1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
96 #include <net/cls_cgroup.h>
99 #include <linux/netfilter.h>
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly
;
112 unsigned int sysctl_net_busy_poll __read_mostly
;
115 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
);
116 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
);
117 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
119 static int sock_close(struct inode
*inode
, struct file
*file
);
120 static __poll_t
sock_poll(struct file
*file
,
121 struct poll_table_struct
*wait
);
122 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
124 static long compat_sock_ioctl(struct file
*file
,
125 unsigned int cmd
, unsigned long arg
);
127 static int sock_fasync(int fd
, struct file
*filp
, int on
);
128 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
129 int offset
, size_t size
, loff_t
*ppos
, int more
);
130 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
131 struct pipe_inode_info
*pipe
, size_t len
,
134 #ifdef CONFIG_PROC_FS
135 static void sock_show_fdinfo(struct seq_file
*m
, struct file
*f
)
137 struct socket
*sock
= f
->private_data
;
139 if (sock
->ops
->show_fdinfo
)
140 sock
->ops
->show_fdinfo(m
, sock
);
143 #define sock_show_fdinfo NULL
147 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
148 * in the operation structures but are done directly via the socketcall() multiplexor.
151 static const struct file_operations socket_file_ops
= {
152 .owner
= THIS_MODULE
,
154 .read_iter
= sock_read_iter
,
155 .write_iter
= sock_write_iter
,
157 .unlocked_ioctl
= sock_ioctl
,
159 .compat_ioctl
= compat_sock_ioctl
,
162 .release
= sock_close
,
163 .fasync
= sock_fasync
,
164 .sendpage
= sock_sendpage
,
165 .splice_write
= generic_splice_sendpage
,
166 .splice_read
= sock_splice_read
,
167 .show_fdinfo
= sock_show_fdinfo
,
170 static const char * const pf_family_names
[] = {
171 [PF_UNSPEC
] = "PF_UNSPEC",
172 [PF_UNIX
] = "PF_UNIX/PF_LOCAL",
173 [PF_INET
] = "PF_INET",
174 [PF_AX25
] = "PF_AX25",
176 [PF_APPLETALK
] = "PF_APPLETALK",
177 [PF_NETROM
] = "PF_NETROM",
178 [PF_BRIDGE
] = "PF_BRIDGE",
179 [PF_ATMPVC
] = "PF_ATMPVC",
181 [PF_INET6
] = "PF_INET6",
182 [PF_ROSE
] = "PF_ROSE",
183 [PF_DECnet
] = "PF_DECnet",
184 [PF_NETBEUI
] = "PF_NETBEUI",
185 [PF_SECURITY
] = "PF_SECURITY",
187 [PF_NETLINK
] = "PF_NETLINK/PF_ROUTE",
188 [PF_PACKET
] = "PF_PACKET",
190 [PF_ECONET
] = "PF_ECONET",
191 [PF_ATMSVC
] = "PF_ATMSVC",
194 [PF_IRDA
] = "PF_IRDA",
195 [PF_PPPOX
] = "PF_PPPOX",
196 [PF_WANPIPE
] = "PF_WANPIPE",
199 [PF_MPLS
] = "PF_MPLS",
201 [PF_TIPC
] = "PF_TIPC",
202 [PF_BLUETOOTH
] = "PF_BLUETOOTH",
203 [PF_IUCV
] = "PF_IUCV",
204 [PF_RXRPC
] = "PF_RXRPC",
205 [PF_ISDN
] = "PF_ISDN",
206 [PF_PHONET
] = "PF_PHONET",
207 [PF_IEEE802154
] = "PF_IEEE802154",
208 [PF_CAIF
] = "PF_CAIF",
211 [PF_VSOCK
] = "PF_VSOCK",
213 [PF_QIPCRTR
] = "PF_QIPCRTR",
216 [PF_MCTP
] = "PF_MCTP",
220 * The protocol list. Each protocol is registered in here.
223 static DEFINE_SPINLOCK(net_family_lock
);
224 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
228 * Move socket addresses back and forth across the kernel/user
229 * divide and look after the messy bits.
233 * move_addr_to_kernel - copy a socket address into kernel space
234 * @uaddr: Address in user space
235 * @kaddr: Address in kernel space
236 * @ulen: Length in user space
238 * The address is copied into kernel space. If the provided address is
239 * too long an error code of -EINVAL is returned. If the copy gives
240 * invalid addresses -EFAULT is returned. On a success 0 is returned.
