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 <linux/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 int sockfs_xattr_get(const struct xattr_handler
*handler
,
324 struct dentry
*dentry
, struct inode
*inode
,
325 const char *suffix
, void *value
, size_t size
)
328 if (dentry
->d_name
.len
+ 1 > size
)
330 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
332 return dentry
->d_name
.len
+ 1;
335 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
336 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
337 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
339 static const struct xattr_handler sockfs_xattr_handler
= {
340 .name
= XATTR_NAME_SOCKPROTONAME
,
341 .get
= sockfs_xattr_get
,
344 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
345 struct dentry
*dentry
, struct inode
*inode
,
346 const char *suffix
, const void *value
,
347 size_t size
, int flags
)
349 /* Handled by LSM. */
353 static const struct xattr_handler sockfs_security_xattr_handler
= {
354 .prefix
= XATTR_SECURITY_PREFIX
,
355 .set
= sockfs_security_xattr_set
,
358 static const struct xattr_handler
*sockfs_xattr_handlers
[] = {
359 &sockfs_xattr_handler
,
360 &sockfs_security_xattr_handler
,
364 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
365 int flags
, const char *dev_name
, void *data
)
367 return mount_pseudo_xattr(fs_type
, "socket:", &sockfs_ops
,
368 sockfs_xattr_handlers
,
369 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
372 static struct vfsmount
*sock_mnt __read_mostly
;
374 static struct file_system_type sock_fs_type
= {
376 .mount
= sockfs_mount
,
377 .kill_sb
= kill_anon_super
,
381 * Obtains the first available file descriptor and sets it up for use.
383 * These functions create file structures and maps them to fd space
384 * of the current process. On success it returns file descriptor
385 * and file struct implicitly stored in sock->file.
386 * Note that another thread may close file descriptor before we return
387 * from this function. We use the fact that now we do not refer
388 * to socket after mapping. If one day we will need it, this
389 * function will increment ref. count on file by 1.
391 * In any case returned fd MAY BE not valid!
392 * This race condition is unavoidable
393 * with shared fd spaces, we cannot solve it inside kernel,
394 * but we take care of internal coherence yet.
397 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
399 struct qstr name
= { .name
= "" };
405 name
.len
= strlen(name
.name
);
406 } else if (sock
->sk
) {
407 name
.name
= sock
->sk
->sk_prot_creator
->name
;
408 name
.len
= strlen(name
.name
);
410 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
411 if (unlikely(!path
.dentry
))
412 return ERR_PTR(-ENOMEM
);
413 path
.mnt
= mntget(sock_mnt
);
415 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
417 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
420 /* drop dentry, keep inode */
421 ihold(d_inode(path
.dentry
));
427 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
428 file
->private_data
= sock
;
431 EXPORT_SYMBOL(sock_alloc_file
);
433 static int sock_map_fd(struct socket
*sock
, int flags
)
435 struct file
*newfile
;
436 int fd
= get_unused_fd_flags(flags
);
437 if (unlikely(fd
< 0))
440 newfile
= sock_alloc_file(sock
, flags
, NULL
);
441 if (likely(!IS_ERR(newfile
))) {
442 fd_install(fd
, newfile
);
447 return PTR_ERR(newfile
);
450 struct socket
*sock_from_file(struct file
*file
, int *err
)
452 if (file
->f_op
== &socket_file_ops
)
453 return file
->private_data
; /* set in sock_map_fd */
458 EXPORT_SYMBOL(sock_from_file
);
461 * sockfd_lookup - Go from a file number to its socket slot
463 * @err: pointer to an error code return
465 * The file handle passed in is locked and the socket it is bound
466 * too is returned. If an error occurs the err pointer is overwritten
467 * with a negative errno code and NULL is returned. The function checks
468 * for both invalid handles and passing a handle which is not a socket.
470 * On a success the socket object pointer is returned.
473 struct socket
*sockfd_lookup(int fd
, int *err
)
484 sock
= sock_from_file(file
, err
);
489 EXPORT_SYMBOL(sockfd_lookup
);
491 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
493 struct fd f
= fdget(fd
);
498 sock
= sock_from_file(f
.file
, err
);
500 *fput_needed
= f
.flags
;
508 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
514 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
524 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
529 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
536 static int sockfs_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
538 int err
= simple_setattr(dentry
, iattr
);
540 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
541 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
543 sock
->sk
->sk_uid
= iattr
->ia_uid
;
549 static const struct inode_operations sockfs_inode_ops
= {
550 .listxattr
= sockfs_listxattr
,
551 .setattr
= sockfs_setattr
,
555 * sock_alloc - allocate a socket
557 * Allocate a new inode and socket object. The two are bound together
558 * and initialised. The socket is then returned. If we are out of inodes
562 struct socket
*sock_alloc(void)
567 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
571 sock
= SOCKET_I(inode
);
573 kmemcheck_annotate_bitfield(sock
, type
);
574 inode
->i_ino
= get_next_ino();
575 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
576 inode
->i_uid
= current_fsuid();
577 inode
->i_gid
= current_fsgid();
578 inode
->i_op
= &sockfs_inode_ops
;
580 this_cpu_add(sockets_in_use
, 1);
583 EXPORT_SYMBOL(sock_alloc
);
586 * sock_release - close a socket
587 * @sock: socket to close
589 * The socket is released from the protocol stack if it has a release
590 * callback, and the inode is then released if the socket is bound to
591 * an inode not a file.
594 void sock_release(struct socket
*sock
)
597 struct module
*owner
= sock
->ops
->owner
;
599 sock
->ops
->release(sock
);
604 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
605 pr_err("%s: fasync list not empty!\n", __func__
);
607 this_cpu_sub(sockets_in_use
, 1);
609 iput(SOCK_INODE(sock
));
614 EXPORT_SYMBOL(sock_release
);
616 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
618 u8 flags
= *tx_flags
;
620 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
621 flags
|= SKBTX_HW_TSTAMP
;
623 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
624 flags
|= SKBTX_SW_TSTAMP
;
626 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
627 flags
|= SKBTX_SCHED_TSTAMP
;
631 EXPORT_SYMBOL(__sock_tx_timestamp
);
633 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
635 int ret
= sock
->ops
->sendmsg(sock
, msg
, msg_data_left(msg
));
636 BUG_ON(ret
== -EIOCBQUEUED
);
640 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
642 int err
= security_socket_sendmsg(sock
, msg
,
645 return err
?: sock_sendmsg_nosec(sock
, msg
);
647 EXPORT_SYMBOL(sock_sendmsg
);
649 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
650 struct kvec
*vec
, size_t num
, size_t size
)
652 iov_iter_kvec(&msg
->msg_iter
, WRITE
| ITER_KVEC
, vec
, num
, size
);
653 return sock_sendmsg(sock
, msg
);
655 EXPORT_SYMBOL(kernel_sendmsg
);
657 static bool skb_is_err_queue(const struct sk_buff
*skb
)
659 /* pkt_type of skbs enqueued on the error queue are set to
660 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
661 * in recvmsg, since skbs received on a local socket will never
662 * have a pkt_type of PACKET_OUTGOING.
