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
)) {
411 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 for a bit */
421 ihold(d_inode(path
.dentry
));
423 /* ... and now kill it properly */
429 file
->f_flags
= O_RDWR
| (flags
& O_NONBLOCK
);
430 file
->private_data
= sock
;
433 EXPORT_SYMBOL(sock_alloc_file
);
435 static int sock_map_fd(struct socket
*sock
, int flags
)
437 struct file
*newfile
;
438 int fd
= get_unused_fd_flags(flags
);
439 if (unlikely(fd
< 0)) {
444 newfile
= sock_alloc_file(sock
, flags
, NULL
);
445 if (likely(!IS_ERR(newfile
))) {
446 fd_install(fd
, newfile
);
451 return PTR_ERR(newfile
);
454 struct socket
*sock_from_file(struct file
*file
, int *err
)
456 if (file
->f_op
== &socket_file_ops
)
457 return file
->private_data
; /* set in sock_map_fd */
462 EXPORT_SYMBOL(sock_from_file
);
465 * sockfd_lookup - Go from a file number to its socket slot
467 * @err: pointer to an error code return
469 * The file handle passed in is locked and the socket it is bound
470 * to is returned. If an error occurs the err pointer is overwritten
471 * with a negative errno code and NULL is returned. The function checks
472 * for both invalid handles and passing a handle which is not a socket.
474 * On a success the socket object pointer is returned.
477 struct socket
*sockfd_lookup(int fd
, int *err
)
488 sock
= sock_from_file(file
, err
);
493 EXPORT_SYMBOL(sockfd_lookup
);
495 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
497 struct fd f
= fdget(fd
);
502 sock
= sock_from_file(f
.file
, err
);
504 *fput_needed
= f
.flags
;
512 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
518 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
528 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
533 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
540 static int sockfs_setattr(struct dentry
*dentry
, struct iattr
*iattr
)
542 int err
= simple_setattr(dentry
, iattr
);
544 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
545 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
548 sock
->sk
->sk_uid
= iattr
->ia_uid
;
556 static const struct inode_operations sockfs_inode_ops
= {
557 .listxattr
= sockfs_listxattr
,
558 .setattr
= sockfs_setattr
,
562 * sock_alloc - allocate a socket
564 * Allocate a new inode and socket object. The two are bound together
565 * and initialised. The socket is then returned. If we are out of inodes
569 struct socket
*sock_alloc(void)
574 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
578 sock
= SOCKET_I(inode
);
580 inode
->i_ino
= get_next_ino();
581 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
582 inode
->i_uid
= current_fsuid();
583 inode
->i_gid
= current_fsgid();
584 inode
->i_op
= &sockfs_inode_ops
;
586 this_cpu_add(sockets_in_use
, 1);
589 EXPORT_SYMBOL(sock_alloc
);
592 * sock_release - close a socket
593 * @sock: socket to close
595 * The socket is released from the protocol stack if it has a release
596 * callback, and the inode is then released if the socket is bound to
597 * an inode not a file.
600 static void __sock_release(struct socket
*sock
, struct inode
*inode
)
603 struct module
*owner
= sock
->ops
->owner
;
607 sock
->ops
->release(sock
);
614 if (rcu_dereference_protected(sock
->wq
, 1)->fasync_list
)
615 pr_err("%s: fasync list not empty!\n", __func__
);
617 this_cpu_sub(sockets_in_use
, 1);
619 iput(SOCK_INODE(sock
));
625 void sock_release(struct socket
*sock
)
627 __sock_release(sock
, NULL
);
629 EXPORT_SYMBOL(sock_release
);
631 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
633 u8 flags
= *tx_flags
;
635 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
636 flags
|= SKBTX_HW_TSTAMP
;
638 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
639 flags
|= SKBTX_SW_TSTAMP
;
641 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
642 flags
|= SKBTX_SCHED_TSTAMP
;
646 EXPORT_SYMBOL(__sock_tx_timestamp
);
648 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
650 int ret
= sock
->ops
->sendmsg(sock
, msg
, msg_data_left(msg
));
651 BUG_ON(ret
== -EIOCBQUEUED
);
655 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
657 int err
= security_socket_sendmsg(sock
, msg
,
660 return err
?: sock_sendmsg_nosec(sock
, msg
);
662 EXPORT_SYMBOL(sock_sendmsg
);
664 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
665 struct kvec
*vec
, size_t num
, size_t size
)
667 iov_iter_kvec(&msg
->msg_iter
, WRITE
| ITER_KVEC
, vec
, num
, size
);
668 return sock_sendmsg(sock
, msg
);
670 EXPORT_SYMBOL(kernel_sendmsg
);
672 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
673 struct kvec
*vec
, size_t num
, size_t size
)
675 struct socket
*sock
= sk
->sk_socket
;
677 if (!sock
->ops
->sendmsg_locked
)
678 return sock_no_sendmsg_locked(sk
, msg
, size
);
680 iov_iter_kvec(&msg
->msg_iter
, WRITE
| ITER_KVEC
, vec
, num
, size
);
682 return sock
->ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
684 EXPORT_SYMBOL(kernel_sendmsg_locked
);
686 static bool skb_is_err_queue(const struct sk_buff
*skb
)
688 /* pkt_type of skbs enqueued on the error queue are set to
689 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
690 * in recvmsg, since skbs received on a local socket will never
691 * have a pkt_type of PACKET_OUTGOING.
693 return skb
->pkt_type
== PACKET_OUTGOING
;
696 /* On transmit, software and hardware timestamps are returned independently.
697 * As the two skb clones share the hardware timestamp, which may be updated
698 * before the software timestamp is received, a hardware TX timestamp may be
699 * returned only if there is no software TX timestamp. Ignore false software
700 * timestamps, which may be made in the __sock_recv_timestamp() call when the
701 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
702 * hardware timestamp.
704 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
706 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
709 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
711 struct scm_ts_pktinfo ts_pktinfo
;
712 struct net_device
*orig_dev
;
714 if (!skb_mac_header_was_set(skb
))
717 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
720 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
722 ts_pktinfo
.if_index
= orig_dev
->ifindex
;
725 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
726 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
727 sizeof(ts_pktinfo
), &ts_pktinfo
);
731 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
733 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
736 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
737 struct scm_timestamping tss
;
738 int empty
= 1, false_tstamp
= 0;
739 struct skb_shared_hwtstamps
*shhwtstamps
=
742 /* Race occurred between timestamp enabling and packet
743 receiving. Fill in the current time for now. */
744 if (need_software_tstamp
&& skb
->tstamp
== 0) {
745 __net_timestamp(skb
);
749 if (need_software_tstamp
) {
750 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
752 skb_get_timestamp(skb
, &tv
);
753 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMP
,
757 skb_get_timestampns(skb
, &ts
);
758 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPNS
,
763 memset(&tss
, 0, sizeof(tss
));
764 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
765 ktime_to_timespec_cond(skb
->tstamp
, tss
.ts
+ 0))
768 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
769 !skb_is_swtx_tstamp(skb
, false_tstamp
) &&
770 ktime_to_timespec_cond(shhwtstamps
->hwtstamp
, tss
.ts
+ 2)) {
772 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
773 !skb_is_err_queue(skb
))
774 put_ts_pktinfo(msg
, skb
);
777 put_cmsg(msg
, SOL_SOCKET
,
778 SCM_TIMESTAMPING
, sizeof(tss
), &tss
);
780 if (skb_is_err_queue(skb
) && skb
->len
&&
781 SKB_EXT_ERR(skb
)->opt_stats
)
782 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
783 skb
->len
, skb
->data
);
786 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
788 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
793 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
795 if (!skb
->wifi_acked_valid
)
798 ack
= skb
->wifi_acked
;
800 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
802 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
804 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
807 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
808 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
809 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
812 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
815 sock_recv_timestamp(msg
, sk
, skb
);
816 sock_recv_drops(msg
, sk
, skb
);
818 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
820 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
823 return sock
->ops
->recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
826 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
828 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
830 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
832 EXPORT_SYMBOL(sock_recvmsg
);
835 * kernel_recvmsg - Receive a message from a socket (kernel space)
836 * @sock: The socket to receive the message from
837 * @msg: Received message
838 * @vec: Input s/g array for message data
839 * @num: Size of input s/g array
840 * @size: Number of bytes to read
841 * @flags: Message flags (MSG_DONTWAIT, etc...)