243 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
245 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
249 if (copy_from_user(kaddr
, uaddr
, ulen
))
251 return audit_sockaddr(ulen
, kaddr
);
255 * move_addr_to_user - copy an address to user space
256 * @kaddr: kernel space address
257 * @klen: length of address in kernel
258 * @uaddr: user space address
259 * @ulen: pointer to user length field
261 * The value pointed to by ulen on entry is the buffer length available.
262 * This is overwritten with the buffer space used. -EINVAL is returned
263 * if an overlong buffer is specified or a negative buffer size. -EFAULT
264 * is returned if either the buffer or the length field are not
266 * After copying the data up to the limit the user specifies, the true
267 * length of the data is written over the length limit the user
268 * specified. Zero is returned for a success.
271 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
272 void __user
*uaddr
, int __user
*ulen
)
277 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
278 err
= get_user(len
, ulen
);
286 if (audit_sockaddr(klen
, kaddr
))
288 if (copy_to_user(uaddr
, kaddr
, len
))
292 * "fromlen shall refer to the value before truncation.."
295 return __put_user(klen
, ulen
);
298 static struct kmem_cache
*sock_inode_cachep __ro_after_init
;
300 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
302 struct socket_alloc
*ei
;
304 ei
= alloc_inode_sb(sb
, sock_inode_cachep
, GFP_KERNEL
);
307 init_waitqueue_head(&ei
->socket
.wq
.wait
);
308 ei
->socket
.wq
.fasync_list
= NULL
;
309 ei
->socket
.wq
.flags
= 0;
311 ei
->socket
.state
= SS_UNCONNECTED
;
312 ei
->socket
.flags
= 0;
313 ei
->socket
.ops
= NULL
;
314 ei
->socket
.sk
= NULL
;
315 ei
->socket
.file
= NULL
;
317 return &ei
->vfs_inode
;
320 static void sock_free_inode(struct inode
*inode
)
322 struct socket_alloc
*ei
;
324 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
325 kmem_cache_free(sock_inode_cachep
, ei
);
328 static void init_once(void *foo
)
330 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
332 inode_init_once(&ei
->vfs_inode
);
335 static void init_inodecache(void)
337 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
338 sizeof(struct socket_alloc
),
340 (SLAB_HWCACHE_ALIGN
|
341 SLAB_RECLAIM_ACCOUNT
|
342 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
344 BUG_ON(sock_inode_cachep
== NULL
);
347 static const struct super_operations sockfs_ops
= {
348 .alloc_inode
= sock_alloc_inode
,
349 .free_inode
= sock_free_inode
,
350 .statfs
= simple_statfs
,
354 * sockfs_dname() is called from d_path().
356 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
358 return dynamic_dname(buffer
, buflen
, "socket:[%lu]",
359 d_inode(dentry
)->i_ino
);
362 static const struct dentry_operations sockfs_dentry_operations
= {
363 .d_dname
= sockfs_dname
,
366 static int sockfs_xattr_get(const struct xattr_handler
*handler
,
367 struct dentry
*dentry
, struct inode
*inode
,
368 const char *suffix
, void *value
, size_t size
)
371 if (dentry
->d_name
.len
+ 1 > size
)
373 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
375 return dentry
->d_name
.len
+ 1;
378 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
379 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
380 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
382 static const struct xattr_handler sockfs_xattr_handler
= {
383 .name
= XATTR_NAME_SOCKPROTONAME
,
384 .get
= sockfs_xattr_get
,
387 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
388 struct user_namespace
*mnt_userns
,
389 struct dentry
*dentry
, struct inode
*inode
,
390 const char *suffix
, const void *value
,
391 size_t size
, int flags
)
393 /* Handled by LSM. */
397 static const struct xattr_handler sockfs_security_xattr_handler
= {
398 .prefix
= XATTR_SECURITY_PREFIX
,
399 .set
= sockfs_security_xattr_set
,
402 static const struct xattr_handler
*sockfs_xattr_handlers
[] = {
403 &sockfs_xattr_handler
,
404 &sockfs_security_xattr_handler
,
408 static int sockfs_init_fs_context(struct fs_context
*fc
)
410 struct pseudo_fs_context
*ctx
= init_pseudo(fc
, SOCKFS_MAGIC
);
413 ctx
->ops
= &sockfs_ops
;
414 ctx
->dops
= &sockfs_dentry_operations
;
415 ctx
->xattr
= sockfs_xattr_handlers
;
419 static struct vfsmount
*sock_mnt __read_mostly
;
421 static struct file_system_type sock_fs_type
= {
423 .init_fs_context
= sockfs_init_fs_context
,
424 .kill_sb
= kill_anon_super
,
428 * Obtains the first available file descriptor and sets it up for use.