664 return skb
->pkt_type
== PACKET_OUTGOING
;
668 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
670 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
673 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
674 struct scm_timestamping tss
;
676 struct skb_shared_hwtstamps
*shhwtstamps
=
679 /* Race occurred between timestamp enabling and packet
680 receiving. Fill in the current time for now. */
681 if (need_software_tstamp
&& skb
->tstamp
== 0)
682 __net_timestamp(skb
);
684 if (need_software_tstamp
) {
685 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
687 skb_get_timestamp(skb
, &tv
);
688 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
692 skb_get_timestampns(skb
, &ts
);
693 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
698 memset(&tss
, 0, sizeof(tss
));
699 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
700 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
703 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
704 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2))
707 put_cmsg(msg
, SOL_SOCKET
,
708 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
710 if (skb_is_err_queue(skb
) && skb
->len
&&
711 SKB_EXT_ERR(skb
)->opt_stats
)
712 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
713 skb
->len
, skb
->data
);
716 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
718 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
723 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
725 if (!skb
->wifi_acked_valid
)
728 ack
= skb
->wifi_acked
;
730 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
732 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
734 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
737 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
738 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
739 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
742 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
745 sock_recv_timestamp(msg
, sk
, skb
);
746 sock_recv_drops(msg
, sk
, skb
);
748 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
750 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
753 return sock
->ops
->recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
756 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
758 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
760 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
762 EXPORT_SYMBOL(sock_recvmsg
);
765 * kernel_recvmsg - Receive a message from a socket (kernel space)
766 * @sock: The socket to receive the message from
767 * @msg: Received message
768 * @vec: Input s/g array for message data
769 * @num: Size of input s/g array
770 * @size: Number of bytes to read
771 * @flags: Message flags (MSG_DONTWAIT, etc...)
773 * On return the msg structure contains the scatter/gather array passed in the
774 * vec argument. The array is modified so that it consists of the unfilled
775 * portion of the original array.
777 * The returned value is the total number of bytes received, or an error.
779 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
780 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
782 mm_segment_t oldfs
= get_fs();
785 iov_iter_kvec(&msg
->msg_iter
, READ
| ITER_KVEC
, vec
, num
, size
);
787 result
= sock_recvmsg(sock
, msg
, flags
);
791 EXPORT_SYMBOL(kernel_recvmsg
);
793 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
794 int offset
, size_t size
, loff_t
*ppos
, int more
)
799 sock
= file
->private_data
;
801 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
802 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
805 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
808 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
809 struct pipe_inode_info
*pipe
, size_t len
,
812 struct socket
*sock
= file
->private_data
;
814 if (unlikely(!sock
->ops
->splice_read
))
817 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
820 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
822 struct file
*file
= iocb
->ki_filp
;
823 struct socket
*sock
= file
->private_data
;
824 struct msghdr msg
= {.msg_iter
= *to
,
828 if (file
->f_flags
& O_NONBLOCK
)
829 msg
.msg_flags
= MSG_DONTWAIT
;
831 if (iocb
->ki_pos
!= 0)
834 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
837 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
842 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
844 struct file
*file
= iocb
->ki_filp
;
845 struct socket
*sock
= file
->private_data
;
846 struct msghdr msg
= {.msg_iter
= *from
,
850 if (iocb
->ki_pos
!= 0)
853 if (file
->f_flags
& O_NONBLOCK
)
854 msg
.msg_flags
= MSG_DONTWAIT
;
856 if (sock
->type
== SOCK_SEQPACKET
)
857 msg
.msg_flags
|= MSG_EOR
;
859 res
= sock_sendmsg(sock
, &msg
);
860 *from
= msg
.msg_iter
;
865 * Atomic setting of ioctl hooks to avoid race
866 * with module unload.
869 static DEFINE_MUTEX(br_ioctl_mutex
);
870 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
872 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
874 mutex_lock(&br_ioctl_mutex
);
875 br_ioctl_hook
= hook
;
876 mutex_unlock(&br_ioctl_mutex
);
878 EXPORT_SYMBOL(brioctl_set
);
880 static DEFINE_MUTEX(vlan_ioctl_mutex
);
881 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
883 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
885 mutex_lock(&vlan_ioctl_mutex
);
886 vlan_ioctl_hook
= hook
;
887 mutex_unlock(&vlan_ioctl_mutex
);
889 EXPORT_SYMBOL(vlan_ioctl_set
);
891 static DEFINE_MUTEX(dlci_ioctl_mutex
);
892 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
894 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
896 mutex_lock(&dlci_ioctl_mutex
);
897 dlci_ioctl_hook
= hook
;
898 mutex_unlock(&dlci_ioctl_mutex
);
900 EXPORT_SYMBOL(dlci_ioctl_set
);
902 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
903 unsigned int cmd
, unsigned long arg
)
906 void __user
*argp
= (void __user
*)arg
;
908 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
911 * If this ioctl is unknown try to hand it down
914 if (err
== -ENOIOCTLCMD
)
915 err
= dev_ioctl(net
, cmd
, argp
);
921 * With an ioctl, arg may well be a user mode pointer, but we don't know
922 * what to do with it - that's up to the protocol still.
925 static struct ns_common
*get_net_ns(struct ns_common
*ns
)
927 return &get_net(container_of(ns
, struct net
, ns
))->ns
;
930 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
934 void __user
*argp
= (void __user
*)arg
;
938 sock
= file
->private_data
;
941 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
942 err
= dev_ioctl(net
, cmd
, argp
);
944 #ifdef CONFIG_WEXT_CORE
945 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
946 err
= dev_ioctl(net
, cmd
, argp
);
953 if (get_user(pid
, (int __user
*)argp
))
955 f_setown(sock
->file
, pid
, 1);
960 err
= put_user(f_getown(sock
->file
),
969 request_module("bridge");
971 mutex_lock(&br_ioctl_mutex
);
973 err
= br_ioctl_hook(net
, cmd
, argp
);
974 mutex_unlock(&br_ioctl_mutex
);
979 if (!vlan_ioctl_hook
)
980 request_module("8021q");
982 mutex_lock(&vlan_ioctl_mutex
);
984 err
= vlan_ioctl_hook(net
, argp
);
985 mutex_unlock(&vlan_ioctl_mutex
);
990 if (!dlci_ioctl_hook
)
991 request_module("dlci");
993 mutex_lock(&dlci_ioctl_mutex
);
995 err
= dlci_ioctl_hook(cmd
, argp
);
996 mutex_unlock(&dlci_ioctl_mutex
);
1000 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1003 err
= open_related_ns(&net
->ns
, get_net_ns
);
1006 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1012 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1015 struct socket
*sock
= NULL
;
1017 err
= security_socket_create(family
, type
, protocol
, 1);
1021 sock
= sock_alloc();
1028 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1040 EXPORT_SYMBOL(sock_create_lite
);
1042 /* No kernel lock held - perfect */
1043 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1045 unsigned int busy_flag
= 0;
1046 struct socket
*sock
;
1049 * We can't return errors to poll, so it's either yes or no.
