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 void 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 BUG_ON(sock_inode_cachep
== NULL
);
302 static const struct super_operations sockfs_ops
= {
303 .alloc_inode
= sock_alloc_inode
,
304 .destroy_inode
= sock_destroy_inode
,
305 .statfs
= simple_statfs
,
309 * sockfs_dname() is called from d_path().
311 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
313 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
314 d_inode(dentry
)->i_ino
);
317 static const struct dentry_operations sockfs_dentry_operations
= {
318 .d_dname
= sockfs_dname
,
321 static int sockfs_xattr_get(const struct xattr_handler
*handler
,
322 struct dentry
*dentry
, struct inode
*inode
,
323 const char *suffix
, void *value
, size_t size
)
326 if (dentry
->d_name
.len
+ 1 > size
)
328 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
330 return dentry
->d_name
.len
+ 1;
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
337 static const struct xattr_handler sockfs_xattr_handler
= {
338 .name
= XATTR_NAME_SOCKPROTONAME
,
339 .get
= sockfs_xattr_get
,
342 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
343 struct dentry
*dentry
, struct inode
*inode
,
344 const char *suffix
, const void *value
,
345 size_t size
, int flags
)
347 /* Handled by LSM. */
351 static const struct xattr_handler sockfs_security_xattr_handler
= {
352 .prefix
= XATTR_SECURITY_PREFIX
,
353 .set
= sockfs_security_xattr_set
,
356 static const struct xattr_handler
*sockfs_xattr_handlers
[] = {
357 &sockfs_xattr_handler
,
358 &sockfs_security_xattr_handler
,
362 static struct dentry
*sockfs_mount(struct file_system_type
*fs_type
,
363 int flags
, const char *dev_name
, void *data
)
365 return mount_pseudo_xattr(fs_type
, "socket:", &sockfs_ops
,
366 sockfs_xattr_handlers
,
367 &sockfs_dentry_operations
, SOCKFS_MAGIC
);
370 static struct vfsmount
*sock_mnt __read_mostly
;
372 static struct file_system_type sock_fs_type
= {
374 .mount
= sockfs_mount
,
375 .kill_sb
= kill_anon_super
,
379 * Obtains the first available file descriptor and sets it up for use.
381 * These functions create file structures and maps them to fd space
382 * of the current process. On success it returns file descriptor
383 * and file struct implicitly stored in sock->file.
384 * Note that another thread may close file descriptor before we return
385 * from this function. We use the fact that now we do not refer
386 * to socket after mapping. If one day we will need it, this
387 * function will increment ref. count on file by 1.
389 * In any case returned fd MAY BE not valid!
390 * This race condition is unavoidable
391 * with shared fd spaces, we cannot solve it inside kernel,
392 * but we take care of internal coherence yet.
395 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
397 struct qstr name
= { .name
= "" };
403 name
.len
= strlen(name
.name
);
404 } else if (sock
->sk
) {
405 name
.name
= sock
->sk
->sk_prot_creator
->name
;
406 name
.len
= strlen(name
.name
);
408 path
.dentry
= d_alloc_pseudo(sock_mnt
->mnt_sb
, &name
);
409 if (unlikely(!path
.dentry
))
410 return ERR_PTR(-ENOMEM
);
411 path
.mnt
= mntget(sock_mnt
);
413 d_instantiate(path
.dentry
, SOCK_INODE(sock
));
415 file
= alloc_file(&path
, FMODE_READ
| FMODE_WRITE
,
418 /* drop dentry, keep inode */
419 ihold(d_inode(path
.dentry
));
425 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
426 file
->private_data
= sock
;
429 EXPORT_SYMBOL(sock_alloc_file
);
431 static int sock_map_fd(struct socket
*sock
, int flags
)
433 struct file
*newfile
;
434 int fd
= get_unused_fd_flags(flags
);
435 if (unlikely(fd
< 0))
438 newfile
= sock_alloc_file(sock
, flags
, NULL
);
439 if (likely(!IS_ERR(newfile
))) {
440 fd_install(fd
, newfile
);
445 return PTR_ERR(newfile
);
448 struct socket
*sock_from_file(struct file
*file
, int *err
)
450 if (file
->f_op
== &socket_file_ops
)
451 return file
->private_data
; /* set in sock_map_fd */
456 EXPORT_SYMBOL(sock_from_file
);
459 * sockfd_lookup - Go from a file number to its socket slot
461 * @err: pointer to an error code return
463 * The file handle passed in is locked and the socket it is bound
464 * too is returned. If an error occurs the err pointer is overwritten
465 * with a negative errno code and NULL is returned. The function checks
466 * for both invalid handles and passing a handle which is not a socket.
468 * On a success the socket object pointer is returned.
471 struct socket
*sockfd_lookup(int fd
, int *err
)
482 sock
= sock_from_file(file
, err
);
487 EXPORT_SYMBOL(sockfd_lookup
);
489 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
491 struct fd f
= fdget(fd
);
496 sock
= sock_from_file(f
.file
, err
);
498 *fput_needed
= f
.flags
;
506 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
512 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
522 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
527 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
534 static int sockfs_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
536 int err
= simple_setattr(dentry
, iattr
);
538 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
539 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
541 sock
->sk
->sk_uid
= iattr
->ia_uid
;
547 static const struct inode_operations sockfs_inode_ops
= {
548 .listxattr
= sockfs_listxattr
,
549 .setattr
= sockfs_setattr
,
553 * sock_alloc - allocate a socket
555 * Allocate a new inode and socket object. The two are bound together
556 * and initialised. The socket is then returned. If we are out of inodes
560 struct socket
*sock_alloc(void)
565 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
569 sock
= SOCKET_I(inode
);
571 kmemcheck_annotate_bitfield(sock
, type
);
572 inode
->i_ino
= get_next_ino();
573 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
574 inode
->i_uid
= current_fsuid();
575 inode
->i_gid
= current_fsgid();
576 inode
->i_op
= &sockfs_inode_ops
;
578 this_cpu_add(sockets_in_use
, 1);
581 EXPORT_SYMBOL(sock_alloc
);
584 * sock_release - close a socket
585 * @sock: socket to close
587 * The socket is released from the protocol stack if it has a release
588 * callback, and the inode is then released if the socket is bound to
589 * an inode not a file.