430 * These functions create file structures and maps them to fd space
431 * of the current process. On success it returns file descriptor
432 * and file struct implicitly stored in sock->file.
433 * Note that another thread may close file descriptor before we return
434 * from this function. We use the fact that now we do not refer
435 * to socket after mapping. If one day we will need it, this
436 * function will increment ref. count on file by 1.
438 * In any case returned fd MAY BE not valid!
439 * This race condition is unavoidable
440 * with shared fd spaces, we cannot solve it inside kernel,
441 * but we take care of internal coherence yet.
445 * sock_alloc_file - Bind a &socket to a &file
447 * @flags: file status flags
448 * @dname: protocol name
450 * Returns the &file bound with @sock, implicitly storing it
451 * in sock->file. If dname is %NULL, sets to "".
452 * On failure the return is a ERR pointer (see linux/err.h).
453 * This function uses GFP_KERNEL internally.
456 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
461 dname
= sock
->sk
? sock
->sk
->sk_prot_creator
->name
: "";
463 file
= alloc_file_pseudo(SOCK_INODE(sock
), sock_mnt
, dname
,
464 O_RDWR
| (flags
& O_NONBLOCK
),
472 file
->private_data
= sock
;
473 stream_open(SOCK_INODE(sock
), file
);
476 EXPORT_SYMBOL(sock_alloc_file
);
478 static int sock_map_fd(struct socket
*sock
, int flags
)
480 struct file
*newfile
;
481 int fd
= get_unused_fd_flags(flags
);
482 if (unlikely(fd
< 0)) {
487 newfile
= sock_alloc_file(sock
, flags
, NULL
);
488 if (!IS_ERR(newfile
)) {
489 fd_install(fd
, newfile
);
494 return PTR_ERR(newfile
);
498 * sock_from_file - Return the &socket bounded to @file.
501 * On failure returns %NULL.
504 struct socket
*sock_from_file(struct file
*file
)
506 if (file
->f_op
== &socket_file_ops
)
507 return file
->private_data
; /* set in sock_alloc_file */
511 EXPORT_SYMBOL(sock_from_file
);
514 * sockfd_lookup - Go from a file number to its socket slot
516 * @err: pointer to an error code return
518 * The file handle passed in is locked and the socket it is bound
519 * to is returned. If an error occurs the err pointer is overwritten
520 * with a negative errno code and NULL is returned. The function checks
521 * for both invalid handles and passing a handle which is not a socket.
523 * On a success the socket object pointer is returned.
526 struct socket
*sockfd_lookup(int fd
, int *err
)
537 sock
= sock_from_file(file
);
544 EXPORT_SYMBOL(sockfd_lookup
);
546 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
548 struct fd f
= fdget(fd
);
553 sock
= sock_from_file(f
.file
);
555 *fput_needed
= f
.flags
& FDPUT_FPUT
;
564 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
570 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
580 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
585 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
592 static int sockfs_setattr(struct user_namespace
*mnt_userns
,
593 struct dentry
*dentry
, struct iattr
*iattr
)
595 int err
= simple_setattr(&init_user_ns
, dentry
, iattr
);
597 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
598 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
601 sock
->sk
->sk_uid
= iattr
->ia_uid
;
609 static const struct inode_operations sockfs_inode_ops
= {
610 .listxattr
= sockfs_listxattr
,
611 .setattr
= sockfs_setattr
,
615 * sock_alloc - allocate a socket
617 * Allocate a new inode and socket object. The two are bound together
618 * and initialised. The socket is then returned. If we are out of inodes
619 * NULL is returned. This functions uses GFP_KERNEL internally.