843 * On return the msg structure contains the scatter/gather array passed in the
844 * vec argument. The array is modified so that it consists of the unfilled
845 * portion of the original array.
847 * The returned value is the total number of bytes received, or an error.
849 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
850 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
852 mm_segment_t oldfs
= get_fs();
855 iov_iter_kvec(&msg
->msg_iter
, READ
| ITER_KVEC
, vec
, num
, size
);
857 result
= sock_recvmsg(sock
, msg
, flags
);
861 EXPORT_SYMBOL(kernel_recvmsg
);
863 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
864 int offset
, size_t size
, loff_t
*ppos
, int more
)
869 sock
= file
->private_data
;
871 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
872 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
875 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
878 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
879 struct pipe_inode_info
*pipe
, size_t len
,
882 struct socket
*sock
= file
->private_data
;
884 if (unlikely(!sock
->ops
->splice_read
))
887 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
890 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
892 struct file
*file
= iocb
->ki_filp
;
893 struct socket
*sock
= file
->private_data
;
894 struct msghdr msg
= {.msg_iter
= *to
,
898 if (file
->f_flags
& O_NONBLOCK
)
899 msg
.msg_flags
= MSG_DONTWAIT
;
901 if (iocb
->ki_pos
!= 0)
904 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
907 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
912 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
914 struct file
*file
= iocb
->ki_filp
;
915 struct socket
*sock
= file
->private_data
;
916 struct msghdr msg
= {.msg_iter
= *from
,
920 if (iocb
->ki_pos
!= 0)
923 if (file
->f_flags
& O_NONBLOCK
)
924 msg
.msg_flags
= MSG_DONTWAIT
;
926 if (sock
->type
== SOCK_SEQPACKET
)
927 msg
.msg_flags
|= MSG_EOR
;
929 res
= sock_sendmsg(sock
, &msg
);
930 *from
= msg
.msg_iter
;
935 * Atomic setting of ioctl hooks to avoid race
936 * with module unload.
939 static DEFINE_MUTEX(br_ioctl_mutex
);
940 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
942 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
944 mutex_lock(&br_ioctl_mutex
);
945 br_ioctl_hook
= hook
;
946 mutex_unlock(&br_ioctl_mutex
);
948 EXPORT_SYMBOL(brioctl_set
);
950 static DEFINE_MUTEX(vlan_ioctl_mutex
);
951 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
953 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
955 mutex_lock(&vlan_ioctl_mutex
);
956 vlan_ioctl_hook
= hook
;
957 mutex_unlock(&vlan_ioctl_mutex
);
959 EXPORT_SYMBOL(vlan_ioctl_set
);
961 static DEFINE_MUTEX(dlci_ioctl_mutex
);
962 static int (*dlci_ioctl_hook
) (unsigned int, void __user
*);
964 void dlci_ioctl_set(int (*hook
) (unsigned int, void __user
*))
966 mutex_lock(&dlci_ioctl_mutex
);
967 dlci_ioctl_hook
= hook
;
968 mutex_unlock(&dlci_ioctl_mutex
);
970 EXPORT_SYMBOL(dlci_ioctl_set
);
972 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
973 unsigned int cmd
, unsigned long arg
)
976 void __user
*argp
= (void __user
*)arg
;
978 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
981 * If this ioctl is unknown try to hand it down
984 if (err
== -ENOIOCTLCMD
)
985 err
= dev_ioctl(net
, cmd
, argp
);
991 * With an ioctl, arg may well be a user mode pointer, but we don't know
992 * what to do with it - that's up to the protocol still.
995 static struct ns_common
*get_net_ns(struct ns_common
*ns
)
997 return &get_net(container_of(ns
, struct net
, ns
))->ns
;
1000 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1002 struct socket
*sock
;
1004 void __user
*argp
= (void __user
*)arg
;
1008 sock
= file
->private_data
;
1011 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15)) {
1012 err
= dev_ioctl(net
, cmd
, argp
);
1014 #ifdef CONFIG_WEXT_CORE
1015 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1016 err
= dev_ioctl(net
, cmd
, argp
);
1023 if (get_user(pid
, (int __user
*)argp
))
1025 err
= f_setown(sock
->file
, pid
, 1);
1029 err
= put_user(f_getown(sock
->file
),
1030 (int __user
*)argp
);
1038 request_module("bridge");
1040 mutex_lock(&br_ioctl_mutex
);
1042 err
= br_ioctl_hook(net
, cmd
, argp
);
1043 mutex_unlock(&br_ioctl_mutex
);
1048 if (!vlan_ioctl_hook
)
1049 request_module("8021q");
1051 mutex_lock(&vlan_ioctl_mutex
);
1052 if (vlan_ioctl_hook
)
1053 err
= vlan_ioctl_hook(net
, argp
);
1054 mutex_unlock(&vlan_ioctl_mutex
);
1059 if (!dlci_ioctl_hook
)
1060 request_module("dlci");
1062 mutex_lock(&dlci_ioctl_mutex
);
1063 if (dlci_ioctl_hook
)
1064 err
= dlci_ioctl_hook(cmd
, argp
);
1065 mutex_unlock(&dlci_ioctl_mutex
);
1069 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1072 err
= open_related_ns(&net
->ns
, get_net_ns
);
1075 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1081 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1084 struct socket
*sock
= NULL
;
1086 err
= security_socket_create(family
, type
, protocol
, 1);
1090 sock
= sock_alloc();
1097 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1109 EXPORT_SYMBOL(sock_create_lite
);
1111 /* No kernel lock held - perfect */
1112 static unsigned int sock_poll(struct file
*file
, poll_table
*wait
)
1114 unsigned int busy_flag
= 0;
1115 struct socket
*sock
;
1118 * We can't return errors to poll, so it's either yes or no.
1120 sock
= file
->private_data
;
1122 if (sk_can_busy_loop(sock
->sk
)) {
1123 /* this socket can poll_ll so tell the system call */
1124 busy_flag
= POLL_BUSY_LOOP
;
1126 /* once, only if requested by syscall */
1127 if (wait
&& (wait
->_key
& POLL_BUSY_LOOP
))
1128 sk_busy_loop(sock
->sk
, 1);
1131 return busy_flag
| sock
->ops
->poll(file
, sock
, wait
);
1134 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1136 struct socket
*sock
= file
->private_data
;
1138 return sock
->ops
->mmap(file
, sock
, vma
);
1141 static int sock_close(struct inode
*inode
, struct file
*filp
)
1143 __sock_release(SOCKET_I(inode
), inode
);
1148 * Update the socket async list
1150 * Fasync_list locking strategy.
1152 * 1. fasync_list is modified only under process context socket lock
1153 * i.e. under semaphore.
1154 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1155 * or under socket lock
1158 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1160 struct socket
*sock
= filp
->private_data
;
1161 struct sock
*sk
= sock
->sk
;
1162 struct socket_wq
*wq
;
1168 wq
= rcu_dereference_protected(sock
->wq
, lockdep_sock_is_held(sk
));
1169 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1171 if (!wq
->fasync_list
)
1172 sock_reset_flag(sk
, SOCK_FASYNC
);
1174 sock_set_flag(sk
, SOCK_FASYNC
);
1180 /* This function may be called only under rcu_lock */
1182 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1184 if (!wq
|| !wq
->fasync_list
)
1188 case SOCK_WAKE_WAITD
:
1189 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1192 case SOCK_WAKE_SPACE
:
1193 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1198 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1201 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1206 EXPORT_SYMBOL(sock_wake_async
);
1208 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1209 struct socket
**res
, int kern
)
1212 struct socket
*sock
;
1213 const struct net_proto_family
*pf
;
1216 * Check protocol is in range
1218 if (family
< 0 || family
>= NPROTO
)
1219 return -EAFNOSUPPORT
;
1220 if (type
< 0 || type
>= SOCK_MAX
)
1225 This uglymoron is moved from INET layer to here to avoid
1226 deadlock in module load.