592 void sock_release(struct socket
*sock
)
595 struct module
*owner
= sock
->ops
->owner
;
597 sock
->ops
->release(sock
);
602 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
603 pr_err("%s: fasync list not empty!\n", __func__
);
605 this_cpu_sub(sockets_in_use
, 1);
607 iput(SOCK_INODE(sock
));
612 EXPORT_SYMBOL(sock_release
);
614 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
616 u8 flags
= *tx_flags
;
618 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
619 flags
|= SKBTX_HW_TSTAMP
;
621 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
622 flags
|= SKBTX_SW_TSTAMP
;
624 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
625 flags
|= SKBTX_SCHED_TSTAMP
;
629 EXPORT_SYMBOL(__sock_tx_timestamp
);
631 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
633 int ret
= sock
->ops
->sendmsg(sock
, msg
, msg_data_left(msg
));
634 BUG_ON(ret
== -EIOCBQUEUED
);
638 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
640 int err
= security_socket_sendmsg(sock
, msg
,
643 return err
?: sock_sendmsg_nosec(sock
, msg
);
645 EXPORT_SYMBOL(sock_sendmsg
);
647 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
648 struct kvec
*vec
, size_t num
, size_t size
)
650 iov_iter_kvec(&msg
->msg_iter
, WRITE
| ITER_KVEC
, vec
, num
, size
);
651 return sock_sendmsg(sock
, msg
);
653 EXPORT_SYMBOL(kernel_sendmsg
);
656 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
658 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
661 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
662 struct scm_timestamping tss
;
664 struct skb_shared_hwtstamps
*shhwtstamps
=
667 /* Race occurred between timestamp enabling and packet
668 receiving. Fill in the current time for now. */
669 if (need_software_tstamp
&& skb
->tstamp
== 0)
670 __net_timestamp(skb
);
672 if (need_software_tstamp
) {
673 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
675 skb_get_timestamp(skb
, &tv
);
676 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
680 skb_get_timestampns(skb
, &ts
);
681 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
686 memset(&tss
, 0, sizeof(tss
));
687 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
688 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
691 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
692 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2))
695 put_cmsg(msg
, SOL_SOCKET
,
696 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
698 if (skb
->len
&& (sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_STATS
))
699 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
700 skb
->len
, skb
->data
);
703 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
705 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
710 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
712 if (!skb
->wifi_acked_valid
)
715 ack
= skb
->wifi_acked
;
717 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
719 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
721 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
724 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
725 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
726 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
729 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
732 sock_recv_timestamp(msg
, sk
, skb
);
733 sock_recv_drops(msg
, sk
, skb
);
735 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
737 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
740 return sock
->ops
->recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
743 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
745 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
747 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
749 EXPORT_SYMBOL(sock_recvmsg
);
752 * kernel_recvmsg - Receive a message from a socket (kernel space)
753 * @sock: The socket to receive the message from
754 * @msg: Received message
755 * @vec: Input s/g array for message data
756 * @num: Size of input s/g array
757 * @size: Number of bytes to read
758 * @flags: Message flags (MSG_DONTWAIT, etc...)
760 * On return the msg structure contains the scatter/gather array passed in the
761 * vec argument. The array is modified so that it consists of the unfilled
762 * portion of the original array.
764 * The returned value is the total number of bytes received, or an error.
766 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
767 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
769 mm_segment_t oldfs
= get_fs();
772 iov_iter_kvec(&msg
->msg_iter
, READ
| ITER_KVEC
, vec
, num
, size
);
774 result
= sock_recvmsg(sock
, msg
, flags
);
778 EXPORT_SYMBOL(kernel_recvmsg
);
780 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
781 int offset
, size_t size
, loff_t
*ppos
, int more
)
786 sock
= file
->private_data
;
788 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
789 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
792 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
795 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
796 struct pipe_inode_info
*pipe
, size_t len
,
799 struct socket
*sock
= file
->private_data
;
801 if (unlikely(!sock
->ops
->splice_read
))
804 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
807 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
809 struct file
*file
= iocb
->ki_filp
;
810 struct socket
*sock
= file
->private_data
;
811 struct msghdr msg
= {.msg_iter
= *to
,
815 if (file
->f_flags
& O_NONBLOCK
)
816 msg
.msg_flags
= MSG_DONTWAIT
;
818 if (iocb
->ki_pos
!= 0)
821 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
824 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
829 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
831 struct file
*file
= iocb
->ki_filp
;
832 struct socket
*sock
= file
->private_data
;
833 struct msghdr msg
= {.msg_iter
= *from
,
837 if (iocb
->ki_pos
!= 0)
840 if (file
->f_flags
& O_NONBLOCK
)
841 msg
.msg_flags
= MSG_DONTWAIT
;
843 if (sock
->type
== SOCK_SEQPACKET
)
844 msg
.msg_flags
|= MSG_EOR
;
846 res
= sock_sendmsg(sock
, &msg
);
847 *from
= msg
.msg_iter
;
852 * Atomic setting of ioctl hooks to avoid race
853 * with module unload.
856 static DEFINE_MUTEX(br_ioctl_mutex
);
857 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
859 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
861 mutex_lock(&br_ioctl_mutex
);
862 br_ioctl_hook
= hook
;
863 mutex_unlock(&br_ioctl_mutex
);
865 EXPORT_SYMBOL(brioctl_set
);
867 static DEFINE_MUTEX(vlan_ioctl_mutex
);
868 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
870 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
872 mutex_lock(&vlan_ioctl_mutex
);
873 vlan_ioctl_hook
= hook
;
874 mutex_unlock(&vlan_ioctl_mutex
);
876 EXPORT_SYMBOL(vlan_ioctl_set
);
878 static DEFINE_MUTEX(dlci_ioctl_mutex
);
879 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
881 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
883 mutex_lock(&dlci_ioctl_mutex
);
884 dlci_ioctl_hook
= hook
;
885 mutex_unlock(&dlci_ioctl_mutex
);
887 EXPORT_SYMBOL(dlci_ioctl_set
);
889 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
890 unsigned int cmd
, unsigned long arg
)
893 void __user
*argp
= (void __user
*)arg
;
895 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
898 * If this ioctl is unknown try to hand it down
901 if (err
== -ENOIOCTLCMD
)
902 err
= dev_ioctl(net
, cmd
, argp
);
908 * With an ioctl, arg may well be a user mode pointer, but we don't know
909 * what to do with it - that's up to the protocol still.
912 static struct ns_common
*get_net_ns(struct ns_common
*ns
)
914 return &get_net(container_of(ns
, struct net
, ns
))->ns
;
917 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
921 void __user
*argp
= (void __user
*)arg
;
925 sock
= file
->private_data
;
928 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
929 err
= dev_ioctl(net
, cmd
, argp
);
931 #ifdef CONFIG_WEXT_CORE
932 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
933 err
= dev_ioctl(net
, cmd
, argp
);
940 if (get_user(pid
, (int __user
*)argp
))
942 f_setown(sock
->file
, pid
, 1);
947 err
= put_user(f_getown(sock
->file
),
956 request_module("bridge");
958 mutex_lock(&br_ioctl_mutex
);
960 err
= br_ioctl_hook(net
, cmd
, argp
);
961 mutex_unlock(&br_ioctl_mutex
);
966 if (!vlan_ioctl_hook
)
967 request_module("8021q");
969 mutex_lock(&vlan_ioctl_mutex
);
971 err
= vlan_ioctl_hook(net
, argp
);
972 mutex_unlock(&vlan_ioctl_mutex
);
977 if (!dlci_ioctl_hook
)
978 request_module("dlci");
980 mutex_lock(&dlci_ioctl_mutex
);
982 err
= dlci_ioctl_hook(cmd
, argp
);
983 mutex_unlock(&dlci_ioctl_mutex
);
987 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
990 err
= open_related_ns(&net
->ns
, get_net_ns
);
993 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
999 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1002 struct socket
*sock
= NULL
;
1004 err
= security_socket_create(family
, type
, protocol
, 1);
1008 sock
= sock_alloc();
1015 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1027 EXPORT_SYMBOL(sock_create_lite
);
1029 /* No kernel lock held - perfect */
1030 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1032 unsigned int busy_flag
= 0;
1033 struct socket
*sock
;
1036 * We can't return errors to poll, so it's either yes or no.