1051 sock
= file
->private_data
;
1053 if (sk_can_busy_loop(sock
->sk
)) {
1054 /* this socket can poll_ll so tell the system call */
1055 busy_flag
= POLL_BUSY_LOOP
;
1057 /* once, only if requested by syscall */
1058 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1059 sk_busy_loop(sock
->sk
, 1);
1062 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1065 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1067 struct socket
*sock
= file
->private_data
;
1069 return sock
->ops
->mmap(file
, sock
, vma
);
1072 static int sock_close(struct inode
*inode
, struct file
*filp
)
1074 sock_release(SOCKET_I(inode
));
1079 * Update the socket async list
1081 * Fasync_list locking strategy.
1083 * 1. fasync_list is modified only under process context socket lock
1084 * i.e. under semaphore.
1085 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1086 * or under socket lock
1089 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1091 struct socket
*sock
= filp
->private_data
;
1092 struct sock
*sk
= sock
->sk
;
1093 struct socket_wq
*wq
;
1099 wq
= rcu_dereference_protected(sock
->wq
, lockdep_sock_is_held(sk
));
1100 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1102 if (!wq
->fasync_list
)
1103 sock_reset_flag(sk
, SOCK_FASYNC
);
1105 sock_set_flag(sk
, SOCK_FASYNC
);
1111 /* This function may be called only under rcu_lock */
1113 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1115 if (!wq
|| !wq
->fasync_list
)
1119 case SOCK_WAKE_WAITD
:
1120 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1123 case SOCK_WAKE_SPACE
:
1124 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1129 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1132 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1137 EXPORT_SYMBOL(sock_wake_async
);
1139 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1140 struct socket
**res
, int kern
)
1143 struct socket
*sock
;
1144 const struct net_proto_family
*pf
;
1147 * Check protocol is in range
1149 if (family
< 0 || family
>= NPROTO
)
1150 return -EAFNOSUPPORT
;
1151 if (type
< 0 || type
>= SOCK_MAX
)
1156 This uglymoron is moved from INET layer to here to avoid
1157 deadlock in module load.
1159 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1160 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1165 err
= security_socket_create(family
, type
, protocol
, kern
);
1170 * Allocate the socket and allow the family to set things up. if
1171 * the protocol is 0, the family is instructed to select an appropriate
1174 sock
= sock_alloc();
1176 net_warn_ratelimited("socket: no more sockets\n");
1177 return -ENFILE
; /* Not exactly a match, but its the
1178 closest posix thing */
1183 #ifdef CONFIG_MODULES
1184 /* Attempt to load a protocol module if the find failed.
1186 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1187 * requested real, full-featured networking support upon configuration.
1188 * Otherwise module support will break!
1190 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1191 request_module("net-pf-%d", family
);
1195 pf
= rcu_dereference(net_families
[family
]);
1196 err
= -EAFNOSUPPORT
;
1201 * We will call the ->create function, that possibly is in a loadable
1202 * module, so we have to bump that loadable module refcnt first.
1204 if (!try_module_get(pf
->owner
))
1207 /* Now protected by module ref count */
1210 err
= pf
->create(net
, sock
, protocol
, kern
);
1212 goto out_module_put
;
1215 * Now to bump the refcnt of the [loadable] module that owns this
1216 * socket at sock_release time we decrement its refcnt.
1218 if (!try_module_get(sock
->ops
->owner
))
1219 goto out_module_busy
;
1222 * Now that we're done with the ->create function, the [loadable]
1223 * module can have its refcnt decremented
1225 module_put(pf
->owner
);
1226 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1228 goto out_sock_release
;
1234 err
= -EAFNOSUPPORT
;
1237 module_put(pf
->owner
);
1244 goto out_sock_release
;
1246 EXPORT_SYMBOL(__sock_create
);
1248 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1250 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1252 EXPORT_SYMBOL(sock_create
);
1254 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1256 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1258 EXPORT_SYMBOL(sock_create_kern
);
1260 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1263 struct socket
*sock
;
1266 /* Check the SOCK_* constants for consistency. */
1267 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1268 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1269 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1270 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1272 flags
= type
& ~SOCK_TYPE_MASK
;
1273 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1275 type
&= SOCK_TYPE_MASK
;
1277 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1278 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1280 retval
= sock_create(family
, type
, protocol
, &sock
);
1284 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1289 /* It may be already another descriptor 8) Not kernel problem. */
1298 * Create a pair of connected sockets.
1301 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1302 int __user
*, usockvec
)
1304 struct socket
*sock1
, *sock2
;
1306 struct file
*newfile1
, *newfile2
;
1309 flags
= type
& ~SOCK_TYPE_MASK
;
1310 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1312 type
&= SOCK_TYPE_MASK
;
1314 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1315 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1318 * Obtain the first socket and check if the underlying protocol
1319 * supports the socketpair call.
1322 err
= sock_create(family
, type
, protocol
, &sock1
);
1326 err
= sock_create(family
, type
, protocol
, &sock2
);
1330 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1332 goto out_release_both
;
1334 fd1
= get_unused_fd_flags(flags
);
1335 if (unlikely(fd1
< 0)) {
1337 goto out_release_both
;
1340 fd2
= get_unused_fd_flags(flags
);
1341 if (unlikely(fd2
< 0)) {
1343 goto out_put_unused_1
;
1346 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1347 if (IS_ERR(newfile1
)) {
1348 err
= PTR_ERR(newfile1
);
1349 goto out_put_unused_both
;
1352 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1353 if (IS_ERR(newfile2
)) {
1354 err
= PTR_ERR(newfile2
);
1358 err
= put_user(fd1
, &usockvec
[0]);
1362 err
= put_user(fd2
, &usockvec
[1]);
1366 audit_fd_pair(fd1
, fd2
);
1368 fd_install(fd1
, newfile1
);
1369 fd_install(fd2
, newfile2
);
1370 /* fd1 and fd2 may be already another descriptors.
1371 * Not kernel problem.
1387 sock_release(sock2
);
1390 out_put_unused_both
:
1395 sock_release(sock2
);
1397 sock_release(sock1
);
1403 * Bind a name to a socket. Nothing much to do here since it's
1404 * the protocol's responsibility to handle the local address.