622 struct socket
*sock_alloc(void)
627 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
631 sock
= SOCKET_I(inode
);
633 inode
->i_ino
= get_next_ino();
634 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
635 inode
->i_uid
= current_fsuid();
636 inode
->i_gid
= current_fsgid();
637 inode
->i_op
= &sockfs_inode_ops
;
641 EXPORT_SYMBOL(sock_alloc
);
643 static void __sock_release(struct socket
*sock
, struct inode
*inode
)
646 struct module
*owner
= sock
->ops
->owner
;
650 sock
->ops
->release(sock
);
658 if (sock
->wq
.fasync_list
)
659 pr_err("%s: fasync list not empty!\n", __func__
);
662 iput(SOCK_INODE(sock
));
669 * sock_release - close a socket
670 * @sock: socket to close
672 * The socket is released from the protocol stack if it has a release
673 * callback, and the inode is then released if the socket is bound to
674 * an inode not a file.
676 void sock_release(struct socket
*sock
)
678 __sock_release(sock
, NULL
);
680 EXPORT_SYMBOL(sock_release
);
682 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
684 u8 flags
= *tx_flags
;
686 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
) {
687 flags
|= SKBTX_HW_TSTAMP
;
689 /* PTP hardware clocks can provide a free running cycle counter
690 * as a time base for virtual clocks. Tell driver to use the
691 * free running cycle counter for timestamp if socket is bound
694 if (tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
695 flags
|= SKBTX_HW_TSTAMP_USE_CYCLES
;
698 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
699 flags
|= SKBTX_SW_TSTAMP
;
701 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
702 flags
|= SKBTX_SCHED_TSTAMP
;
706 EXPORT_SYMBOL(__sock_tx_timestamp
);
708 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket
*, struct msghdr
*,
710 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket
*, struct msghdr
*,
712 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
714 int ret
= INDIRECT_CALL_INET(sock
->ops
->sendmsg
, inet6_sendmsg
,
715 inet_sendmsg
, sock
, msg
,
717 BUG_ON(ret
== -EIOCBQUEUED
);
722 * sock_sendmsg - send a message through @sock
724 * @msg: message to send
726 * Sends @msg through @sock, passing through LSM.
727 * Returns the number of bytes sent, or an error code.
729 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
731 int err
= security_socket_sendmsg(sock
, msg
,
734 return err
?: sock_sendmsg_nosec(sock
, msg
);
736 EXPORT_SYMBOL(sock_sendmsg
);
739 * kernel_sendmsg - send a message through @sock (kernel-space)
741 * @msg: message header
743 * @num: vec array length
744 * @size: total message data size
746 * Builds the message data with @vec and sends it through @sock.
747 * Returns the number of bytes sent, or an error code.
750 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
751 struct kvec
*vec
, size_t num
, size_t size
)
753 iov_iter_kvec(&msg
->msg_iter
, WRITE
, vec
, num
, size
);
754 return sock_sendmsg(sock
, msg
);
756 EXPORT_SYMBOL(kernel_sendmsg
);
759 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
761 * @msg: message header
762 * @vec: output s/g array
763 * @num: output s/g array length
764 * @size: total message data size
766 * Builds the message data with @vec and sends it through @sock.
767 * Returns the number of bytes sent, or an error code.
768 * Caller must hold @sk.
771 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
772 struct kvec
*vec
, size_t num
, size_t size
)
774 struct socket
*sock
= sk
->sk_socket
;
776 if (!sock
->ops
->sendmsg_locked
)
777 return sock_no_sendmsg_locked(sk
, msg
, size
);
779 iov_iter_kvec(&msg
->msg_iter
, WRITE
, vec
, num
, size
);
781 return sock
->ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
783 EXPORT_SYMBOL(kernel_sendmsg_locked
);
785 static bool skb_is_err_queue(const struct sk_buff
*skb
)
787 /* pkt_type of skbs enqueued on the error queue are set to
788 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
789 * in recvmsg, since skbs received on a local socket will never
790 * have a pkt_type of PACKET_OUTGOING.
792 return skb
->pkt_type
== PACKET_OUTGOING
;
795 /* On transmit, software and hardware timestamps are returned independently.
796 * As the two skb clones share the hardware timestamp, which may be updated
797 * before the software timestamp is received, a hardware TX timestamp may be
798 * returned only if there is no software TX timestamp. Ignore false software
799 * timestamps, which may be made in the __sock_recv_timestamp() call when the
800 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
801 * hardware timestamp.