1038 sock
= file
->private_data
;
1040 if (sk_can_busy_loop(sock
->sk
)) {
1041 /* this socket can poll_ll so tell the system call */
1042 busy_flag
= POLL_BUSY_LOOP
;
1044 /* once, only if requested by syscall */
1045 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1046 sk_busy_loop(sock
->sk
, 1);
1049 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1052 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1054 struct socket
*sock
= file
->private_data
;
1056 return sock
->ops
->mmap(file
, sock
, vma
);
1059 static int sock_close(struct inode
*inode
, struct file
*filp
)
1061 sock_release(SOCKET_I(inode
));
1066 * Update the socket async list
1068 * Fasync_list locking strategy.
1070 * 1. fasync_list is modified only under process context socket lock
1071 * i.e. under semaphore.
1072 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1073 * or under socket lock
1076 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1078 struct socket
*sock
= filp
->private_data
;
1079 struct sock
*sk
= sock
->sk
;
1080 struct socket_wq
*wq
;
1086 wq
= rcu_dereference_protected(sock
->wq
, lockdep_sock_is_held(sk
));
1087 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1089 if (!wq
->fasync_list
)
1090 sock_reset_flag(sk
, SOCK_FASYNC
);
1092 sock_set_flag(sk
, SOCK_FASYNC
);
1098 /* This function may be called only under rcu_lock */
1100 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1102 if (!wq
|| !wq
->fasync_list
)
1106 case SOCK_WAKE_WAITD
:
1107 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1110 case SOCK_WAKE_SPACE
:
1111 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1116 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1119 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1124 EXPORT_SYMBOL(sock_wake_async
);
1126 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1127 struct socket
**res
, int kern
)
1130 struct socket
*sock
;
1131 const struct net_proto_family
*pf
;
1134 * Check protocol is in range
1136 if (family
< 0 || family
>= NPROTO
)
1137 return -EAFNOSUPPORT
;
1138 if (type
< 0 || type
>= SOCK_MAX
)
1143 This uglymoron is moved from INET layer to here to avoid
1144 deadlock in module load.
1146 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1147 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1152 err
= security_socket_create(family
, type
, protocol
, kern
);
1157 * Allocate the socket and allow the family to set things up. if
1158 * the protocol is 0, the family is instructed to select an appropriate
1161 sock
= sock_alloc();
1163 net_warn_ratelimited("socket: no more sockets\n");
1164 return -ENFILE
; /* Not exactly a match, but its the
1165 closest posix thing */
1170 #ifdef CONFIG_MODULES
1171 /* Attempt to load a protocol module if the find failed.
1173 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1174 * requested real, full-featured networking support upon configuration.
1175 * Otherwise module support will break!
1177 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1178 request_module("net-pf-%d", family
);
1182 pf
= rcu_dereference(net_families
[family
]);
1183 err
= -EAFNOSUPPORT
;
1188 * We will call the ->create function, that possibly is in a loadable
1189 * module, so we have to bump that loadable module refcnt first.
1191 if (!try_module_get(pf
->owner
))
1194 /* Now protected by module ref count */
1197 err
= pf
->create(net
, sock
, protocol
, kern
);
1199 goto out_module_put
;
1202 * Now to bump the refcnt of the [loadable] module that owns this
1203 * socket at sock_release time we decrement its refcnt.
1205 if (!try_module_get(sock
->ops
->owner
))
1206 goto out_module_busy
;
1209 * Now that we're done with the ->create function, the [loadable]
1210 * module can have its refcnt decremented
1212 module_put(pf
->owner
);
1213 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1215 goto out_sock_release
;
1221 err
= -EAFNOSUPPORT
;
1224 module_put(pf
->owner
);
1231 goto out_sock_release
;
1233 EXPORT_SYMBOL(__sock_create
);
1235 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1237 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1239 EXPORT_SYMBOL(sock_create
);
1241 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1243 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1245 EXPORT_SYMBOL(sock_create_kern
);
1247 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1250 struct socket
*sock
;
1253 /* Check the SOCK_* constants for consistency. */
1254 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1255 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1256 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1257 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1259 flags
= type
& ~SOCK_TYPE_MASK
;
1260 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1262 type
&= SOCK_TYPE_MASK
;
1264 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1265 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1267 retval
= sock_create(family
, type
, protocol
, &sock
);
1271 retval
= sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1276 /* It may be already another descriptor 8) Not kernel problem. */
1285 * Create a pair of connected sockets.
1288 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1289 int __user
*, usockvec
)
1291 struct socket
*sock1
, *sock2
;
1293 struct file
*newfile1
, *newfile2
;
1296 flags
= type
& ~SOCK_TYPE_MASK
;
1297 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1299 type
&= SOCK_TYPE_MASK
;
1301 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1302 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1305 * Obtain the first socket and check if the underlying protocol
1306 * supports the socketpair call.
1309 err
= sock_create(family
, type
, protocol
, &sock1
);
1313 err
= sock_create(family
, type
, protocol
, &sock2
);
1317 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1319 goto out_release_both
;
1321 fd1
= get_unused_fd_flags(flags
);
1322 if (unlikely(fd1
< 0)) {
1324 goto out_release_both
;
1327 fd2
= get_unused_fd_flags(flags
);
1328 if (unlikely(fd2
< 0)) {
1330 goto out_put_unused_1
;
1333 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1334 if (IS_ERR(newfile1
)) {
1335 err
= PTR_ERR(newfile1
);
1336 goto out_put_unused_both
;
1339 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1340 if (IS_ERR(newfile2
)) {
1341 err
= PTR_ERR(newfile2
);
1345 err
= put_user(fd1
, &usockvec
[0]);
1349 err
= put_user(fd2
, &usockvec
[1]);
1353 audit_fd_pair(fd1
, fd2
);
1355 fd_install(fd1
, newfile1
);
1356 fd_install(fd2
, newfile2
);
1357 /* fd1 and fd2 may be already another descriptors.
1358 * Not kernel problem.