1228 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1229 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1234 err
= security_socket_create(family
, type
, protocol
, kern
);
1239 * Allocate the socket and allow the family to set things up. if
1240 * the protocol is 0, the family is instructed to select an appropriate
1243 sock
= sock_alloc();
1245 net_warn_ratelimited("socket: no more sockets\n");
1246 return -ENFILE
; /* Not exactly a match, but its the
1247 closest posix thing */
1252 #ifdef CONFIG_MODULES
1253 /* Attempt to load a protocol module if the find failed.
1255 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1256 * requested real, full-featured networking support upon configuration.
1257 * Otherwise module support will break!
1259 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1260 request_module("net-pf-%d", family
);
1264 pf
= rcu_dereference(net_families
[family
]);
1265 err
= -EAFNOSUPPORT
;
1270 * We will call the ->create function, that possibly is in a loadable
1271 * module, so we have to bump that loadable module refcnt first.
1273 if (!try_module_get(pf
->owner
))
1276 /* Now protected by module ref count */
1279 err
= pf
->create(net
, sock
, protocol
, kern
);
1281 goto out_module_put
;
1284 * Now to bump the refcnt of the [loadable] module that owns this
1285 * socket at sock_release time we decrement its refcnt.
1287 if (!try_module_get(sock
->ops
->owner
))
1288 goto out_module_busy
;
1291 * Now that we're done with the ->create function, the [loadable]
1292 * module can have its refcnt decremented
1294 module_put(pf
->owner
);
1295 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1297 goto out_sock_release
;
1303 err
= -EAFNOSUPPORT
;
1306 module_put(pf
->owner
);
1313 goto out_sock_release
;
1315 EXPORT_SYMBOL(__sock_create
);
1317 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1319 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1321 EXPORT_SYMBOL(sock_create
);
1323 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1325 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1327 EXPORT_SYMBOL(sock_create_kern
);
1329 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1332 struct socket
*sock
;
1335 /* Check the SOCK_* constants for consistency. */
1336 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1337 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1338 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1339 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1341 flags
= type
& ~SOCK_TYPE_MASK
;
1342 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1344 type
&= SOCK_TYPE_MASK
;
1346 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1347 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1349 retval
= sock_create(family
, type
, protocol
, &sock
);
1353 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1357 * Create a pair of connected sockets.
1360 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1361 int __user
*, usockvec
)
1363 struct socket
*sock1
, *sock2
;
1365 struct file
*newfile1
, *newfile2
;
1368 flags
= type
& ~SOCK_TYPE_MASK
;
1369 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1371 type
&= SOCK_TYPE_MASK
;
1373 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1374 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1377 * reserve descriptors and make sure we won't fail
1378 * to return them to userland.
1380 fd1
= get_unused_fd_flags(flags
);
1381 if (unlikely(fd1
< 0))
1384 fd2
= get_unused_fd_flags(flags
);
1385 if (unlikely(fd2
< 0)) {
1390 err
= put_user(fd1
, &usockvec
[0]);
1394 err
= put_user(fd2
, &usockvec
[1]);
1399 * Obtain the first socket and check if the underlying protocol
1400 * supports the socketpair call.
1403 err
= sock_create(family
, type
, protocol
, &sock1
);
1404 if (unlikely(err
< 0))
1407 err
= sock_create(family
, type
, protocol
, &sock2
);
1408 if (unlikely(err
< 0)) {
1409 sock_release(sock1
);
1413 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1414 if (unlikely(err
< 0)) {
1415 sock_release(sock2
);
1416 sock_release(sock1
);
1420 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1421 if (IS_ERR(newfile1
)) {
1422 err
= PTR_ERR(newfile1
);
1423 sock_release(sock2
);
1427 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1428 if (IS_ERR(newfile2
)) {
1429 err
= PTR_ERR(newfile2
);
1434 audit_fd_pair(fd1
, fd2
);
1436 fd_install(fd1
, newfile1
);
1437 fd_install(fd2
, newfile2
);
1447 * Bind a name to a socket. Nothing much to do here since it's
1448 * the protocol's responsibility to handle the local address.
1450 * We move the socket address to kernel space before we call
1451 * the protocol layer (having also checked the address is ok).
1454 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1456 struct socket
*sock
;
1457 struct sockaddr_storage address
;
1458 int err
, fput_needed
;
1460 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1462 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1464 err
= security_socket_bind(sock
,
1465 (struct sockaddr
*)&address
,
1468 err
= sock
->ops
->bind(sock
,
1472 fput_light(sock
->file
, fput_needed
);
1478 * Perform a listen. Basically, we allow the protocol to do anything
1479 * necessary for a listen, and if that works, we mark the socket as
1480 * ready for listening.
1483 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1485 struct socket
*sock
;
1486 int err
, fput_needed
;
1489 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1491 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1492 if ((unsigned int)backlog
> somaxconn
)
1493 backlog
= somaxconn
;
1495 err
= security_socket_listen(sock
, backlog
);
1497 err
= sock
->ops
->listen(sock
, backlog
);
1499 fput_light(sock
->file
, fput_needed
);
1505 * For accept, we attempt to create a new socket, set up the link
1506 * with the client, wake up the client, then return the new
1507 * connected fd. We collect the address of the connector in kernel
1508 * space and move it to user at the very end. This is unclean because
1509 * we open the socket then return an error.
1511 * 1003.1g adds the ability to recvmsg() to query connection pending
1512 * status to recvmsg. We need to add that support in a way thats
1513 * clean when we restucture accept also.
1516 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1517 int __user
*, upeer_addrlen
, int, flags
)
1519 struct socket
*sock
, *newsock
;
1520 struct file
*newfile
;
1521 int err
, len
, newfd
, fput_needed
;
1522 struct sockaddr_storage address
;
1524 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1527 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1528 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1530 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1535 newsock
= sock_alloc();
1539 newsock
->type
= sock
->type
;
1540 newsock
->ops
= sock
->ops
;
1543 * We don't need try_module_get here, as the listening socket (sock)
1544 * has the protocol module (sock->ops->owner) held.
1546 __module_get(newsock
->ops
->owner
);
1548 newfd
= get_unused_fd_flags(flags
);
1549 if (unlikely(newfd
< 0)) {
1551 sock_release(newsock
);
1554 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1555 if (IS_ERR(newfile
)) {
1556 err
= PTR_ERR(newfile
);
1557 put_unused_fd(newfd
);
1561 err
= security_socket_accept(sock
, newsock
);
1565 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
, false);
1569 if (upeer_sockaddr
) {
1570 if (newsock
->ops
->getname(newsock
, (struct sockaddr
*)&address
,
1572 err
= -ECONNABORTED
;
1575 err
= move_addr_to_user(&address
,
1576 len
, upeer_sockaddr
, upeer_addrlen
);
1581 /* File flags are not inherited via accept() unlike another OSes. */
1583 fd_install(newfd
, newfile
);
1587 fput_light(sock
->file
, fput_needed
);
1592 put_unused_fd(newfd
);
1596 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1597 int __user
*, upeer_addrlen
)
1599 return sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1603 * Attempt to connect to a socket with the server address. The address
1604 * is in user space so we verify it is OK and move it to kernel space.
1606 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1609 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1610 * other SEQPACKET protocols that take time to connect() as it doesn't
1611 * include the -EINPROGRESS status for such sockets.