1406 * We move the socket address to kernel space before we call
1407 * the protocol layer (having also checked the address is ok).
1410 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1412 struct socket
*sock
;
1413 struct sockaddr_storage address
;
1414 int err
, fput_needed
;
1416 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1418 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1420 err
= security_socket_bind(sock
,
1421 (struct sockaddr
*)&address
,
1424 err
= sock
->ops
->bind(sock
,
1428 fput_light(sock
->file
, fput_needed
);
1434 * Perform a listen. Basically, we allow the protocol to do anything
1435 * necessary for a listen, and if that works, we mark the socket as
1436 * ready for listening.
1439 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1441 struct socket
*sock
;
1442 int err
, fput_needed
;
1445 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1447 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1448 if ((unsigned int)backlog
> somaxconn
)
1449 backlog
= somaxconn
;
1451 err
= security_socket_listen(sock
, backlog
);
1453 err
= sock
->ops
->listen(sock
, backlog
);
1455 fput_light(sock
->file
, fput_needed
);
1461 * For accept, we attempt to create a new socket, set up the link
1462 * with the client, wake up the client, then return the new
1463 * connected fd. We collect the address of the connector in kernel
1464 * space and move it to user at the very end. This is unclean because
1465 * we open the socket then return an error.
1467 * 1003.1g adds the ability to recvmsg() to query connection pending
1468 * status to recvmsg. We need to add that support in a way thats
1469 * clean when we restucture accept also.
1472 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1473 int __user
*, upeer_addrlen
, int, flags
)
1475 struct socket
*sock
, *newsock
;
1476 struct file
*newfile
;
1477 int err
, len
, newfd
, fput_needed
;
1478 struct sockaddr_storage address
;
1480 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1483 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1484 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1486 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1491 newsock
= sock_alloc();
1495 newsock
->type
= sock
->type
;
1496 newsock
->ops
= sock
->ops
;
1499 * We don't need try_module_get here, as the listening socket (sock)
1500 * has the protocol module (sock->ops->owner) held.
1502 __module_get(newsock
->ops
->owner
);
1504 newfd
= get_unused_fd_flags(flags
);
1505 if (unlikely(newfd
< 0)) {
1507 sock_release(newsock
);
1510 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1511 if (IS_ERR(newfile
)) {
1512 err
= PTR_ERR(newfile
);
1513 put_unused_fd(newfd
);
1514 sock_release(newsock
);
1518 err
= security_socket_accept(sock
, newsock
);
1522 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
);
1526 if (upeer_sockaddr
) {
1527 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1529 err
= -ECONNABORTED
;
1532 err
= move_addr_to_user(&address
,
1533 len
, upeer_sockaddr
, upeer_addrlen
);
1538 /* File flags are not inherited via accept() unlike another OSes. */
1540 fd_install(newfd
, newfile
);
1544 fput_light(sock
->file
, fput_needed
);
1549 put_unused_fd(newfd
);
1553 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1554 int __user
*, upeer_addrlen
)
1556 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1560 * Attempt to connect to a socket with the server address. The address
1561 * is in user space so we verify it is OK and move it to kernel space.
1563 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1566 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1567 * other SEQPACKET protocols that take time to connect() as it doesn't
1568 * include the -EINPROGRESS status for such sockets.
1571 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1574 struct socket
*sock
;
1575 struct sockaddr_storage address
;
1576 int err
, fput_needed
;
1578 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1581 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1586 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1590 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1591 sock
->file
->f_flags
);
1593 fput_light(sock
->file
, fput_needed
);
1599 * Get the local address ('name') of a socket object. Move the obtained
1600 * name to user space.
1603 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1604 int __user
*, usockaddr_len
)
1606 struct socket
*sock
;
1607 struct sockaddr_storage address
;
1608 int len
, err
, fput_needed
;
1610 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1614 err
= security_socket_getsockname(sock
);
1618 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1621 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1624 fput_light(sock
->file
, fput_needed
);
1630 * Get the remote address ('name') of a socket object. Move the obtained
1631 * name to user space.
1634 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1635 int __user
*, usockaddr_len
)
1637 struct socket
*sock
;
1638 struct sockaddr_storage address
;
1639 int len
, err
, fput_needed
;
1641 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1643 err
= security_socket_getpeername(sock
);
1645 fput_light(sock
->file
, fput_needed
);
1650 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1653 err
= move_addr_to_user(&address
, len
, usockaddr
,
1655 fput_light(sock
->file
, fput_needed
);
1661 * Send a datagram to a given address. We move the address into kernel
1662 * space and check the user space data area is readable before invoking
1666 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1667 unsigned int, flags
, struct sockaddr __user
*, addr
,
1670 struct socket
*sock
;
1671 struct sockaddr_storage address
;
1677 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
1680 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1684 msg
.msg_name
= NULL
;
1685 msg
.msg_control
= NULL
;
1686 msg
.msg_controllen
= 0;
1687 msg
.msg_namelen
= 0;
1689 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1692 msg
.msg_name
= (struct sockaddr
*)&address
;
1693 msg
.msg_namelen
= addr_len
;
1695 if (sock
->file
->f_flags
& O_NONBLOCK
)
1696 flags
|= MSG_DONTWAIT
;
1697 msg
.msg_flags
= flags
;
1698 err
= sock_sendmsg(sock
, &msg
);
1701 fput_light(sock
->file
, fput_needed
);
1707 * Send a datagram down a socket.
1710 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1711 unsigned int, flags
)
1713 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1717 * Receive a frame from the socket and optionally record the address of the
1718 * sender. We verify the buffers are writable and if needed move the
1719 * sender address from kernel to user space.
1722 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1723 unsigned int, flags
, struct sockaddr __user
*, addr
,
1724 int __user
*, addr_len
)
1726 struct socket
*sock
;
1729 struct sockaddr_storage address
;
1733 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
1736 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1740 msg
.msg_control
= NULL
;
1741 msg
.msg_controllen
= 0;
1742 /* Save some cycles and don't copy the address if not needed */
1743 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1744 /* We assume all kernel code knows the size of sockaddr_storage */
1745 msg
.msg_namelen
= 0;
1746 msg
.msg_iocb
= NULL
;
1747 if (sock
->file
->f_flags
& O_NONBLOCK
)
1748 flags
|= MSG_DONTWAIT
;
1749 err
= sock_recvmsg(sock
, &msg
, flags
);
1751 if (err
>= 0 && addr
!= NULL
) {
1752 err2
= move_addr_to_user(&address
,
1753 msg
.msg_namelen
, addr
, addr_len
);
1758 fput_light(sock
->file
, fput_needed
);
1764 * Receive a datagram from a socket.
1767 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1768 unsigned int, flags
)
1770 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1774 * Set a socket option. Because we don't know the option lengths we have
1775 * to pass the user mode parameter for the protocols to sort out.