1374 sock_release(sock2
);
1377 out_put_unused_both
:
1382 sock_release(sock2
);
1384 sock_release(sock1
);
1390 * Bind a name to a socket. Nothing much to do here since it's
1391 * the protocol's responsibility to handle the local address.
1393 * We move the socket address to kernel space before we call
1394 * the protocol layer (having also checked the address is ok).
1397 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1399 struct socket
*sock
;
1400 struct sockaddr_storage address
;
1401 int err
, fput_needed
;
1403 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1405 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1407 err
= security_socket_bind(sock
,
1408 (struct sockaddr
*)&address
,
1411 err
= sock
->ops
->bind(sock
,
1415 fput_light(sock
->file
, fput_needed
);
1421 * Perform a listen. Basically, we allow the protocol to do anything
1422 * necessary for a listen, and if that works, we mark the socket as
1423 * ready for listening.
1426 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1428 struct socket
*sock
;
1429 int err
, fput_needed
;
1432 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1434 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1435 if ((unsigned int)backlog
> somaxconn
)
1436 backlog
= somaxconn
;
1438 err
= security_socket_listen(sock
, backlog
);
1440 err
= sock
->ops
->listen(sock
, backlog
);
1442 fput_light(sock
->file
, fput_needed
);
1448 * For accept, we attempt to create a new socket, set up the link
1449 * with the client, wake up the client, then return the new
1450 * connected fd. We collect the address of the connector in kernel
1451 * space and move it to user at the very end. This is unclean because
1452 * we open the socket then return an error.
1454 * 1003.1g adds the ability to recvmsg() to query connection pending
1455 * status to recvmsg. We need to add that support in a way thats
1456 * clean when we restucture accept also.
1459 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1460 int __user
*, upeer_addrlen
, int, flags
)
1462 struct socket
*sock
, *newsock
;
1463 struct file
*newfile
;
1464 int err
, len
, newfd
, fput_needed
;
1465 struct sockaddr_storage address
;
1467 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1470 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1471 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1473 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1478 newsock
= sock_alloc();
1482 newsock
->type
= sock
->type
;
1483 newsock
->ops
= sock
->ops
;
1486 * We don't need try_module_get here, as the listening socket (sock)
1487 * has the protocol module (sock->ops->owner) held.
1489 __module_get(newsock
->ops
->owner
);
1491 newfd
= get_unused_fd_flags(flags
);
1492 if (unlikely(newfd
< 0)) {
1494 sock_release(newsock
);
1497 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1498 if (IS_ERR(newfile
)) {
1499 err
= PTR_ERR(newfile
);
1500 put_unused_fd(newfd
);
1501 sock_release(newsock
);
1505 err
= security_socket_accept(sock
, newsock
);
1509 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
, false);
1513 if (upeer_sockaddr
) {
1514 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1516 err
= -ECONNABORTED
;
1519 err
= move_addr_to_user(&address
,
1520 len
, upeer_sockaddr
, upeer_addrlen
);
1525 /* File flags are not inherited via accept() unlike another OSes. */
1527 fd_install(newfd
, newfile
);
1531 fput_light(sock
->file
, fput_needed
);
1536 put_unused_fd(newfd
);
1540 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1541 int __user
*, upeer_addrlen
)
1543 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1547 * Attempt to connect to a socket with the server address. The address
1548 * is in user space so we verify it is OK and move it to kernel space.
1550 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1553 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1554 * other SEQPACKET protocols that take time to connect() as it doesn't
1555 * include the -EINPROGRESS status for such sockets.
1558 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1561 struct socket
*sock
;
1562 struct sockaddr_storage address
;
1563 int err
, fput_needed
;
1565 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1568 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1573 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1577 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1578 sock
->file
->f_flags
);
1580 fput_light(sock
->file
, fput_needed
);
1586 * Get the local address ('name') of a socket object. Move the obtained
1587 * name to user space.
1590 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1591 int __user
*, usockaddr_len
)
1593 struct socket
*sock
;
1594 struct sockaddr_storage address
;
1595 int len
, err
, fput_needed
;
1597 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1601 err
= security_socket_getsockname(sock
);
1605 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1608 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1611 fput_light(sock
->file
, fput_needed
);
1617 * Get the remote address ('name') of a socket object. Move the obtained
1618 * name to user space.
1621 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1622 int __user
*, usockaddr_len
)
1624 struct socket
*sock
;
1625 struct sockaddr_storage address
;
1626 int len
, err
, fput_needed
;
1628 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1630 err
= security_socket_getpeername(sock
);
1632 fput_light(sock
->file
, fput_needed
);
1637 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1640 err
= move_addr_to_user(&address
, len
, usockaddr
,
1642 fput_light(sock
->file
, fput_needed
);
1648 * Send a datagram to a given address. We move the address into kernel
1649 * space and check the user space data area is readable before invoking
1653 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1654 unsigned int, flags
, struct sockaddr __user
*, addr
,
1657 struct socket
*sock
;
1658 struct sockaddr_storage address
;
1664 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
1667 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1671 msg
.msg_name
= NULL
;
1672 msg
.msg_control
= NULL
;
1673 msg
.msg_controllen
= 0;
1674 msg
.msg_namelen
= 0;
1676 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1679 msg
.msg_name
= (struct sockaddr
*)&address
;
1680 msg
.msg_namelen
= addr_len
;
1682 if (sock
->file
->f_flags
& O_NONBLOCK
)
1683 flags
|= MSG_DONTWAIT
;
1684 msg
.msg_flags
= flags
;
1685 err
= sock_sendmsg(sock
, &msg
);
1688 fput_light(sock
->file
, fput_needed
);
1694 * Send a datagram down a socket.
1697 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1698 unsigned int, flags
)
1700 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1704 * Receive a frame from the socket and optionally record the address of the
1705 * sender. We verify the buffers are writable and if needed move the
1706 * sender address from kernel to user space.
1709 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1710 unsigned int, flags
, struct sockaddr __user
*, addr
,
1711 int __user
*, addr_len
)
1713 struct socket
*sock
;
1716 struct sockaddr_storage address
;
1720 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
1723 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1727 msg
.msg_control
= NULL
;
1728 msg
.msg_controllen
= 0;
1729 /* Save some cycles and don't copy the address if not needed */
1730 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1731 /* We assume all kernel code knows the size of sockaddr_storage */
1732 msg
.msg_namelen
= 0;
1733 msg
.msg_iocb
= NULL
;
1735 if (sock
->file
->f_flags
& O_NONBLOCK
)
1736 flags
|= MSG_DONTWAIT
;
1737 err
= sock_recvmsg(sock
, &msg
, flags
);
1739 if (err
>= 0 && addr
!= NULL
) {
1740 err2
= move_addr_to_user(&address
,
1741 msg
.msg_namelen
, addr
, addr_len
);
1746 fput_light(sock
->file
, fput_needed
);
1752 * Receive a datagram from a socket.