1614 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1617 struct socket
*sock
;
1618 struct sockaddr_storage address
;
1619 int err
, fput_needed
;
1621 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1624 err
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1629 security_socket_connect(sock
, (struct sockaddr
*)&address
, addrlen
);
1633 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)&address
, addrlen
,
1634 sock
->file
->f_flags
);
1636 fput_light(sock
->file
, fput_needed
);
1642 * Get the local address ('name') of a socket object. Move the obtained
1643 * name to user space.
1646 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1647 int __user
*, usockaddr_len
)
1649 struct socket
*sock
;
1650 struct sockaddr_storage address
;
1651 int len
, err
, fput_needed
;
1653 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1657 err
= security_socket_getsockname(sock
);
1661 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
, 0);
1664 err
= move_addr_to_user(&address
, len
, usockaddr
, usockaddr_len
);
1667 fput_light(sock
->file
, fput_needed
);
1673 * Get the remote address ('name') of a socket object. Move the obtained
1674 * name to user space.
1677 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1678 int __user
*, usockaddr_len
)
1680 struct socket
*sock
;
1681 struct sockaddr_storage address
;
1682 int len
, err
, fput_needed
;
1684 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1686 err
= security_socket_getpeername(sock
);
1688 fput_light(sock
->file
, fput_needed
);
1693 sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, &len
,
1696 err
= move_addr_to_user(&address
, len
, usockaddr
,
1698 fput_light(sock
->file
, fput_needed
);
1704 * Send a datagram to a given address. We move the address into kernel
1705 * space and check the user space data area is readable before invoking
1709 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
1710 unsigned int, flags
, struct sockaddr __user
*, addr
,
1713 struct socket
*sock
;
1714 struct sockaddr_storage address
;
1720 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
1723 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1727 msg
.msg_name
= NULL
;
1728 msg
.msg_control
= NULL
;
1729 msg
.msg_controllen
= 0;
1730 msg
.msg_namelen
= 0;
1732 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
1735 msg
.msg_name
= (struct sockaddr
*)&address
;
1736 msg
.msg_namelen
= addr_len
;
1738 if (sock
->file
->f_flags
& O_NONBLOCK
)
1739 flags
|= MSG_DONTWAIT
;
1740 msg
.msg_flags
= flags
;
1741 err
= sock_sendmsg(sock
, &msg
);
1744 fput_light(sock
->file
, fput_needed
);
1750 * Send a datagram down a socket.
1753 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
1754 unsigned int, flags
)
1756 return sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
1760 * Receive a frame from the socket and optionally record the address of the
1761 * sender. We verify the buffers are writable and if needed move the
1762 * sender address from kernel to user space.
1765 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
1766 unsigned int, flags
, struct sockaddr __user
*, addr
,
1767 int __user
*, addr_len
)
1769 struct socket
*sock
;
1772 struct sockaddr_storage address
;
1776 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
1779 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1783 msg
.msg_control
= NULL
;
1784 msg
.msg_controllen
= 0;
1785 /* Save some cycles and don't copy the address if not needed */
1786 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
1787 /* We assume all kernel code knows the size of sockaddr_storage */
1788 msg
.msg_namelen
= 0;
1789 msg
.msg_iocb
= NULL
;
1791 if (sock
->file
->f_flags
& O_NONBLOCK
)
1792 flags
|= MSG_DONTWAIT
;
1793 err
= sock_recvmsg(sock
, &msg
, flags
);
1795 if (err
>= 0 && addr
!= NULL
) {
1796 err2
= move_addr_to_user(&address
,
1797 msg
.msg_namelen
, addr
, addr_len
);
1802 fput_light(sock
->file
, fput_needed
);
1808 * Receive a datagram from a socket.
1811 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
1812 unsigned int, flags
)
1814 return sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
1818 * Set a socket option. Because we don't know the option lengths we have
1819 * to pass the user mode parameter for the protocols to sort out.
1822 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
1823 char __user
*, optval
, int, optlen
)
1825 int err
, fput_needed
;
1826 struct socket
*sock
;
1831 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1833 err
= security_socket_setsockopt(sock
, level
, optname
);
1837 if (level
== SOL_SOCKET
)
1839 sock_setsockopt(sock
, level
, optname
, optval
,
1843 sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1846 fput_light(sock
->file
, fput_needed
);
1852 * Get a socket option. Because we don't know the option lengths we have
1853 * to pass a user mode parameter for the protocols to sort out.
1856 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
1857 char __user
*, optval
, int __user
*, optlen
)
1859 int err
, fput_needed
;
1860 struct socket
*sock
;
1862 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1864 err
= security_socket_getsockopt(sock
, level
, optname
);
1868 if (level
== SOL_SOCKET
)
1870 sock_getsockopt(sock
, level
, optname
, optval
,
1874 sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
1877 fput_light(sock
->file
, fput_needed
);
1883 * Shutdown a socket.
1886 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
1888 int err
, fput_needed
;
1889 struct socket
*sock
;
1891 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1893 err
= security_socket_shutdown(sock
, how
);
1895 err
= sock
->ops
->shutdown(sock
, how
);
1896 fput_light(sock
->file
, fput_needed
);
1901 /* A couple of helpful macros for getting the address of the 32/64 bit
1902 * fields which are the same type (int / unsigned) on our platforms.
1904 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1905 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1906 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1908 struct used_address
{
1909 struct sockaddr_storage name
;
1910 unsigned int name_len
;
1913 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
1914 struct user_msghdr __user
*umsg
,
1915 struct sockaddr __user
**save_addr
,
1918 struct user_msghdr msg
;
1921 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
1924 kmsg
->msg_control
= (void __force
*)msg
.msg_control
;
1925 kmsg
->msg_controllen
= msg
.msg_controllen
;
1926 kmsg
->msg_flags
= msg
.msg_flags
;
1928 kmsg
->msg_namelen
= msg
.msg_namelen
;
1930 kmsg
->msg_namelen
= 0;
1932 if (kmsg
->msg_namelen
< 0)
1935 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
1936 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
1939 *save_addr
= msg
.msg_name
;
1941 if (msg
.msg_name
&& kmsg
->msg_namelen
) {
1943 err
= move_addr_to_kernel(msg
.msg_name
,
1950 kmsg
->msg_name
= NULL
;
1951 kmsg
->msg_namelen
= 0;
1954 if (msg
.msg_iovlen
> UIO_MAXIOV
)
1957 kmsg
->msg_iocb
= NULL
;
1959 return import_iovec(save_addr
? READ
: WRITE
,
1960 msg
.msg_iov
, msg
.msg_iovlen
,
1961 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
1964 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
1965 struct msghdr
*msg_sys
, unsigned int flags
,
1966 struct used_address
*used_address
,
1967 unsigned int allowed_msghdr_flags
)
1969 struct compat_msghdr __user
*msg_compat
=
1970 (struct compat_msghdr __user
*)msg
;
1971 struct sockaddr_storage address
;
1972 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
1973 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
1974 __aligned(sizeof(__kernel_size_t
));
1975 /* 20 is size of ipv6_pktinfo */
1976 unsigned char *ctl_buf
= ctl
;
1980 msg_sys
->msg_name
= &address
;
1982 if (MSG_CMSG_COMPAT
& flags
)
1983 err
= get_compat_msghdr(msg_sys
, msg_compat
, NULL
, &iov
);
1985 err
= copy_msghdr_from_user(msg_sys
, msg
, NULL
, &iov
);
1991 if (msg_sys
->msg_controllen
> INT_MAX
)
1993 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
1994 ctl_len
= msg_sys
->msg_controllen
;
1995 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
1997 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2001 ctl_buf
= msg_sys
->msg_control
;
2002 ctl_len
= msg_sys
->msg_controllen
;
2003 } else if (ctl_len
) {
2004 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2005 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2006 if (ctl_len
> sizeof(ctl
)) {
2007 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2008 if (ctl_buf
== NULL
)
2013 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2014 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2015 * checking falls down on this.