1778 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1779 char __user
*, optval
, int, optlen
)
1781 int err
, fput_needed
;
1782 struct socket
*sock
;
1787 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1789 err
= security_socket_setsockopt(sock
, level
, optname
);
1793 if (level
== SOL_SOCKET
)
1795 sock_setsockopt(sock
, level
, optname
, optval
,
1799 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1802 fput_light(sock
->file
, fput_needed
);
1808 * Get a socket option. Because we don't know the option lengths we have
1809 * to pass a user mode parameter for the protocols to sort out.
1812 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1813 char __user
*, optval
, int __user
*, optlen
)
1815 int err
, fput_needed
;
1816 struct socket
*sock
;
1818 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1820 err
= security_socket_getsockopt(sock
, level
, optname
);
1824 if (level
== SOL_SOCKET
)
1826 sock_getsockopt(sock
, level
, optname
, optval
,
1830 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1833 fput_light(sock
->file
, fput_needed
);
1839 * Shutdown a socket.
1842 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1844 int err
, fput_needed
;
1845 struct socket
*sock
;
1847 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1849 err
= security_socket_shutdown(sock
, how
);
1851 err
= sock
->ops
->shutdown(sock
, how
);
1852 fput_light(sock
->file
, fput_needed
);
1857 /* A couple of helpful macros for getting the address of the 32/64 bit
1858 * fields which are the same type (int / unsigned) on our platforms.
1860 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1861 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1862 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1864 struct used_address
{
1865 struct sockaddr_storage name
;
1866 unsigned int name_len
;
1869 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
1870 struct user_msghdr __user
*umsg
,
1871 struct sockaddr __user
**save_addr
,
1874 struct sockaddr __user
*uaddr
;
1875 struct iovec __user
*uiov
;
1879 if (!access_ok(VERIFY_READ
, umsg
, sizeof(*umsg
)) ||
1880 __get_user(uaddr
, &umsg
->msg_name
) ||
1881 __get_user(kmsg
->msg_namelen
, &umsg
->msg_namelen
) ||
1882 __get_user(uiov
, &umsg
->msg_iov
) ||
1883 __get_user(nr_segs
, &umsg
->msg_iovlen
) ||
1884 __get_user(kmsg
->msg_control
, &umsg
->msg_control
) ||
1885 __get_user(kmsg
->msg_controllen
, &umsg
->msg_controllen
) ||
1886 __get_user(kmsg
->msg_flags
, &umsg
->msg_flags
))
1890 kmsg
->msg_namelen
= 0;
1892 if (kmsg
->msg_namelen
< 0)
1895 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1896 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1901 if (uaddr
&& kmsg
->msg_namelen
) {
1903 err
= move_addr_to_kernel(uaddr
, kmsg
->msg_namelen
,
1909 kmsg
->msg_name
= NULL
;
1910 kmsg
->msg_namelen
= 0;
1913 if (nr_segs
> UIO_MAXIOV
)
1916 kmsg
->msg_iocb
= NULL
;
1918 return import_iovec(save_addr
? READ
: WRITE
, uiov
, nr_segs
,
1919 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
1922 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
1923 struct msghdr
*msg_sys
, unsigned int flags
,
1924 struct used_address
*used_address
,
1925 unsigned int allowed_msghdr_flags
)
1927 struct compat_msghdr __user
*msg_compat
=
1928 (struct compat_msghdr __user
*)msg
;
1929 struct sockaddr_storage address
;
1930 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1931 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1932 __aligned(sizeof(__kernel_size_t
));
1933 /* 20 is size of ipv6_pktinfo */
1934 unsigned char *ctl_buf
= ctl
;
1938 msg_sys
->msg_name
= &address
;
1940 if (MSG_CMSG_COMPAT
& flags
)
1941 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
1943 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
1949 if (msg_sys
->msg_controllen
> INT_MAX
)
1951 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
1952 ctl_len
= msg_sys
->msg_controllen
;
1953 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1955 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
1959 ctl_buf
= msg_sys
->msg_control
;
1960 ctl_len
= msg_sys
->msg_controllen
;
1961 } else if (ctl_len
) {
1962 if (ctl_len
> sizeof(ctl
)) {
1963 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1964 if (ctl_buf
== NULL
)
1969 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1970 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1971 * checking falls down on this.
1973 if (copy_from_user(ctl_buf
,
1974 (void __user __force
*)msg_sys
->msg_control
,
1977 msg_sys
->msg_control
= ctl_buf
;
1979 msg_sys
->msg_flags
= flags
;
1981 if (sock
->file
->f_flags
& O_NONBLOCK
)
1982 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
1984 * If this is sendmmsg() and current destination address is same as
1985 * previously succeeded address, omit asking LSM's decision.
1986 * used_address->name_len is initialized to UINT_MAX so that the first
1987 * destination address never matches.
1989 if (used_address
&& msg_sys
->msg_name
&&
1990 used_address
->name_len
== msg_sys
->msg_namelen
&&
1991 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
1992 used_address
->name_len
)) {
1993 err
= sock_sendmsg_nosec(sock
, msg_sys
);
1996 err
= sock_sendmsg(sock
, msg_sys
);
1998 * If this is sendmmsg() and sending to current destination address was
1999 * successful, remember it.