1755 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1756 unsigned int, flags
)
1758 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1762 * Set a socket option. Because we don't know the option lengths we have
1763 * to pass the user mode parameter for the protocols to sort out.
1766 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1767 char __user
*, optval
, int, optlen
)
1769 int err
, fput_needed
;
1770 struct socket
*sock
;
1775 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1777 err
= security_socket_setsockopt(sock
, level
, optname
);
1781 if (level
== SOL_SOCKET
)
1783 sock_setsockopt(sock
, level
, optname
, optval
,
1787 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1790 fput_light(sock
->file
, fput_needed
);
1796 * Get a socket option. Because we don't know the option lengths we have
1797 * to pass a user mode parameter for the protocols to sort out.
1800 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1801 char __user
*, optval
, int __user
*, optlen
)
1803 int err
, fput_needed
;
1804 struct socket
*sock
;
1806 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1808 err
= security_socket_getsockopt(sock
, level
, optname
);
1812 if (level
== SOL_SOCKET
)
1814 sock_getsockopt(sock
, level
, optname
, optval
,
1818 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1821 fput_light(sock
->file
, fput_needed
);
1827 * Shutdown a socket.
1830 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1832 int err
, fput_needed
;
1833 struct socket
*sock
;
1835 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1837 err
= security_socket_shutdown(sock
, how
);
1839 err
= sock
->ops
->shutdown(sock
, how
);
1840 fput_light(sock
->file
, fput_needed
);
1845 /* A couple of helpful macros for getting the address of the 32/64 bit
1846 * fields which are the same type (int / unsigned) on our platforms.
1848 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1849 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1850 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1852 struct used_address
{
1853 struct sockaddr_storage name
;
1854 unsigned int name_len
;
1857 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
1858 struct user_msghdr __user
*umsg
,
1859 struct sockaddr __user
**save_addr
,
1862 struct sockaddr __user
*uaddr
;
1863 struct iovec __user
*uiov
;
1867 if (!access_ok(VERIFY_READ
, umsg
, sizeof(*umsg
)) ||
1868 __get_user(uaddr
, &umsg
->msg_name
) ||
1869 __get_user(kmsg
->msg_namelen
, &umsg
->msg_namelen
) ||
1870 __get_user(uiov
, &umsg
->msg_iov
) ||
1871 __get_user(nr_segs
, &umsg
->msg_iovlen
) ||
1872 __get_user(kmsg
->msg_control
, &umsg
->msg_control
) ||
1873 __get_user(kmsg
->msg_controllen
, &umsg
->msg_controllen
) ||
1874 __get_user(kmsg
->msg_flags
, &umsg
->msg_flags
))
1878 kmsg
->msg_namelen
= 0;
1880 if (kmsg
->msg_namelen
< 0)
1883 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1884 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1889 if (uaddr
&& kmsg
->msg_namelen
) {
1891 err
= move_addr_to_kernel(uaddr
, kmsg
->msg_namelen
,
1897 kmsg
->msg_name
= NULL
;
1898 kmsg
->msg_namelen
= 0;
1901 if (nr_segs
> UIO_MAXIOV
)
1904 kmsg
->msg_iocb
= NULL
;
1906 return import_iovec(save_addr
? READ
: WRITE
, uiov
, nr_segs
,
1907 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
1910 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
1911 struct msghdr
*msg_sys
, unsigned int flags
,
1912 struct used_address
*used_address
,
1913 unsigned int allowed_msghdr_flags
)
1915 struct compat_msghdr __user
*msg_compat
=
1916 (struct compat_msghdr __user
*)msg
;
1917 struct sockaddr_storage address
;
1918 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1919 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1920 __aligned(sizeof(__kernel_size_t
));
1921 /* 20 is size of ipv6_pktinfo */
1922 unsigned char *ctl_buf
= ctl
;
1926 msg_sys
->msg_name
= &address
;
1928 if (MSG_CMSG_COMPAT
& flags
)
1929 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
1931 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
1937 if (msg_sys
->msg_controllen
> INT_MAX
)
1939 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
1940 ctl_len
= msg_sys
->msg_controllen
;
1941 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1943 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
1947 ctl_buf
= msg_sys
->msg_control
;
1948 ctl_len
= msg_sys
->msg_controllen
;
1949 } else if (ctl_len
) {
1950 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
1951 CMSG_ALIGN(sizeof(struct cmsghdr
)));
1952 if (ctl_len
> sizeof(ctl
)) {
1953 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
1954 if (ctl_buf
== NULL
)
1959 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1960 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1961 * checking falls down on this.
1963 if (copy_from_user(ctl_buf
,
1964 (void __user __force
*)msg_sys
->msg_control
,
1967 msg_sys
->msg_control
= ctl_buf
;
1969 msg_sys
->msg_flags
= flags
;
1971 if (sock
->file
->f_flags
& O_NONBLOCK
)
1972 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
1974 * If this is sendmmsg() and current destination address is same as
1975 * previously succeeded address, omit asking LSM's decision.
1976 * used_address->name_len is initialized to UINT_MAX so that the first
1977 * destination address never matches.
1979 if (used_address
&& msg_sys
->msg_name
&&
1980 used_address
->name_len
== msg_sys
->msg_namelen
&&
1981 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
1982 used_address
->name_len
)) {
1983 err
= sock_sendmsg_nosec(sock
, msg_sys
);
1986 err
= sock_sendmsg(sock
, msg_sys
);
1988 * If this is sendmmsg() and sending to current destination address was
1989 * successful, remember it.