803 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
805 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
808 static ktime_t
get_timestamp(struct sock
*sk
, struct sk_buff
*skb
, int *if_index
)
810 bool cycles
= sk
->sk_tsflags
& SOF_TIMESTAMPING_BIND_PHC
;
811 struct skb_shared_hwtstamps
*shhwtstamps
= skb_hwtstamps(skb
);
812 struct net_device
*orig_dev
;
816 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
818 *if_index
= orig_dev
->ifindex
;
819 hwtstamp
= netdev_get_tstamp(orig_dev
, shhwtstamps
, cycles
);
821 hwtstamp
= shhwtstamps
->hwtstamp
;
828 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
,
831 struct scm_ts_pktinfo ts_pktinfo
;
832 struct net_device
*orig_dev
;
834 if (!skb_mac_header_was_set(skb
))
837 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
841 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
843 if_index
= orig_dev
->ifindex
;
846 ts_pktinfo
.if_index
= if_index
;
848 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
849 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
850 sizeof(ts_pktinfo
), &ts_pktinfo
);
854 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
856 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
859 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
860 int new_tstamp
= sock_flag(sk
, SOCK_TSTAMP_NEW
);
861 struct scm_timestamping_internal tss
;
863 int empty
= 1, false_tstamp
= 0;
864 struct skb_shared_hwtstamps
*shhwtstamps
=
869 /* Race occurred between timestamp enabling and packet
870 receiving. Fill in the current time for now. */
871 if (need_software_tstamp
&& skb
->tstamp
== 0) {
872 __net_timestamp(skb
);
876 if (need_software_tstamp
) {
877 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
879 struct __kernel_sock_timeval tv
;
881 skb_get_new_timestamp(skb
, &tv
);
882 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_NEW
,
885 struct __kernel_old_timeval tv
;
887 skb_get_timestamp(skb
, &tv
);
888 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_OLD
,
893 struct __kernel_timespec ts
;
895 skb_get_new_timestampns(skb
, &ts
);
896 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_NEW
,
899 struct __kernel_old_timespec ts
;
901 skb_get_timestampns(skb
, &ts
);
902 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_OLD
,
908 memset(&tss
, 0, sizeof(tss
));
909 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
910 ktime_to_timespec64_cond(skb
->tstamp
, tss
.ts
+ 0))
913 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
914 !skb_is_swtx_tstamp(skb
, false_tstamp
)) {
916 if (skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP_NETDEV
)
917 hwtstamp
= get_timestamp(sk
, skb
, &if_index
);
919 hwtstamp
= shhwtstamps
->hwtstamp
;
921 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
922 hwtstamp
= ptp_convert_timestamp(&hwtstamp
,
925 if (ktime_to_timespec64_cond(hwtstamp
, tss
.ts
+ 2)) {
928 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
929 !skb_is_err_queue(skb
))
930 put_ts_pktinfo(msg
, skb
, if_index
);
934 if (sock_flag(sk
, SOCK_TSTAMP_NEW
))
935 put_cmsg_scm_timestamping64(msg
, &tss
);
937 put_cmsg_scm_timestamping(msg
, &tss
);
939 if (skb_is_err_queue(skb
) && skb
->len
&&
940 SKB_EXT_ERR(skb
)->opt_stats
)
941 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
942 skb
->len
, skb
->data
);
945 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
947 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
952 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
954 if (!skb
->wifi_acked_valid
)
957 ack
= skb
->wifi_acked
;
959 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
961 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
963 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
966 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
967 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
968 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
971 static void sock_recv_mark(struct msghdr
*msg
, struct sock
*sk
,
974 if (sock_flag(sk
, SOCK_RCVMARK
) && skb
)
975 put_cmsg(msg
, SOL_SOCKET
, SO_MARK
, sizeof(__u32
),
979 void __sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
982 sock_recv_timestamp(msg
, sk
, skb
);
983 sock_recv_drops(msg
, sk
, skb
);
984 sock_recv_mark(msg
, sk
, skb
);
986 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs
);
988 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket
*, struct msghdr
*,
990 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket
*, struct msghdr
*,
992 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
995 return INDIRECT_CALL_INET(sock
->ops
->recvmsg
, inet6_recvmsg
,
996 inet_recvmsg
, sock
, msg
, msg_data_left(msg
),
1001 * sock_recvmsg - receive a message from @sock
1003 * @msg: message to receive
1004 * @flags: message flags
1006 * Receives @msg from @sock, passing through LSM. Returns the total number
1007 * of bytes received, or an error.