2017 if (copy_from_user(ctl_buf
,
2018 (void __user __force
*)msg_sys
->msg_control
,
2021 msg_sys
->msg_control
= ctl_buf
;
2023 msg_sys
->msg_flags
= flags
;
2025 if (sock
->file
->f_flags
& O_NONBLOCK
)
2026 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2028 * If this is sendmmsg() and current destination address is same as
2029 * previously succeeded address, omit asking LSM's decision.
2030 * used_address->name_len is initialized to UINT_MAX so that the first
2031 * destination address never matches.
2033 if (used_address
&& msg_sys
->msg_name
&&
2034 used_address
->name_len
== msg_sys
->msg_namelen
&&
2035 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2036 used_address
->name_len
)) {
2037 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2040 err
= sock_sendmsg(sock
, msg_sys
);
2042 * If this is sendmmsg() and sending to current destination address was
2043 * successful, remember it.
2045 if (used_address
&& err
>= 0) {
2046 used_address
->name_len
= msg_sys
->msg_namelen
;
2047 if (msg_sys
->msg_name
)
2048 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2049 used_address
->name_len
);
2054 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2061 * BSD sendmsg interface
2064 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2066 int fput_needed
, err
;
2067 struct msghdr msg_sys
;
2068 struct socket
*sock
;
2070 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2074 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2076 fput_light(sock
->file
, fput_needed
);
2081 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2083 if (flags
& MSG_CMSG_COMPAT
)
2085 return __sys_sendmsg(fd
, msg
, flags
);
2089 * Linux sendmmsg interface
2092 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2095 int fput_needed
, err
, datagrams
;
2096 struct socket
*sock
;
2097 struct mmsghdr __user
*entry
;
2098 struct compat_mmsghdr __user
*compat_entry
;
2099 struct msghdr msg_sys
;
2100 struct used_address used_address
;
2101 unsigned int oflags
= flags
;
2103 if (vlen
> UIO_MAXIOV
)
2108 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2112 used_address
.name_len
= UINT_MAX
;
2114 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2118 while (datagrams
< vlen
) {
2119 if (datagrams
== vlen
- 1)
2122 if (MSG_CMSG_COMPAT
& flags
) {
2123 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2124 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2127 err
= __put_user(err
, &compat_entry
->msg_len
);
2130 err
= ___sys_sendmsg(sock
,
2131 (struct user_msghdr __user
*)entry
,
2132 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2135 err
= put_user(err
, &entry
->msg_len
);
2142 if (msg_data_left(&msg_sys
))
2147 fput_light(sock
->file
, fput_needed
);
2149 /* We only return an error if no datagrams were able to be sent */
2156 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2157 unsigned int, vlen
, unsigned int, flags
)
2159 if (flags
& MSG_CMSG_COMPAT
)
2161 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
);
2164 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2165 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2167 struct compat_msghdr __user
*msg_compat
=
2168 (struct compat_msghdr __user
*)msg
;
2169 struct iovec iovstack
[UIO_FASTIOV
];
2170 struct iovec
*iov
= iovstack
;
2171 unsigned long cmsg_ptr
;
2175 /* kernel mode address */
2176 struct sockaddr_storage addr
;
2178 /* user mode address pointers */
2179 struct sockaddr __user
*uaddr
;
2180 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2182 msg_sys
->msg_name
= &addr
;
2184 if (MSG_CMSG_COMPAT
& flags
)
2185 err
= get_compat_msghdr(msg_sys
, msg_compat
, &uaddr
, &iov
);
2187 err
= copy_msghdr_from_user(msg_sys
, msg
, &uaddr
, &iov
);
2191 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2192 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2194 /* We assume all kernel code knows the size of sockaddr_storage */
2195 msg_sys
->msg_namelen
= 0;
2197 if (sock
->file
->f_flags
& O_NONBLOCK
)
2198 flags
|= MSG_DONTWAIT
;
2199 err
= (nosec
? sock_recvmsg_nosec
: sock_recvmsg
)(sock
, msg_sys
, flags
);
2204 if (uaddr
!= NULL
) {
2205 err
= move_addr_to_user(&addr
,
2206 msg_sys
->msg_namelen
, uaddr
,
2211 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2215 if (MSG_CMSG_COMPAT
& flags
)
2216 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2217 &msg_compat
->msg_controllen
);
2219 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2220 &msg
->msg_controllen
);
2231 * BSD recvmsg interface
2234 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned flags
)
2236 int fput_needed
, err
;
2237 struct msghdr msg_sys
;
2238 struct socket
*sock
;
2240 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2244 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2246 fput_light(sock
->file
, fput_needed
);
2251 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2252 unsigned int, flags
)
2254 if (flags
& MSG_CMSG_COMPAT
)
2256 return __sys_recvmsg(fd
, msg
, flags
);
2260 * Linux recvmmsg interface
2263 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2264 unsigned int flags
, struct timespec
*timeout
)
2266 int fput_needed
, err
, datagrams
;
2267 struct socket
*sock
;
2268 struct mmsghdr __user
*entry
;
2269 struct compat_mmsghdr __user
*compat_entry
;
2270 struct msghdr msg_sys
;
2271 struct timespec64 end_time
;
2272 struct timespec64 timeout64
;
2275 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2281 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2285 err
= sock_error(sock
->sk
);
2292 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2294 while (datagrams
< vlen
) {
2296 * No need to ask LSM for more than the first datagram.
2298 if (MSG_CMSG_COMPAT
& flags
) {
2299 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2300 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2304 err
= __put_user(err
, &compat_entry
->msg_len
);
2307 err
= ___sys_recvmsg(sock
,
2308 (struct user_msghdr __user
*)entry
,
2309 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2313 err
= put_user(err
, &entry
->msg_len
);
2321 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2322 if (flags
& MSG_WAITFORONE
)
2323 flags
|= MSG_DONTWAIT
;
2326 ktime_get_ts64(&timeout64
);
2327 *timeout
= timespec64_to_timespec(
2328 timespec64_sub(end_time
, timeout64
));
2329 if (timeout
->tv_sec
< 0) {
2330 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2334 /* Timeout, return less than vlen datagrams */
2335 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2339 /* Out of band data, return right away */
2340 if (msg_sys
.msg_flags
& MSG_OOB
)
2348 if (datagrams
== 0) {
2354 * We may return less entries than requested (vlen) if the
2355 * sock is non block and there aren't enough datagrams...
2357 if (err
!= -EAGAIN
) {
2359 * ... or if recvmsg returns an error after we
2360 * received some datagrams, where we record the
2361 * error to return on the next call or if the
2362 * app asks about it using getsockopt(SO_ERROR).
2364 sock
->sk
->sk_err
= -err
;
2367 fput_light(sock
->file
, fput_needed
);
2372 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2373 unsigned int, vlen
, unsigned int, flags
,
2374 struct timespec __user
*, timeout
)
2377 struct timespec timeout_sys
;
2379 if (flags
& MSG_CMSG_COMPAT
)
2383 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2385 if (copy_from_user(&timeout_sys
, timeout
, sizeof(timeout_sys
)))
2388 datagrams
= __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2390 if (datagrams
> 0 &&
2391 copy_to_user(timeout
, &timeout_sys
, sizeof(timeout_sys
)))
2392 datagrams
= -EFAULT
;
2397 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2398 /* Argument list sizes for sys_socketcall */
2399 #define AL(x) ((x) * sizeof(unsigned long))
2400 static const unsigned char nargs
[21] = {
2401 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2402 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2403 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2410 * System call vectors.
2412 * Argument checking cleaned up. Saved 20% in size.
2413 * This function doesn't need to set the kernel lock because
2414 * it is set by the callees.