1991 if (used_address
&& err
>= 0) {
1992 used_address
->name_len
= msg_sys
->msg_namelen
;
1993 if (msg_sys
->msg_name
)
1994 memcpy(&used_address
->name
, msg_sys
->msg_name
,
1995 used_address
->name_len
);
2000 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2007 * BSD sendmsg interface
2010 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2012 int fput_needed
, err
;
2013 struct msghdr msg_sys
;
2014 struct socket
*sock
;
2016 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2020 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2022 fput_light(sock
->file
, fput_needed
);
2027 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2029 if (flags
& MSG_CMSG_COMPAT
)
2031 return __sys_sendmsg(fd
, msg
, flags
);
2035 * Linux sendmmsg interface
2038 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2041 int fput_needed
, err
, datagrams
;
2042 struct socket
*sock
;
2043 struct mmsghdr __user
*entry
;
2044 struct compat_mmsghdr __user
*compat_entry
;
2045 struct msghdr msg_sys
;
2046 struct used_address used_address
;
2047 unsigned int oflags
= flags
;
2049 if (vlen
> UIO_MAXIOV
)
2054 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2058 used_address
.name_len
= UINT_MAX
;
2060 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2064 while (datagrams
< vlen
) {
2065 if (datagrams
== vlen
- 1)
2068 if (MSG_CMSG_COMPAT
& flags
) {
2069 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2070 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2073 err
= __put_user(err
, &compat_entry
->msg_len
);
2076 err
= ___sys_sendmsg(sock
,
2077 (struct user_msghdr __user
*)entry
,
2078 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2081 err
= put_user(err
, &entry
->msg_len
);
2088 if (msg_data_left(&msg_sys
))
2093 fput_light(sock
->file
, fput_needed
);
2095 /* We only return an error if no datagrams were able to be sent */
2102 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2103 unsigned int, vlen
, unsigned int, flags
)
2105 if (flags
& MSG_CMSG_COMPAT
)
2107 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2110 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2111 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2113 struct compat_msghdr __user
*msg_compat
=
2114 (struct compat_msghdr __user
*)msg
;
2115 struct iovec iovstack
[UIO_FASTIOV
];
2116 struct iovec
*iov
= iovstack
;
2117 unsigned long cmsg_ptr
;
2121 /* kernel mode address */
2122 struct sockaddr_storage addr
;
2124 /* user mode address pointers */
2125 struct sockaddr __user
*uaddr
;
2126 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2128 msg_sys
->msg_name
= &addr
;
2130 if (MSG_CMSG_COMPAT
& flags
)
2131 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2133 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2137 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2138 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2140 /* We assume all kernel code knows the size of sockaddr_storage */
2141 msg_sys
->msg_namelen
= 0;
2143 if (sock
->file
->f_flags
& O_NONBLOCK
)
2144 flags
|= MSG_DONTWAIT
;
2145 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
, flags
);
2150 if (uaddr
!= NULL
) {
2151 err
= move_addr_to_user(&addr
,
2152 msg_sys
->msg_namelen
, uaddr
,
2157 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2161 if (MSG_CMSG_COMPAT
& flags
)
2162 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2163 &msg_compat
->msg_controllen
);
2165 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2166 &msg
->msg_controllen
);
2177 * BSD recvmsg interface
2180 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2182 int fput_needed
, err
;
2183 struct msghdr msg_sys
;
2184 struct socket
*sock
;
2186 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2190 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2192 fput_light(sock
->file
, fput_needed
);
2197 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2198 unsigned int, flags
)
2200 if (flags
& MSG_CMSG_COMPAT
)
2202 return __sys_recvmsg(fd
, msg
, flags
);
2206 * Linux recvmmsg interface
2209 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2210 unsigned int flags
, struct timespec
*timeout
)
2212 int fput_needed
, err
, datagrams
;
2213 struct socket
*sock
;
2214 struct mmsghdr __user
*entry
;
2215 struct compat_mmsghdr __user
*compat_entry
;
2216 struct msghdr msg_sys
;
2217 struct timespec64 end_time
;
2218 struct timespec64 timeout64
;
2221 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2227 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2231 err
= sock_error(sock
->sk
);
2238 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2240 while (datagrams
< vlen
) {
2242 * No need to ask LSM for more than the first datagram.
2244 if (MSG_CMSG_COMPAT
& flags
) {
2245 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2246 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2250 err
= __put_user(err
, &compat_entry
->msg_len
);
2253 err
= ___sys_recvmsg(sock
,
2254 (struct user_msghdr __user
*)entry
,
2255 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2259 err
= put_user(err
, &entry
->msg_len
);
2267 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2268 if (flags
& MSG_WAITFORONE
)
2269 flags
|= MSG_DONTWAIT
;
2272 ktime_get_ts64(&timeout64
);
2273 *timeout
= timespec64_to_timespec(
2274 timespec64_sub(end_time
, timeout64
));
2275 if (timeout
->tv_sec
< 0) {
2276 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2280 /* Timeout, return less than vlen datagrams */
2281 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2285 /* Out of band data, return right away */
2286 if (msg_sys
.msg_flags
& MSG_OOB
)
2294 if (datagrams
== 0) {
2300 * We may return less entries than requested (vlen) if the
2301 * sock is non block and there aren't enough datagrams...
2303 if (err
!= -EAGAIN
) {
2305 * ... or if recvmsg returns an error after we
2306 * received some datagrams, where we record the
2307 * error to return on the next call or if the
2308 * app asks about it using getsockopt(SO_ERROR).
2310 sock
->sk
->sk_err
= -err
;
2313 fput_light(sock
->file
, fput_needed
);
2318 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2319 unsigned int, vlen
, unsigned int, flags
,
2320 struct timespec __user
*, timeout
)
2323 struct timespec timeout_sys
;
2325 if (flags
& MSG_CMSG_COMPAT
)
2329 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2331 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2334 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2336 if (datagrams
> 0 &&
2337 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2338 datagrams
= -EFAULT
;
2343 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2344 /* Argument list sizes for sys_socketcall */
2345 #define AL(x) ((x) * sizeof(unsigned long))
2346 static const unsigned char nargs
[21] = {
2347 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2348 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2349 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2356 * System call vectors.
2358 * Argument checking cleaned up. Saved 20% in size.
2359 * This function doesn't need to set the kernel lock because
2360 * it is set by the callees.
2363 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2365 unsigned long a
[AUDITSC_ARGS
];
2366 unsigned long a0
, a1
;
2370 if (call
< 1 || call
> SYS_SENDMMSG
)
2374 if (len
> sizeof(a
))
2377 /* copy_from_user should be SMP safe. */
2378 if (copy_from_user(a
, args
, len
))
2381 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2390 err
= sys_socket(a0
, a1
, a
[2]);
2393 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2396 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2399 err
= sys_listen(a0
, a1
);
2402 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2403 (int __user
*)a
[2], 0);
2405 case SYS_GETSOCKNAME
:
2407 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2408 (int __user
*)a
[2]);
2410 case SYS_GETPEERNAME
:
2412 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2413 (int __user
*)a
[2]);
2415 case SYS_SOCKETPAIR
:
2416 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2419 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2422 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2423 (struct sockaddr __user
*)a
[4], a
[5]);
2426 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2429 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2430 (struct sockaddr __user
*)a
[4],
2431 (int __user
*)a
[5]);
2434 err
= sys_shutdown(a0
, a1
);
2436 case SYS_SETSOCKOPT
:
2437 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2439 case SYS_GETSOCKOPT
:
2441 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2442 (int __user
*)a
[4]);
2445 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2448 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2451 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2454 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2455 (struct timespec __user
*)a
[4]);
2458 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2459 (int __user
*)a
[2], a
[3]);
2468 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2471 * sock_register - add a socket protocol handler
2472 * @ops: description of protocol
2474 * This function is called by a protocol handler that wants to
2475 * advertise its address family, and have it linked into the
2476 * socket interface. The value ops->family corresponds to the
2477 * socket system call protocol family.
2479 int sock_register(const struct net_proto_family
*ops
)
2483 if (ops
->family
>= NPROTO
) {
2484 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2488 spin_lock(&net_family_lock
);
2489 if (rcu_dereference_protected(net_families
[ops
->family
],
2490 lockdep_is_held(&net_family_lock
)))
2493 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2496 spin_unlock(&net_family_lock
);
2498 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2501 EXPORT_SYMBOL(sock_register
);
2504 * sock_unregister - remove a protocol handler
2505 * @family: protocol family to remove
2507 * This function is called by a protocol handler that wants to
2508 * remove its address family, and have it unlinked from the
2509 * new socket creation.