2001 if (used_address
&& err
>= 0) {
2002 used_address
->name_len
= msg_sys
->msg_namelen
;
2003 if (msg_sys
->msg_name
)
2004 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2005 used_address
->name_len
);
2010 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2017 * BSD sendmsg interface
2020 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2022 int fput_needed
, err
;
2023 struct msghdr msg_sys
;
2024 struct socket
*sock
;
2026 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2030 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2032 fput_light(sock
->file
, fput_needed
);
2037 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2039 if (flags
& MSG_CMSG_COMPAT
)
2041 return __sys_sendmsg(fd
, msg
, flags
);
2045 * Linux sendmmsg interface
2048 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2051 int fput_needed
, err
, datagrams
;
2052 struct socket
*sock
;
2053 struct mmsghdr __user
*entry
;
2054 struct compat_mmsghdr __user
*compat_entry
;
2055 struct msghdr msg_sys
;
2056 struct used_address used_address
;
2057 unsigned int oflags
= flags
;
2059 if (vlen
> UIO_MAXIOV
)
2064 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2068 used_address
.name_len
= UINT_MAX
;
2070 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2074 while (datagrams
< vlen
) {
2075 if (datagrams
== vlen
- 1)
2078 if (MSG_CMSG_COMPAT
& flags
) {
2079 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2080 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2083 err
= __put_user(err
, &compat_entry
->msg_len
);
2086 err
= ___sys_sendmsg(sock
,
2087 (struct user_msghdr __user
*)entry
,
2088 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2091 err
= put_user(err
, &entry
->msg_len
);
2098 if (msg_data_left(&msg_sys
))
2103 fput_light(sock
->file
, fput_needed
);
2105 /* We only return an error if no datagrams were able to be sent */
2112 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2113 unsigned int, vlen
, unsigned int, flags
)
2115 if (flags
& MSG_CMSG_COMPAT
)
2117 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2120 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2121 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2123 struct compat_msghdr __user
*msg_compat
=
2124 (struct compat_msghdr __user
*)msg
;
2125 struct iovec iovstack
[UIO_FASTIOV
];
2126 struct iovec
*iov
= iovstack
;
2127 unsigned long cmsg_ptr
;
2131 /* kernel mode address */
2132 struct sockaddr_storage addr
;
2134 /* user mode address pointers */
2135 struct sockaddr __user
*uaddr
;
2136 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2138 msg_sys
->msg_name
= &addr
;
2140 if (MSG_CMSG_COMPAT
& flags
)
2141 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2143 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2147 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2148 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2150 /* We assume all kernel code knows the size of sockaddr_storage */
2151 msg_sys
->msg_namelen
= 0;
2153 if (sock
->file
->f_flags
& O_NONBLOCK
)
2154 flags
|= MSG_DONTWAIT
;
2155 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
, flags
);
2160 if (uaddr
!= NULL
) {
2161 err
= move_addr_to_user(&addr
,
2162 msg_sys
->msg_namelen
, uaddr
,
2167 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2171 if (MSG_CMSG_COMPAT
& flags
)
2172 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2173 &msg_compat
->msg_controllen
);
2175 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2176 &msg
->msg_controllen
);
2187 * BSD recvmsg interface
2190 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2192 int fput_needed
, err
;
2193 struct msghdr msg_sys
;
2194 struct socket
*sock
;
2196 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2200 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2202 fput_light(sock
->file
, fput_needed
);
2207 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2208 unsigned int, flags
)
2210 if (flags
& MSG_CMSG_COMPAT
)
2212 return __sys_recvmsg(fd
, msg
, flags
);
2216 * Linux recvmmsg interface
2219 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2220 unsigned int flags
, struct timespec
*timeout
)
2222 int fput_needed
, err
, datagrams
;
2223 struct socket
*sock
;
2224 struct mmsghdr __user
*entry
;
2225 struct compat_mmsghdr __user
*compat_entry
;
2226 struct msghdr msg_sys
;
2227 struct timespec64 end_time
;
2228 struct timespec64 timeout64
;
2231 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2237 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2241 err
= sock_error(sock
->sk
);
2248 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2250 while (datagrams
< vlen
) {
2252 * No need to ask LSM for more than the first datagram.
2254 if (MSG_CMSG_COMPAT
& flags
) {
2255 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2256 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2260 err
= __put_user(err
, &compat_entry
->msg_len
);
2263 err
= ___sys_recvmsg(sock
,
2264 (struct user_msghdr __user
*)entry
,
2265 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2269 err
= put_user(err
, &entry
->msg_len
);
2277 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2278 if (flags
& MSG_WAITFORONE
)
2279 flags
|= MSG_DONTWAIT
;
2282 ktime_get_ts64(&timeout64
);
2283 *timeout
= timespec64_to_timespec(
2284 timespec64_sub(end_time
, timeout64
));
2285 if (timeout
->tv_sec
< 0) {
2286 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2290 /* Timeout, return less than vlen datagrams */
2291 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2295 /* Out of band data, return right away */
2296 if (msg_sys
.msg_flags
& MSG_OOB
)
2304 if (datagrams
== 0) {
2310 * We may return less entries than requested (vlen) if the
2311 * sock is non block and there aren't enough datagrams...
2313 if (err
!= -EAGAIN
) {
2315 * ... or if recvmsg returns an error after we
2316 * received some datagrams, where we record the
2317 * error to return on the next call or if the
2318 * app asks about it using getsockopt(SO_ERROR).
2320 sock
->sk
->sk_err
= -err
;
2323 fput_light(sock
->file
, fput_needed
);
2328 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2329 unsigned int, vlen
, unsigned int, flags
,
2330 struct timespec __user
*, timeout
)
2333 struct timespec timeout_sys
;
2335 if (flags
& MSG_CMSG_COMPAT
)
2339 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2341 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2344 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2346 if (datagrams
> 0 &&
2347 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2348 datagrams
= -EFAULT
;
2353 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2354 /* Argument list sizes for sys_socketcall */
2355 #define AL(x) ((x) * sizeof(unsigned long))
2356 static const unsigned char nargs
[21] = {
2357 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2358 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2359 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2366 * System call vectors.
2368 * Argument checking cleaned up. Saved 20% in size.
2369 * This function doesn't need to set the kernel lock because
2370 * it is set by the callees.
2373 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2375 unsigned long a
[AUDITSC_ARGS
];
2376 unsigned long a0
, a1
;
2380 if (call
< 1 || call
> SYS_SENDMMSG
)
2384 if (len
> sizeof(a
))
2387 /* copy_from_user should be SMP safe. */
2388 if (copy_from_user(a
, args
, len
))
2391 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2400 err
= sys_socket(a0
, a1
, a
[2]);
2403 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2406 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2409 err
= sys_listen(a0
, a1
);
2412 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2413 (int __user
*)a
[2], 0);
2415 case SYS_GETSOCKNAME
:
2417 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2418 (int __user
*)a
[2]);
2420 case SYS_GETPEERNAME
:
2422 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2423 (int __user
*)a
[2]);
2425 case SYS_SOCKETPAIR
:
2426 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2429 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2432 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2433 (struct sockaddr __user
*)a
[4], a
[5]);
2436 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2439 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2440 (struct sockaddr __user
*)a
[4],
2441 (int __user
*)a
[5]);
2444 err
= sys_shutdown(a0
, a1
);
2446 case SYS_SETSOCKOPT
:
2447 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2449 case SYS_GETSOCKOPT
:
2451 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2452 (int __user
*)a
[4]);
2455 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2458 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2461 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2464 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2465 (struct timespec __user
*)a
[4]);
2468 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2469 (int __user
*)a
[2], a
[3]);
2478 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2481 * sock_register - add a socket protocol handler
2482 * @ops: description of protocol
2484 * This function is called by a protocol handler that wants to
2485 * advertise its address family, and have it linked into the
2486 * socket interface. The value ops->family corresponds to the
2487 * socket system call protocol family.
2489 int sock_register(const struct net_proto_family
*ops
)
2493 if (ops
->family
>= NPROTO
) {
2494 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2498 spin_lock(&net_family_lock
);
2499 if (rcu_dereference_protected(net_families
[ops
->family
],
2500 lockdep_is_held(&net_family_lock
)))
2503 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2506 spin_unlock(&net_family_lock
);
2508 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2511 EXPORT_SYMBOL(sock_register
);
2514 * sock_unregister - remove a protocol handler
2515 * @family: protocol family to remove
2517 * This function is called by a protocol handler that wants to
2518 * remove its address family, and have it unlinked from the
2519 * new socket creation.