2417 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2419 unsigned long a
[AUDITSC_ARGS
];
2420 unsigned long a0
, a1
;
2424 if (call
< 1 || call
> SYS_SENDMMSG
)
2428 if (len
> sizeof(a
))
2431 /* copy_from_user should be SMP safe. */
2432 if (copy_from_user(a
, args
, len
))
2435 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2444 err
= sys_socket(a0
, a1
, a
[2]);
2447 err
= sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2450 err
= sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2453 err
= sys_listen(a0
, a1
);
2456 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2457 (int __user
*)a
[2], 0);
2459 case SYS_GETSOCKNAME
:
2461 sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2462 (int __user
*)a
[2]);
2464 case SYS_GETPEERNAME
:
2466 sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2467 (int __user
*)a
[2]);
2469 case SYS_SOCKETPAIR
:
2470 err
= sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2473 err
= sys_send(a0
, (void __user
*)a1
, a
[2], a
[3]);
2476 err
= sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2477 (struct sockaddr __user
*)a
[4], a
[5]);
2480 err
= sys_recv(a0
, (void __user
*)a1
, a
[2], a
[3]);
2483 err
= sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2484 (struct sockaddr __user
*)a
[4],
2485 (int __user
*)a
[5]);
2488 err
= sys_shutdown(a0
, a1
);
2490 case SYS_SETSOCKOPT
:
2491 err
= sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3], a
[4]);
2493 case SYS_GETSOCKOPT
:
2495 sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2496 (int __user
*)a
[4]);
2499 err
= sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2502 err
= sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3]);
2505 err
= sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
, a
[2]);
2508 err
= sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2], a
[3],
2509 (struct timespec __user
*)a
[4]);
2512 err
= sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2513 (int __user
*)a
[2], a
[3]);
2522 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2525 * sock_register - add a socket protocol handler
2526 * @ops: description of protocol
2528 * This function is called by a protocol handler that wants to
2529 * advertise its address family, and have it linked into the
2530 * socket interface. The value ops->family corresponds to the
2531 * socket system call protocol family.
2533 int sock_register(const struct net_proto_family
*ops
)
2537 if (ops
->family
>= NPROTO
) {
2538 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
2542 spin_lock(&net_family_lock
);
2543 if (rcu_dereference_protected(net_families
[ops
->family
],
2544 lockdep_is_held(&net_family_lock
)))
2547 rcu_assign_pointer(net_families
[ops
->family
], ops
);
2550 spin_unlock(&net_family_lock
);
2552 pr_info("NET: Registered protocol family %d\n", ops
->family
);
2555 EXPORT_SYMBOL(sock_register
);
2558 * sock_unregister - remove a protocol handler
2559 * @family: protocol family to remove
2561 * This function is called by a protocol handler that wants to
2562 * remove its address family, and have it unlinked from the
2563 * new socket creation.
2565 * If protocol handler is a module, then it can use module reference
2566 * counts to protect against new references. If protocol handler is not
2567 * a module then it needs to provide its own protection in
2568 * the ops->create routine.
2570 void sock_unregister(int family
)
2572 BUG_ON(family
< 0 || family
>= NPROTO
);
2574 spin_lock(&net_family_lock
);
2575 RCU_INIT_POINTER(net_families
[family
], NULL
);
2576 spin_unlock(&net_family_lock
);
2580 pr_info("NET: Unregistered protocol family %d\n", family
);
2582 EXPORT_SYMBOL(sock_unregister
);
2584 static int __init
sock_init(void)
2588 * Initialize the network sysctl infrastructure.
2590 err
= net_sysctl_init();
2595 * Initialize skbuff SLAB cache
2600 * Initialize the protocols module.
2605 err
= register_filesystem(&sock_fs_type
);
2608 sock_mnt
= kern_mount(&sock_fs_type
);
2609 if (IS_ERR(sock_mnt
)) {
2610 err
= PTR_ERR(sock_mnt
);
2614 /* The real protocol initialization is performed in later initcalls.
2617 #ifdef CONFIG_NETFILTER
2618 err
= netfilter_init();
2623 ptp_classifier_init();
2629 unregister_filesystem(&sock_fs_type
);
2634 core_initcall(sock_init
); /* early initcall */
2636 static int __init
jit_init(void)
2638 #ifdef CONFIG_BPF_JIT_ALWAYS_ON
2643 pure_initcall(jit_init
);
2645 #ifdef CONFIG_PROC_FS
2646 void socket_seq_show(struct seq_file
*seq
)
2651 for_each_possible_cpu(cpu
)
2652 counter
+= per_cpu(sockets_in_use
, cpu
);
2654 /* It can be negative, by the way. 8) */
2658 seq_printf(seq
, "sockets: used %d\n", counter
);
2660 #endif /* CONFIG_PROC_FS */
2662 #ifdef CONFIG_COMPAT
2663 static int do_siocgstamp(struct net
*net
, struct socket
*sock
,
2664 unsigned int cmd
, void __user
*up
)
2666 mm_segment_t old_fs
= get_fs();
2671 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&ktv
);
2674 err
= compat_put_timeval(&ktv
, up
);
2679 static int do_siocgstampns(struct net
*net
, struct socket
*sock
,
2680 unsigned int cmd
, void __user
*up
)
2682 mm_segment_t old_fs
= get_fs();
2683 struct timespec kts
;
2687 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)&kts
);
2690 err
= compat_put_timespec(&kts
, up
);
2695 static int dev_ifname32(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2697 struct ifreq __user
*uifr
;
2700 uifr
= compat_alloc_user_space(sizeof(struct ifreq
));
2701 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2704 err
= dev_ioctl(net
, SIOCGIFNAME
, uifr
);
2708 if (copy_in_user(uifr32
, uifr
, sizeof(struct compat_ifreq
)))
2714 static int dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
2716 struct compat_ifconf ifc32
;
2718 struct ifconf __user
*uifc
;
2719 struct compat_ifreq __user
*ifr32
;
2720 struct ifreq __user
*ifr
;
2724 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
2727 memset(&ifc
, 0, sizeof(ifc
));
2728 if (ifc32
.ifcbuf
== 0) {
2732 uifc
= compat_alloc_user_space(sizeof(struct ifconf
));
2734 size_t len
= ((ifc32
.ifc_len
/ sizeof(struct compat_ifreq
)) + 1) *
2735 sizeof(struct ifreq
);
2736 uifc
= compat_alloc_user_space(sizeof(struct ifconf
) + len
);
2738 ifr
= ifc
.ifc_req
= (void __user
*)(uifc
+ 1);
2739 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2740 for (i
= 0; i
< ifc32
.ifc_len
; i
+= sizeof(struct compat_ifreq
)) {
2741 if (copy_in_user(ifr
, ifr32
, sizeof(struct compat_ifreq
)))
2747 if (copy_to_user(uifc
, &ifc
, sizeof(struct ifconf
)))
2750 err
= dev_ioctl(net
, SIOCGIFCONF
, uifc
);
2754 if (copy_from_user(&ifc
, uifc
, sizeof(struct ifconf
)))
2758 ifr32
= compat_ptr(ifc32
.ifcbuf
);
2760 i
+ sizeof(struct compat_ifreq
) <= ifc32
.ifc_len
&& j
< ifc
.ifc_len
;
2761 i
+= sizeof(struct compat_ifreq
), j
+= sizeof(struct ifreq
)) {
2762 if (copy_in_user(ifr32
, ifr
, sizeof(struct compat_ifreq
)))
2768 if (ifc32
.ifcbuf
== 0) {
2769 /* Translate from 64-bit structure multiple to
2773 i
= ((i
/ sizeof(struct ifreq
)) * sizeof(struct compat_ifreq
));
2778 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
2784 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
2786 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
2787 bool convert_in
= false, convert_out
= false;
2788 size_t buf_size
= ALIGN(sizeof(struct ifreq
), 8);
2789 struct ethtool_rxnfc __user
*rxnfc
;
2790 struct ifreq __user
*ifr
;
2791 u32 rule_cnt
= 0, actual_rule_cnt
;
2796 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
2799 compat_rxnfc
= compat_ptr(data
);
2801 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
2804 /* Most ethtool structures are defined without padding.
2805 * Unfortunately struct ethtool_rxnfc is an exception.