2511 * If protocol handler is a module, then it can use module reference
2512 * counts to protect against new references. If protocol handler is not
2513 * a module then it needs to provide its own protection in
2514 * the ops->create routine.
2516 void sock_unregister(int family
)
2518 BUG_ON(family
< 0 || family
>= NPROTO
);
2520 spin_lock(&net_family_lock
);
2521 RCU_INIT_POINTER(net_families
[family
], NULL
);
2522 spin_unlock(&net_family_lock
);
2526 pr_info("NET: Unregistered protocol family %d\n", family
);
2528 EXPORT_SYMBOL(sock_unregister
);
2530 static int __init
sock_init(void)
2534 * Initialize the network sysctl infrastructure.
2536 err
= net_sysctl_init();
2541 * Initialize skbuff SLAB cache
2546 * Initialize the protocols module.
2551 err
= register_filesystem(&sock_fs_type
);
2554 sock_mnt
= kern_mount(&sock_fs_type
);
2555 if (IS_ERR(sock_mnt
)) {
2556 err
= PTR_ERR(sock_mnt
);
2560 /* The real protocol initialization is performed in later initcalls.
2563 #ifdef CONFIG_NETFILTER
2564 err
= netfilter_init();
2569 ptp_classifier_init();
2575 unregister_filesystem(&sock_fs_type
);
2580 core_initcall(sock_init
); /* early initcall */
2582 #ifdef CONFIG_PROC_FS
2583 void socket_seq_show(struct seq_file
*seq
)
2588 for_each_possible_cpu(cpu
)
2589 counter
+= per_cpu(sockets_in_use
, cpu
);
2591 /* It can be negative, by the way. 8) */
2595 seq_printf(seq
, "sockets: used %d\n", counter
);
2597 #endif /* CONFIG_PROC_FS */
2599 #ifdef CONFIG_COMPAT
2600 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2601 unsigned int cmd
, void __user
*up
)
2603 mm_segment_t old_fs
= get_fs();
2608 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2611 err
= compat_put_timeval(&ktv
, up
);
2616 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2617 unsigned int cmd
, void __user
*up
)
2619 mm_segment_t old_fs
= get_fs();
2620 struct timespec kts
;
2624 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2627 err
= compat_put_timespec(&kts
, up
);
2632 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2634 struct ifreq __user
*uifr
;
2637 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2638 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2641 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2645 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2651 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2653 struct compat_ifconf ifc32
;
2655 struct ifconf __user
*uifc
;
2656 struct compat_ifreq __user
*ifr32
;
2657 struct ifreq __user
*ifr
;
2661 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2664 memset(&ifc
, 0, sizeof(ifc
));
2665 if (ifc32
.ifcbuf
== 0) {
2669 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2671 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2672 sizeof(struct ifreq
);
2673 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2675 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2676 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2677 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2678 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2684 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2687 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2691 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2695 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2697 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2698 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2699 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2705 if (ifc32
.ifcbuf
== 0) {
2706 /* Translate from 64-bit structure multiple to
2710 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2715 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2721 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2723 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2724 bool convert_in
= false, convert_out
= false;
2725 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2726 struct ethtool_rxnfc __user
*rxnfc
;
2727 struct ifreq __user
*ifr
;
2728 u32 rule_cnt
= 0, actual_rule_cnt
;
2733 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2736 compat_rxnfc
= compat_ptr(data
);
2738 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2741 /* Most ethtool structures are defined without padding.
2742 * Unfortunately struct ethtool_rxnfc is an exception.
2747 case ETHTOOL_GRXCLSRLALL
:
2748 /* Buffer size is variable */
2749 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2751 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2753 buf_size
+= rule_cnt
* sizeof(u32
);
2755 case ETHTOOL_GRXRINGS
:
2756 case ETHTOOL_GRXCLSRLCNT
:
2757 case ETHTOOL_GRXCLSRULE
:
2758 case ETHTOOL_SRXCLSRLINS
:
2761 case ETHTOOL_SRXCLSRLDEL
:
2762 buf_size
+= sizeof(struct ethtool_rxnfc
);
2767 ifr
= compat_alloc_user_space(buf_size
);
2768 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2770 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2773 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2774 &ifr
->ifr_ifru
.ifru_data
))
2778 /* We expect there to be holes between fs.m_ext and
2779 * fs.ring_cookie and at the end of fs, but nowhere else.
2781 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2782 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2783 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2784 sizeof(rxnfc
->fs
.m_ext
));
2786 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2787 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2788 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2789 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2791 if (copy_in_user(rxnfc
, compat_rxnfc
,
2792 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2793 (void __user
*)rxnfc
) ||
2794 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2795 &compat_rxnfc
->fs
.ring_cookie
,
2796 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2797 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2798 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2799 sizeof(rxnfc
->rule_cnt
)))
2803 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2808 if (copy_in_user(compat_rxnfc
, rxnfc
,
2809 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2810 (const void __user
*)rxnfc
) ||
2811 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2812 &rxnfc
->fs
.ring_cookie
,
2813 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2814 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2815 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2816 sizeof(rxnfc
->rule_cnt
)))
2819 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2820 /* As an optimisation, we only copy the actual
2821 * number of rules that the underlying
2822 * function returned. Since Mallory might
2823 * change the rule count in user memory, we
2824 * check that it is less than the rule count
2825 * originally given (as the user buffer size),
2826 * which has been range-checked.
