than the rule count
2777 * originally given (as the user buffer size),
2778 * which has been range-checked.
2780 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2782 if (actual_rule_cnt
< rule_cnt
)
2783 rule_cnt
= actual_rule_cnt
;
2784 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2785 &rxnfc
->rule_locs
[0],
2786 rule_cnt
* sizeof(u32
)))
2794 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2797 compat_uptr_t uptr32
;
2798 struct ifreq __user
*uifr
;
2800 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2801 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2804 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2807 uptr
= compat_ptr(uptr32
);
2809 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2812 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2815 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2816 struct compat_ifreq __user
*ifr32
)
2819 mm_segment_t old_fs
;
2823 case SIOCBONDENSLAVE
:
2824 case SIOCBONDRELEASE
:
2825 case SIOCBONDSETHWADDR
:
2826 case SIOCBONDCHANGEACTIVE
:
2827 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2832 err
= dev_ioctl(net
, cmd
,
2833 (struct ifreq __user __force
*) &kifr
);
2838 return -ENOIOCTLCMD
;
2842 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2843 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
2844 struct compat_ifreq __user
*u_ifreq32
)
2846 struct ifreq __user
*u_ifreq64
;
2847 char tmp_buf
[IFNAMSIZ
];
2848 void __user
*data64
;
2851 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2854 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2856 data64
= compat_ptr(data32
);
2858 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2860 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2863 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2866 return dev_ioctl(net
, cmd
, u_ifreq64
);
2869 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2870 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2872 struct ifreq __user
*uifr
;
2875 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2876 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2879 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2890 case SIOCGIFBRDADDR
:
2891 case SIOCGIFDSTADDR
:
2892 case SIOCGIFNETMASK
:
2897 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2905 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2906 struct compat_ifreq __user
*uifr32
)
2909 struct compat_ifmap __user
*uifmap32
;
2910 mm_segment_t old_fs
;
2913 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2914 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2915 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2916 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2917 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2918 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2919 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2920 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2926 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
2929 if (cmd
== SIOCGIFMAP
&& !err
) {
2930 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2931 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2932 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2933 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2934 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2935 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2936 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2945 struct sockaddr rt_dst
; /* target address */
2946 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2947 struct sockaddr rt_genmask
; /* target network mask (IP) */
2948 unsigned short rt_flags
;
2951 unsigned char rt_tos
;
2952 unsigned char rt_class
;
2954 short rt_metric
; /* +1 for binary compatibility! */
2955 /* char * */ u32 rt_dev
; /* forcing the device at add */
2956 u32 rt_mtu
; /* per route MTU/Window */
2957 u32 rt_window
; /* Window clamping */
2958 unsigned short rt_irtt
; /* Initial RTT */
2961 struct in6_rtmsg32
{
2962 struct in6_addr rtmsg_dst
;
2963 struct in6_addr rtmsg_src
;
2964 struct in6_addr rtmsg_gateway
;
2974 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
2975 unsigned int cmd
, void __user
*argp
)
2979 struct in6_rtmsg r6
;
2983 mm_segment_t old_fs
= get_fs();
2985 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
2986 struct in6_rtmsg32 __user
*ur6
= argp
;
2987 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
2988 3 * sizeof(struct in6_addr
));
2989 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
2990 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
2991 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
2992 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
2993 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
2994 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
2995 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
2999 struct rtentry32 __user
*ur4
= argp
;
3000 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3001 3 * sizeof(struct sockaddr
));