2810 case ETHTOOL_GRXCLSRLALL
:
2811 /* Buffer size is variable */
2812 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
2814 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
2816 buf_size
+= rule_cnt
* sizeof(u32
);
2818 case ETHTOOL_GRXRINGS
:
2819 case ETHTOOL_GRXCLSRLCNT
:
2820 case ETHTOOL_GRXCLSRULE
:
2821 case ETHTOOL_SRXCLSRLINS
:
2824 case ETHTOOL_SRXCLSRLDEL
:
2825 buf_size
+= sizeof(struct ethtool_rxnfc
);
2830 ifr
= compat_alloc_user_space(buf_size
);
2831 rxnfc
= (void __user
*)ifr
+ ALIGN(sizeof(struct ifreq
), 8);
2833 if (copy_in_user(&ifr
->ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
2836 if (put_user(convert_in
? rxnfc
: compat_ptr(data
),
2837 &ifr
->ifr_ifru
.ifru_data
))
2841 /* We expect there to be holes between fs.m_ext and
2842 * fs.ring_cookie and at the end of fs, but nowhere else.
2844 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
2845 sizeof(compat_rxnfc
->fs
.m_ext
) !=
2846 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
2847 sizeof(rxnfc
->fs
.m_ext
));
2849 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
2850 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
2851 offsetof(struct ethtool_rxnfc
, fs
.location
) -
2852 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
2854 if (copy_in_user(rxnfc
, compat_rxnfc
,
2855 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2856 (void __user
*)rxnfc
) ||
2857 copy_in_user(&rxnfc
->fs
.ring_cookie
,
2858 &compat_rxnfc
->fs
.ring_cookie
,
2859 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
2860 (void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2861 copy_in_user(&rxnfc
->rule_cnt
, &compat_rxnfc
->rule_cnt
,
2862 sizeof(rxnfc
->rule_cnt
)))
2866 ret
= dev_ioctl(net
, SIOCETHTOOL
, ifr
);
2871 if (copy_in_user(compat_rxnfc
, rxnfc
,
2872 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
2873 (const void __user
*)rxnfc
) ||
2874 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
2875 &rxnfc
->fs
.ring_cookie
,
2876 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
2877 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
2878 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
2879 sizeof(rxnfc
->rule_cnt
)))
2882 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
2883 /* As an optimisation, we only copy the actual
2884 * number of rules that the underlying
2885 * function returned. Since Mallory might
2886 * change the rule count in user memory, we
2887 * check that it is less than the rule count
2888 * originally given (as the user buffer size),
2889 * which has been range-checked.
2891 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
2893 if (actual_rule_cnt
< rule_cnt
)
2894 rule_cnt
= actual_rule_cnt
;
2895 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
2896 &rxnfc
->rule_locs
[0],
2897 rule_cnt
* sizeof(u32
)))
2905 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
2908 compat_uptr_t uptr32
;
2909 struct ifreq __user
*uifr
;
2911 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2912 if (copy_in_user(uifr
, uifr32
, sizeof(struct compat_ifreq
)))
2915 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
2918 uptr
= compat_ptr(uptr32
);
2920 if (put_user(uptr
, &uifr
->ifr_settings
.ifs_ifsu
.raw_hdlc
))
2923 return dev_ioctl(net
, SIOCWANDEV
, uifr
);
2926 static int bond_ioctl(struct net
*net
, unsigned int cmd
,
2927 struct compat_ifreq __user
*ifr32
)
2930 mm_segment_t old_fs
;
2934 case SIOCBONDENSLAVE
:
2935 case SIOCBONDRELEASE
:
2936 case SIOCBONDSETHWADDR
:
2937 case SIOCBONDCHANGEACTIVE
:
2938 if (copy_from_user(&kifr
, ifr32
, sizeof(struct compat_ifreq
)))
2943 err
= dev_ioctl(net
, cmd
,
2944 (struct ifreq __user __force
*) &kifr
);
2949 return -ENOIOCTLCMD
;
2953 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2954 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
2955 struct compat_ifreq __user
*u_ifreq32
)
2957 struct ifreq __user
*u_ifreq64
;
2958 char tmp_buf
[IFNAMSIZ
];
2959 void __user
*data64
;
2962 if (copy_from_user(&tmp_buf
[0], &(u_ifreq32
->ifr_ifrn
.ifrn_name
[0]),
2965 if (get_user(data32
, &u_ifreq32
->ifr_ifru
.ifru_data
))
2967 data64
= compat_ptr(data32
);
2969 u_ifreq64
= compat_alloc_user_space(sizeof(*u_ifreq64
));
2971 if (copy_to_user(&u_ifreq64
->ifr_ifrn
.ifrn_name
[0], &tmp_buf
[0],
2974 if (put_user(data64
, &u_ifreq64
->ifr_ifru
.ifru_data
))
2977 return dev_ioctl(net
, cmd
, u_ifreq64
);
2980 static int dev_ifsioc(struct net
*net
, struct socket
*sock
,
2981 unsigned int cmd
, struct compat_ifreq __user
*uifr32
)
2983 struct ifreq __user
*uifr
;
2986 uifr
= compat_alloc_user_space(sizeof(*uifr
));
2987 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
2990 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3001 case SIOCGIFBRDADDR
:
3002 case SIOCGIFDSTADDR
:
3003 case SIOCGIFNETMASK
:
3008 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3016 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3017 struct compat_ifreq __user
*uifr32
)
3020 struct compat_ifmap __user
*uifmap32
;
3021 mm_segment_t old_fs
;
3024 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3025 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3026 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3027 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3028 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3029 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3030 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3031 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3037 err
= dev_ioctl(net
, cmd
, (void __user __force
*)&ifr
);
3040 if (cmd
== SIOCGIFMAP
&& !err
) {
3041 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3042 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3043 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3044 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3045 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3046 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3047 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3056 struct sockaddr rt_dst
; /* target address */
3057 struct sockaddr rt_gateway
; /* gateway addr (RTF_GATEWAY) */
3058 struct sockaddr rt_genmask
; /* target network mask (IP) */
3059 unsigned short rt_flags
;
3062 unsigned char rt_tos
;
3063 unsigned char rt_class
;
3065 short rt_metric
; /* +1 for binary compatibility! */
3066 /* char * */ u32 rt_dev
; /* forcing the device at add */
3067 u32 rt_mtu
; /* per route MTU/Window */
3068 u32 rt_window
; /* Window clamping */
3069 unsigned short rt_irtt
; /* Initial RTT */
3072 struct in6_rtmsg32
{
3073 struct in6_addr rtmsg_dst
;
3074 struct in6_addr rtmsg_src
;
3075 struct in6_addr rtmsg_gateway
;
3085 static int routing_ioctl(struct net
*net
, struct socket
*sock
,
3086 unsigned int cmd
, void __user
*argp
)
3090 struct in6_rtmsg r6
;
3094 mm_segment_t old_fs
= get_fs();
3096 if (sock
&& sock
->sk
&& sock
->sk
->sk_family
== AF_INET6
) { /* ipv6 */
3097 struct in6_rtmsg32 __user
*ur6
= argp
;
3098 ret
= copy_from_user(&r6
.rtmsg_dst
, &(ur6
->rtmsg_dst
),
3099 3 * sizeof(struct in6_addr
));
3100 ret
|= get_user(r6
.rtmsg_type
, &(ur6
->rtmsg_type
));
3101 ret
|= get_user(r6
.rtmsg_dst_len
, &(ur6
->rtmsg_dst_len
));
3102 ret
|= get_user(r6
.rtmsg_src_len
, &(ur6
->rtmsg_src_len
));
3103 ret