2521 * If protocol handler is a module, then it can use module reference
2522 * counts to protect against new references. If protocol handler is not
2523 * a module then it needs to provide its own protection in
2524 * the ops->create routine.
2526 void sock_unregister(int family
)
2528 BUG_ON(family
< 0 || family
>= NPROTO
);
2530 spin_lock(&net_family_lock
);
2531 RCU_INIT_POINTER(net_families
[family
], NULL
);
2532 spin_unlock(&net_family_lock
);
2536 pr_info("NET: Unregistered protocol family %d\n", family
);
2538 EXPORT_SYMBOL(sock_unregister
);
2540 static int __init
sock_init(void)
2544 * Initialize the network sysctl infrastructure.
2546 err
= net_sysctl_init();
2551 * Initialize skbuff SLAB cache
2556 * Initialize the protocols module.
2561 err
= register_filesystem(&sock_fs_type
);
2564 sock_mnt
= kern_mount(&sock_fs_type
);
2565 if (IS_ERR(sock_mnt
)) {
2566 err
= PTR_ERR(sock_mnt
);
2570 /* The real protocol initialization is performed in later initcalls.
2573 #ifdef CONFIG_NETFILTER
2574 err
= netfilter_init();
2579 ptp_classifier_init();
2585 unregister_filesystem(&sock_fs_type
);
2590 core_initcall(sock_init
); /* early initcall */
2592 #ifdef CONFIG_PROC_FS
2593 void socket_seq_show(struct seq_file
*seq
)
2598 for_each_possible_cpu(cpu
)
2599 counter
+= per_cpu(sockets_in_use
, cpu
);
2601 /* It can be negative, by the way. 8) */
2605 seq_printf(seq
, "sockets: used %d\n", counter
);
2607 #endif /* CONFIG_PROC_FS */
2609 #ifdef CONFIG_COMPAT
2610 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2611 unsigned int cmd
, void __user
*up
)
2613 mm_segment_t old_fs
= get_fs();
2618 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2621 err
= compat_put_timeval(&ktv
, up
);
2626 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2627 unsigned int cmd
, void __user
*up
)
2629 mm_segment_t old_fs
= get_fs();
2630 struct timespec kts
;
2634 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2637 err
= compat_put_timespec(&kts
, up
);
2642 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2644 struct ifreq __user
*uifr
;
2647 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2648 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2651 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2655 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2661 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2663 struct compat_ifconf ifc32
;
2665 struct ifconf __user
*uifc
;
2666 struct compat_ifreq __user
*ifr32
;
2667 struct ifreq __user
*ifr
;
2671 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2674 memset(&ifc
, 0, sizeof(ifc
));
2675 if (ifc32
.ifcbuf
== 0) {
2679 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2681 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2682 sizeof(struct ifreq
);
2683 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2685 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2686 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2687 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2688 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2694 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2697 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2701 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2705 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2707 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2708 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2709 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2715 if (ifc32
.ifcbuf
== 0) {
2716 /* Translate from 64-bit structure multiple to
2720 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2725 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2731 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2733 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2734 bool convert_in
= false, convert_out
= false;
2735 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2736 struct ethtool_rxnfc __user
*rxnfc
;
2737 struct ifreq __user
*ifr
;
2738 u32 rule_cnt
= 0, actual_rule_cnt
;
2743 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2746 compat_rxnfc
= compat_ptr(data
);
2748 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2751 /* Most ethtool structures are defined without padding.
2752 * Unfortunately struct ethtool_rxnfc is an exception.
2757 case ETHTOOL_GRXCLSRLALL
:
2758 /* Buffer size is variable */
2759 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2761 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2763 buf_size
+= rule_cnt
* sizeof(u32
);
2765 case ETHTOOL_GRXRINGS
:
2766 case ETHTOOL_GRXCLSRLCNT
:
2767 case ETHTOOL_GRXCLSRULE
:
2768 case ETHTOOL_SRXCLSRLINS
:
2771 case ETHTOOL_SRXCLSRLDEL
:
2772 buf_size
+= sizeof(struct ethtool_rxnfc
);
2777 ifr
= compat_alloc_user_space(buf_size
);
2778 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2780 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2783 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2784 &ifr
->ifr_ifru
.ifru_data
))
2788 /* We expect there to be holes between fs.m_ext and
2789 * fs.ring_cookie and at the end of fs, but nowhere else.
2791 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2792 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2793 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2794 sizeof(rxnfc
->fs
.m_ext
));
2796 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2797 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2798 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2799 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2801 if (copy_in_user(rxnfc
, compat_rxnfc
,
2802 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2803 (void __user
*)rxnfc
) ||
2804 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2805 &compat_rxnfc
->fs
.ring_cookie
,
2806 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2807 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2808 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2809 sizeof(rxnfc
->rule_cnt
)))
2813 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2818 if (copy_in_user(compat_rxnfc
, rxnfc
,
2819 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2820 (const void __user
*)rxnfc
) ||
2821 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2822 &rxnfc
->fs
.ring_cookie
,
2823 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2824 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2825 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2826 sizeof(rxnfc
->rule_cnt
)))
2829 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2830 /* As an optimisation, we only copy the actual
2831 * number of rules that the underlying
2832 * function returned. Since Mallory might
2833 * change the rule count in user memory, we
2834 * check that it is less than the rule count
2835 * originally given (as the user buffer size),
2836 * which has been range-checked.