3002 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3003 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3004 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3005 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3006 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3007 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3009 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3010 r4
.rt_dev
= (char __user __force
*)devname
;
3024 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3031 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3032 * for some operations; this forces use of the newer bridge-utils that
3033 * use compatible ioctls
3035 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3039 if (get_user(tmp
, argp
))
3041 if (tmp
== BRCTL_GET_VERSION
)
3042 return BRCTL_VERSION
+ 1;
3046 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3047 unsigned int cmd
, unsigned long arg
)
3049 void __user
*argp
= compat_ptr(arg
);
3050 struct sock
*sk
= sock
->sk
;
3051 struct net
*net
= sock_net(sk
);
3053 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3054 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3059 return old_bridge_ioctl(argp
);
3061 return dev_ifname32(net
, argp
);
3063 return dev_ifconf(net
, argp
);
3065 return ethtool_ioctl(net
, argp
);
3067 return compat_siocwandev(net
, argp
);
3070 return compat_sioc_ifmap(net
, cmd
, argp
);
3071 case SIOCBONDENSLAVE
:
3072 case SIOCBONDRELEASE
:
3073 case SIOCBONDSETHWADDR
:
3074 case SIOCBONDCHANGEACTIVE
:
3075 return bond_ioctl(net
, cmd
, argp
);
3078 return routing_ioctl(net
, sock
, cmd
, argp
);
3080 return do_siocgstamp(net
, sock
, cmd
, argp
);
3082 return do_siocgstampns(net
, sock
, cmd
, argp
);
3083 case SIOCBONDSLAVEINFOQUERY
:
3084 case SIOCBONDINFOQUERY
:
3087 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3099 return sock_ioctl(file
, cmd
, arg
);
3116 case SIOCSIFHWBROADCAST
:
3118 case SIOCGIFBRDADDR
:
3119 case SIOCSIFBRDADDR
:
3120 case SIOCGIFDSTADDR
:
3121 case SIOCSIFDSTADDR
:
3122 case SIOCGIFNETMASK
:
3123 case SIOCSIFNETMASK
:
3134 return dev_ifsioc(net
, sock
, cmd
, argp
);
3140 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3143 return -ENOIOCTLCMD
;
3146 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3149 struct socket
*sock
= file
->private_data
;
3150 int ret
= -ENOIOCTLCMD
;
3157 if (sock
->ops
->compat_ioctl
)
3158 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3160 if (ret
== -ENOIOCTLCMD
&&
3161 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3162 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3164 if (ret
== -ENOIOCTLCMD
)
3165 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3171 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3173 return sock
->ops
->bind(sock
, addr
, addrlen
);
3175 EXPORT_SYMBOL(kernel_bind
);
3177 int kernel_listen(struct socket
*sock
, int backlog
)
3179 return sock
->ops
->listen(sock
, backlog
);
3181 EXPORT_SYMBOL(kernel_listen
);
3183 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3185 struct sock
*sk
= sock
->sk
;
3188 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3193 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3195 sock_release(*newsock
);
3200 (*newsock
)->ops
= sock
->ops
;
3201 __module_get((*newsock
)->ops
->owner
);
3206 EXPORT_SYMBOL(kernel_accept
);
3208 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3211 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3213 EXPORT_SYMBOL(kernel_connect
);
3215 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3218 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3220 EXPORT_SYMBOL(kernel_getsockname
);
3222 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3225 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3227 EXPORT_SYMBOL(kernel_getpeername
);
3229 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3230 char *optval
, int *optlen
)
3232 mm_segment_t oldfs
= get_fs();
3233 char __user
*uoptval
;
3234 int __user
*uoptlen
;
3237 uoptval
= (char __user __force
*) optval
;
3238 uoptlen
= (int __user __force
*) optlen
;
3241 if (level
== SOL_SOCKET
)
3242 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3244 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3249 EXPORT_SYMBOL(kernel_getsockopt
);
3251 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3252 char *optval
, unsigned int optlen
)
3254 mm_segment_t oldfs
= get_fs();
3255 char __user
*uoptval
;
3258 uoptval
= (char __user __force
*) optval
;
3261 if (level
== SOL_SOCKET
)
3262 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3264 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3269 EXPORT_SYMBOL(kernel_setsockopt
);
3271 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3272 size_t size
, int flags
)
3274 if (sock
->ops
->sendpage
)
3275 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3277 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3279 EXPORT_SYMBOL(kernel_sendpage
);
3281 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3283 mm_segment_t oldfs
= get_fs();
3287 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3292 EXPORT_SYMBOL(kernel_sock_ioctl
);
3294 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3296 return sock
->ops
->shutdown(sock
, how
);
3298 EXPORT_SYMBOL(kernel_sock_shutdown
);