|= get_user(r6
.rtmsg_metric
, &(ur6
->rtmsg_metric
));
3104 ret
|= get_user(r6
.rtmsg_info
, &(ur6
->rtmsg_info
));
3105 ret
|= get_user(r6
.rtmsg_flags
, &(ur6
->rtmsg_flags
));
3106 ret
|= get_user(r6
.rtmsg_ifindex
, &(ur6
->rtmsg_ifindex
));
3110 struct rtentry32 __user
*ur4
= argp
;
3111 ret
= copy_from_user(&r4
.rt_dst
, &(ur4
->rt_dst
),
3112 3 * sizeof(struct sockaddr
));
3113 ret
|= get_user(r4
.rt_flags
, &(ur4
->rt_flags
));
3114 ret
|= get_user(r4
.rt_metric
, &(ur4
->rt_metric
));
3115 ret
|= get_user(r4
.rt_mtu
, &(ur4
->rt_mtu
));
3116 ret
|= get_user(r4
.rt_window
, &(ur4
->rt_window
));
3117 ret
|= get_user(r4
.rt_irtt
, &(ur4
->rt_irtt
));
3118 ret
|= get_user(rtdev
, &(ur4
->rt_dev
));
3120 ret
|= copy_from_user(devname
, compat_ptr(rtdev
), 15);
3121 r4
.rt_dev
= (char __user __force
*)devname
;
3135 ret
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long) r
);
3142 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3143 * for some operations; this forces use of the newer bridge-utils that
3144 * use compatible ioctls
3146 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3150 if (get_user(tmp
, argp
))
3152 if (tmp
== BRCTL_GET_VERSION
)
3153 return BRCTL_VERSION
+ 1;
3157 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3158 unsigned int cmd
, unsigned long arg
)
3160 void __user
*argp
= compat_ptr(arg
);
3161 struct sock
*sk
= sock
->sk
;
3162 struct net
*net
= sock_net(sk
);
3164 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3165 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3170 return old_bridge_ioctl(argp
);
3172 return dev_ifname32(net
, argp
);
3174 return dev_ifconf(net
, argp
);
3176 return ethtool_ioctl(net
, argp
);
3178 return compat_siocwandev(net
, argp
);
3181 return compat_sioc_ifmap(net
, cmd
, argp
);
3182 case SIOCBONDENSLAVE
:
3183 case SIOCBONDRELEASE
:
3184 case SIOCBONDSETHWADDR
:
3185 case SIOCBONDCHANGEACTIVE
:
3186 return bond_ioctl(net
, cmd
, argp
);
3189 return routing_ioctl(net
, sock
, cmd
, argp
);
3191 return do_siocgstamp(net
, sock
, cmd
, argp
);
3193 return do_siocgstampns(net
, sock
, cmd
, argp
);
3194 case SIOCBONDSLAVEINFOQUERY
:
3195 case SIOCBONDINFOQUERY
:
3198 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3211 return sock_ioctl(file
, cmd
, arg
);
3228 case SIOCSIFHWBROADCAST
:
3230 case SIOCGIFBRDADDR
:
3231 case SIOCSIFBRDADDR
:
3232 case SIOCGIFDSTADDR
:
3233 case SIOCSIFDSTADDR
:
3234 case SIOCGIFNETMASK
:
3235 case SIOCSIFNETMASK
:
3246 return dev_ifsioc(net
, sock
, cmd
, argp
);
3252 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3255 return -ENOIOCTLCMD
;
3258 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3261 struct socket
*sock
= file
->private_data
;
3262 int ret
= -ENOIOCTLCMD
;
3269 if (sock
->ops
->compat_ioctl
)
3270 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3272 if (ret
== -ENOIOCTLCMD
&&
3273 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3274 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3276 if (ret
== -ENOIOCTLCMD
)
3277 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3283 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3285 return sock
->ops
->bind(sock
, addr
, addrlen
);
3287 EXPORT_SYMBOL(kernel_bind
);
3289 int kernel_listen(struct socket
*sock
, int backlog
)
3291 return sock
->ops
->listen(sock
, backlog
);
3293 EXPORT_SYMBOL(kernel_listen
);
3295 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3297 struct sock
*sk
= sock
->sk
;
3300 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3305 err
= sock
->ops
->accept(sock
, *newsock
, flags
, true);
3307 sock_release(*newsock
);
3312 (*newsock
)->ops
= sock
->ops
;
3313 __module_get((*newsock
)->ops
->owner
);
3318 EXPORT_SYMBOL(kernel_accept
);
3320 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3323 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3325 EXPORT_SYMBOL(kernel_connect
);
3327 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
,
3330 return sock
->ops
->getname(sock
, addr
, addrlen
, 0);
3332 EXPORT_SYMBOL(kernel_getsockname
);
3334 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
,
3337 return sock
->ops
->getname(sock
, addr
, addrlen
, 1);
3339 EXPORT_SYMBOL(kernel_getpeername
);
3341 int kernel_getsockopt(struct socket
*sock
, int level
, int optname
,
3342 char *optval
, int *optlen
)
3344 mm_segment_t oldfs
= get_fs();
3345 char __user
*uoptval
;
3346 int __user
*uoptlen
;
3349 uoptval
= (char __user __force
*) optval
;
3350 uoptlen
= (int __user __force
*) optlen
;
3353 if (level
== SOL_SOCKET
)
3354 err
= sock_getsockopt(sock
, level
, optname
, uoptval
, uoptlen
);
3356 err
= sock
->ops
->getsockopt(sock
, level
, optname
, uoptval
,
3361 EXPORT_SYMBOL(kernel_getsockopt
);
3363 int kernel_setsockopt(struct socket
*sock
, int level
, int optname
,
3364 char *optval
, unsigned int optlen
)
3366 mm_segment_t oldfs
= get_fs();
3367 char __user
*uoptval
;
3370 uoptval
= (char __user __force
*) optval
;
3373 if (level
== SOL_SOCKET
)
3374 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
3376 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
3381 EXPORT_SYMBOL(kernel_setsockopt
);
3383 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3384 size_t size
, int flags
)
3386 if (sock
->ops
->sendpage
)
3387 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3389 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3391 EXPORT_SYMBOL(kernel_sendpage
);
3393 int kernel_sendpage_locked(struct sock
*sk
, struct page
*page
, int offset
,
3394 size_t size
, int flags
)
3396 struct socket
*sock
= sk
->sk_socket
;
3398 if (sock
->ops
->sendpage_locked
)
3399 return sock
->ops
->sendpage_locked(sk
, page
, offset
, size
,
3402 return sock_no_sendpage_locked(sk
, page
, offset
, size
, flags
);
3404 EXPORT_SYMBOL(kernel_sendpage_locked
);
3406 int kernel_sock_ioctl(struct socket
*sock
, int cmd
, unsigned long arg
)
3408 mm_segment_t oldfs
= get_fs();
3412 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
3417 EXPORT_SYMBOL(kernel_sock_ioctl
);
3419 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3421 return sock
->ops
->shutdown(sock
, how
);
3423 EXPORT_SYMBOL(kernel_sock_shutdown
);
3425 /* This routine returns the IP overhead imposed by a socket i.e.
3426 * the length of the underlying IP header, depending on whether
3427 * this is an IPv4 or IPv6 socket and the length from IP options turned
3428 * on at the socket. Assumes that the caller has a lock on the socket.
3430 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3432 struct inet_sock
*inet
;
3433 struct ip_options_rcu
*opt
;
3435 #if IS_ENABLED(CONFIG_IPV6)
3436 struct ipv6_pinfo
*np
;
3437 struct ipv6_txoptions
*optv6
= NULL
;
3438 #endif /* IS_ENABLED(CONFIG_IPV6) */
3443 switch (sk
->sk_family
) {
3446 overhead
+= sizeof(struct iphdr
);
3447 opt
= rcu_dereference_protected(inet
->inet_opt
,
3448 sock_owned_by_user(sk
));
3450 overhead
+= opt
->opt
.optlen
;
3452 #if IS_ENABLED(CONFIG_IPV6)
3455 overhead
+= sizeof(struct ipv6hdr
);
3457 optv6
= rcu_dereference_protected(np
->opt
,
3458 sock_owned_by_user(sk
));
3460 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3462 #endif /* IS_ENABLED(CONFIG_IPV6) */
3463 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3467 EXPORT_SYMBOL(kernel_sock_ip_overhead
);