2838 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2840 if (actual_rule_cnt
< rule_cnt
)
2841 rule_cnt
= actual_rule_cnt
;
2842 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2843 &rxnfc
->rule_locs
[0],
2844 rule_cnt
* sizeof(u32
)))
2852 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2855 compat_uptr_t uptr32
;
2856 struct ifreq __user
*uifr
;
2858 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2859 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2862 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2865 uptr
= compat_ptr(uptr32
);
2867 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2870 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2873 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2874 struct compat_ifreq __user
*ifr32
)
2877 mm_segment_t old_fs
;
2881 case SIOCBONDENSLAVE
:
2882 case SIOCBONDRELEASE
:
2883 case SIOCBONDSETHWADDR
:
2884 case SIOCBONDCHANGEACTIVE
:
2885 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2890 err
= dev_ioctl(net
, cmd
,
2891 (struct ifreq __user __force
*) &kifr
);
2896 return -ENOIOCTLCMD
;
2900 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2901 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
2902 struct compat_ifreq __user
*u_ifreq32
)
2904 struct ifreq __user
*u_ifreq64
;
2905 char tmp_buf
[IFNAMSIZ
];
2906 void __user
*data64
;
2909 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2912 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2914 data64
= compat_ptr(data32
);
2916 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2918 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2921 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2924 return dev_ioctl(net
, cmd
, u_ifreq64
);
2927 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2928 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2930 struct ifreq __user
*uifr
;
2933 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2934 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2937 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2948 case SIOCGIFBRDADDR
:
2949 case SIOCGIFDSTADDR
:
2950 case SIOCGIFNETMASK
:
2955 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2963 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2964 struct compat_ifreq __user
*uifr32
)
2967 struct compat_ifmap __user
*uifmap32
;
2968 mm_segment_t old_fs
;
2971 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2972 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2973 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2974 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2975 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2976 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2977 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2978 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2984 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
2987 if (cmd
== SIOCGIFMAP
&& !err
) {
2988 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2989 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2990 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2991 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2992 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2993 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2994 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3003 struct sockaddr rt_dst
; /* target address */
3004 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3005 struct sockaddr rt_genmask
; /* target network mask (IP) */
3006 unsigned short rt_flags
;
3009 unsigned char rt_tos
;
3010 unsigned char rt_class
;
3012 short rt_metric
; /* +1 for binary compatibility! */
3013 /* char * */ u32 rt_dev
; /* forcing the device at add */
3014 u32 rt_mtu
; /* per route MTU/Window */
3015 u32 rt_window
; /* Window clamping */
3016 unsigned short rt_irtt
; /* Initial RTT */
3019 struct in6_rtmsg32
{
3020 struct in6_addr rtmsg_dst
;
3021 struct in6_addr rtmsg_src
;
3022 struct in6_addr rtmsg_gateway
;
3032 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3033 unsigned int cmd
, void __user
*argp
)
3037 struct in6_rtmsg r6
;
3041 mm_segment_t old_fs
= get_fs();
3043 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3044 struct in6_rtmsg32 __user
*ur6
= argp
;
3045 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3046 3 * sizeof(struct in6_addr
));
3047 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3048 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3049 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3050 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3051 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3052 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3053 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3057 struct rtentry32 __user
*ur4
= argp
;
3058 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3059 3 * sizeof(struct sockaddr
));
3060 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3061 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3062 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3063 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3064 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3065 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3067 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3068 r4
.rt_dev
= (char __user __force
*)devname
;
3082 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3089 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3090 * for some operations; this forces use of the newer bridge-utils that
3091 * use compatible ioctls
3093 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3097 if (get_user(tmp
, argp
))
3099 if (tmp
== BRCTL_GET_VERSION
)
3100 return BRCTL_VERSION
+ 1;
3104 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3105 unsigned int cmd
, unsigned long arg
)
3107 void __user
*argp
= compat_ptr(arg
);
3108 struct sock
*sk
= sock
->sk
;
3109 struct net
*net
= sock_net(sk
);
3111 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3112 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3117 return old_bridge_ioctl(argp
);
3119 return dev_ifname32(net
, argp
);
3121 return dev_ifconf(net
, argp
);
3123 return ethtool_ioctl(net
, argp
);
3125 return compat_siocwandev(net
, argp
);
3128 return compat_sioc_ifmap(net
, cmd
, argp
);
3129 case SIOCBONDENSLAVE
:
3130 case SIOCBONDRELEASE
:
3131 case SIOCBONDSETHWADDR
:
3132 case SIOCBONDCHANGEACTIVE
:
3133 return bond_ioctl(net
, cmd
, argp
);
3136 return routing_ioctl(net
, sock
, cmd
, argp
);
3138 return do_siocgstamp(net
, sock
, cmd
, argp
);
3140 return do_siocgstampns(net
, sock
, cmd
, argp
);
3141 case SIOCBONDSLAVEINFOQUERY
:
3142 case SIOCBONDINFOQUERY
:
3145 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3158 return sock_ioctl(file
, cmd
, arg
);
3175 case SIOCSIFHWBROADCAST
:
3177 case SIOCGIFBRDADDR
:
3178 case SIOCSIFBRDADDR
:
3179 case SIOCGIFDSTADDR
:
3180 case SIOCSIFDSTADDR
:
3181 case SIOCGIFNETMASK
:
3182 case SIOCSIFNETMASK
:
3193 return dev_ifsioc(net
, sock
, cmd
, argp
);
3199 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3202 return -ENOIOCTLCMD
;
3205 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3208 struct socket
*sock
= file
->private_data
;
3209 int ret
= -ENOIOCTLCMD
;
3216 if (sock
->ops
->compat_ioctl
)
3217 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3219 if (ret
== -ENOIOCTLCMD
&&
3220 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3221 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3223 if (ret
== -ENOIOCTLCMD
)
3224 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3230 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3232 return sock
->ops
->bind(sock
, addr
, addrlen
);
3234 EXPORT_SYMBOL(kernel_bind
);
3236 int kernel_listen(struct socket
*sock
, int backlog
)
3238 return sock
->ops
->listen(sock
, backlog
);
3240 EXPORT_SYMBOL(kernel_listen
);
3242 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3244 struct sock
*sk
= sock
->sk
;
3247 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3252 err
= sock
->ops
->accept(sock
, *newsock
, flags
);
3254 sock_release(*newsock
);
3259 (*newsock
)->ops
= sock
->ops
;
3260 __module_get((*newsock
)->ops
->owner
);
3265 EXPORT_SYMBOL(kernel_accept
);
3267 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3270 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3272 EXPORT_SYMBOL(kernel_connect
);
3274 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3277 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3279 EXPORT_SYMBOL(kernel_getsockname
);
3281 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3284 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3286 EXPORT_SYMBOL(kernel_getpeername
);
3288 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3289 char *optval
, int *optlen
)
3291 mm_segment_t oldfs
= get_fs();
3292 char __user
*uoptval
;
3293 int __user
*uoptlen
;
3296 uoptval
= (char __user __force
*) optval
;
3297 uoptlen
= (int __user __force
*) optlen
;
3300 if (level
== SOL_SOCKET
)
3301 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3303 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3308 EXPORT_SYMBOL(kernel_getsockopt
);
3310 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3311 char *optval
, unsigned int optlen
)
3313 mm_segment_t oldfs
= get_fs();
3314 char __user
*uoptval
;
3317 uoptval
= (char __user __force
*) optval
;
3320 if (level
== SOL_SOCKET
)
3321 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3323 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3328 EXPORT_SYMBOL(kernel_setsockopt
);
3330 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3331 size_t size
, int flags
)
3333 if (sock
->ops
->sendpage
)
3334 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3336 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3338 EXPORT_SYMBOL(kernel_sendpage
);
3340 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3342 mm_segment_t oldfs
= get_fs();
3346 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3351 EXPORT_SYMBOL(kernel_sock_ioctl
);
3353 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3355 return sock
->ops
->shutdown(sock
, how
);
3357 EXPORT_SYMBOL(kernel_sock_shutdown
);