2828 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2830 if (actual_rule_cnt
< rule_cnt
)
2831 rule_cnt
= actual_rule_cnt
;
2832 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2833 &rxnfc
->rule_locs
[0],
2834 rule_cnt
* sizeof(u32
)))
2842 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2845 compat_uptr_t uptr32
;
2846 struct ifreq __user
*uifr
;
2848 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2849 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2852 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2855 uptr
= compat_ptr(uptr32
);
2857 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2860 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2863 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2864 struct compat_ifreq __user
*ifr32
)
2867 mm_segment_t old_fs
;
2871 case SIOCBONDENSLAVE
:
2872 case SIOCBONDRELEASE
:
2873 case SIOCBONDSETHWADDR
:
2874 case SIOCBONDCHANGEACTIVE
:
2875 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2880 err
= dev_ioctl(net
, cmd
,
2881 (struct ifreq __user __force
*) &kifr
);
2886 return -ENOIOCTLCMD
;
2890 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2891 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
2892 struct compat_ifreq __user
*u_ifreq32
)
2894 struct ifreq __user
*u_ifreq64
;
2895 char tmp_buf
[IFNAMSIZ
];
2896 void __user
*data64
;
2899 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2902 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2904 data64
= compat_ptr(data32
);
2906 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2908 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2911 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2914 return dev_ioctl(net
, cmd
, u_ifreq64
);
2917 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2918 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2920 struct ifreq __user
*uifr
;
2923 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2924 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2927 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
2938 case SIOCGIFBRDADDR
:
2939 case SIOCGIFDSTADDR
:
2940 case SIOCGIFNETMASK
:
2945 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
2953 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
2954 struct compat_ifreq __user
*uifr32
)
2957 struct compat_ifmap __user
*uifmap32
;
2958 mm_segment_t old_fs
;
2961 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
2962 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
2963 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2964 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2965 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2966 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2967 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2968 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2974 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
2977 if (cmd
== SIOCGIFMAP
&& !err
) {
2978 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
2979 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
2980 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
2981 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
2982 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
2983 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
2984 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
2993 struct sockaddr rt_dst
; /* target address */
2994 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
2995 struct sockaddr rt_genmask
; /* target network mask (IP) */
2996 unsigned short rt_flags
;
2999 unsigned char rt_tos
;
3000 unsigned char rt_class
;
3002 short rt_metric
; /* +1 for binary compatibility! */
3003 /* char * */ u32 rt_dev
; /* forcing the device at add */
3004 u32 rt_mtu
; /* per route MTU/Window */
3005 u32 rt_window
; /* Window clamping */
3006 unsigned short rt_irtt
; /* Initial RTT */
3009 struct in6_rtmsg32
{
3010 struct in6_addr rtmsg_dst
;
3011 struct in6_addr rtmsg_src
;
3012 struct in6_addr rtmsg_gateway
;
3022 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3023 unsigned int cmd
, void __user
*argp
)
3027 struct in6_rtmsg r6
;
3031 mm_segment_t old_fs
= get_fs();
3033 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3034 struct in6_rtmsg32 __user
*ur6
= argp
;
3035 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3036 3 * sizeof(struct in6_addr
));
3037 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3038 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3039 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3040 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3041 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3042 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3043 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3047 struct rtentry32 __user
*ur4
= argp
;
3048 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3049 3 * sizeof(struct sockaddr
));
3050 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3051 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3052 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3053 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3054 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3055 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3057 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3058 r4
.rt_dev
= (char __user __force
*)devname
;
3072 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3079 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3080 * for some operations; this forces use of the newer bridge-utils that
3081 * use compatible ioctls
3083 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3087 if (get_user(tmp
, argp
))
3089 if (tmp
== BRCTL_GET_VERSION
)
3090 return BRCTL_VERSION
+ 1;
3094 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3095 unsigned int cmd
, unsigned long arg
)
3097 void __user
*argp
= compat_ptr(arg
);
3098 struct sock
*sk
= sock
->sk
;
3099 struct net
*net
= sock_net(sk
);
3101 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3102 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3107 return old_bridge_ioctl(argp
);
3109 return dev_ifname32(net
, argp
);
3111 return dev_ifconf(net
, argp
);
3113 return ethtool_ioctl(net
, argp
);
3115 return compat_siocwandev(net
, argp
);
3118 return compat_sioc_ifmap(net
, cmd
, argp
);
3119 case SIOCBONDENSLAVE
:
3120 case SIOCBONDRELEASE
:
3121 case SIOCBONDSETHWADDR
:
3122 case SIOCBONDCHANGEACTIVE
:
3123 return bond_ioctl(net
, cmd
, argp
);
3126 return routing_ioctl(net
, sock
, cmd
, argp
);
3128 return do_siocgstamp(net
, sock
, cmd
, argp
);
3130 return do_siocgstampns(net
, sock
, cmd
, argp
);
3131 case SIOCBONDSLAVEINFOQUERY
:
3132 case SIOCBONDINFOQUERY
:
3135 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3148 return sock_ioctl(file
, cmd
, arg
);
3165 case SIOCSIFHWBROADCAST
:
3167 case SIOCGIFBRDADDR
:
3168 case SIOCSIFBRDADDR
:
3169 case SIOCGIFDSTADDR
:
3170 case SIOCSIFDSTADDR
:
3171 case SIOCGIFNETMASK
:
3172 case SIOCSIFNETMASK
:
3183 return dev_ifsioc(net
, sock
, cmd
, argp
);
3189 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3192 return -ENOIOCTLCMD
;
3195 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3198 struct socket
*sock
= file
->private_data
;
3199 int ret
= -ENOIOCTLCMD
;
3206 if (sock
->ops
->compat_ioctl
)
3207 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3209 if (ret
== -ENOIOCTLCMD
&&
3210 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3211 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3213 if (ret
== -ENOIOCTLCMD
)
3214 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3220 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3222 return sock
->ops
->bind(sock
, addr
, addrlen
);
3224 EXPORT_SYMBOL(kernel_bind
);
3226 int kernel_listen(struct socket
*sock
, int backlog
)
3228 return sock
->ops
->listen(sock
, backlog
);
3230 EXPORT_SYMBOL(kernel_listen
);
3232 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3234 struct sock
*sk
= sock
->sk
;
3237 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3242 err
= sock
->ops
->accept(sock
, *newsock
, flags
, true);
3244 sock_release(*newsock
);
3249 (*newsock
)->ops
= sock
->ops
;
3250 __module_get((*newsock
)->ops
->owner
);
3255 EXPORT_SYMBOL(kernel_accept
);
3257 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3260 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3262 EXPORT_SYMBOL(kernel_connect
);
3264 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3267 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3269 EXPORT_SYMBOL(kernel_getsockname
);
3271 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3274 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3276 EXPORT_SYMBOL(kernel_getpeername
);
3278 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3279 char *optval
, int *optlen
)
3281 mm_segment_t oldfs
= get_fs();
3282 char __user
*uoptval
;
3283 int __user
*uoptlen
;
3286 uoptval
= (char __user __force
*) optval
;
3287 uoptlen
= (int __user __force
*) optlen
;
3290 if (level
== SOL_SOCKET
)
3291 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3293 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3298 EXPORT_SYMBOL(kernel_getsockopt
);
3300 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3301 char *optval
, unsigned int optlen
)
3303 mm_segment_t oldfs
= get_fs();
3304 char __user
*uoptval
;
3307 uoptval
= (char __user __force
*) optval
;
3310 if (level
== SOL_SOCKET
)
3311 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3313 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3318 EXPORT_SYMBOL(kernel_setsockopt
);
3320 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3321 size_t size
, int flags
)
3323 if (sock
->ops
->sendpage
)
3324 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3326 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3328 EXPORT_SYMBOL(kernel_sendpage
);
3330 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3332 mm_segment_t oldfs
= get_fs();
3336 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3341 EXPORT_SYMBOL(kernel_sock_ioctl
);
3343 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3345 return sock
->ops
->shutdown(sock
, how
);
3347 EXPORT_SYMBOL(kernel_sock_shutdown
);