1009 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
1011 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
1013 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
1015 EXPORT_SYMBOL(sock_recvmsg
);
1018 * kernel_recvmsg - Receive a message from a socket (kernel space)
1019 * @sock: The socket to receive the message from
1020 * @msg: Received message
1021 * @vec: Input s/g array for message data
1022 * @num: Size of input s/g array
1023 * @size: Number of bytes to read
1024 * @flags: Message flags (MSG_DONTWAIT, etc...)
1026 * On return the msg structure contains the scatter/gather array passed in the
1027 * vec argument. The array is modified so that it consists of the unfilled
1028 * portion of the original array.
1030 * The returned value is the total number of bytes received, or an error.
1033 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
1034 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
1036 msg
->msg_control_is_user
= false;
1037 iov_iter_kvec(&msg
->msg_iter
, READ
, vec
, num
, size
);
1038 return sock_recvmsg(sock
, msg
, flags
);
1040 EXPORT_SYMBOL(kernel_recvmsg
);
1042 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
1043 int offset
, size_t size
, loff_t
*ppos
, int more
)
1045 struct socket
*sock
;
1048 sock
= file
->private_data
;
1050 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
1051 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1054 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
1057 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
1058 struct pipe_inode_info
*pipe
, size_t len
,
1061 struct socket
*sock
= file
->private_data
;
1063 if (unlikely(!sock
->ops
->splice_read
))
1064 return generic_file_splice_read(file
, ppos
, pipe
, len
, flags
);
1066 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
1069 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1071 struct file
*file
= iocb
->ki_filp
;
1072 struct socket
*sock
= file
->private_data
;
1073 struct msghdr msg
= {.msg_iter
= *to
,
1077 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1078 msg
.msg_flags
= MSG_DONTWAIT
;
1080 if (iocb
->ki_pos
!= 0)
1083 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
1086 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
1091 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1093 struct file
*file
= iocb
->ki_filp
;
1094 struct socket
*sock
= file
->private_data
;
1095 struct msghdr msg
= {.msg_iter
= *from
,
1099 if (iocb
->ki_pos
!= 0)
1102 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1103 msg
.msg_flags
= MSG_DONTWAIT
;
1105 if (sock
->type
== SOCK_SEQPACKET
)
1106 msg
.msg_flags
|= MSG_EOR
;
1108 res
= sock_sendmsg(sock
, &msg
);
1109 *from
= msg
.msg_iter
;
1114 * Atomic setting of ioctl hooks to avoid race
1115 * with module unload.
1118 static DEFINE_MUTEX(br_ioctl_mutex
);
1119 static int (*br_ioctl_hook
)(struct net
*net
, struct net_bridge
*br
,
1120 unsigned int cmd
, struct ifreq
*ifr
,
1123 void brioctl_set(int (*hook
)(struct net
*net
, struct net_bridge
*br
,
1124 unsigned int cmd
, struct ifreq
*ifr
,
1127 mutex_lock(&br_ioctl_mutex
);
1128 br_ioctl_hook
= hook
;
1129 mutex_unlock(&br_ioctl_mutex
);
1131 EXPORT_SYMBOL(brioctl_set
);
1133 int br_ioctl_call(struct net
*net
, struct net_bridge
*br
, unsigned int cmd
,
1134 struct ifreq
*ifr
, void __user
*uarg
)
1139 request_module("bridge");
1141 mutex_lock(&br_ioctl_mutex
);
1143 err
= br_ioctl_hook(net
, br
, cmd
, ifr
, uarg
);
1144 mutex_unlock(&br_ioctl_mutex
);
1149 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1150 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1152 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1154 mutex_lock(&vlan_ioctl_mutex
);
1155 vlan_ioctl_hook
= hook
;
1156 mutex_unlock(&vlan_ioctl_mutex
);
1158 EXPORT_SYMBOL(vlan_ioctl_set
);
1160 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1161 unsigned int cmd
, unsigned long arg
)
1166 void __user
*argp
= (void __user
*)arg
;
1169 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1172 * If this ioctl is unknown try to hand it down
1173 * to the NIC driver.
1175 if (err
!= -ENOIOCTLCMD
)
1178 if (!is_socket_ioctl_cmd(cmd
))
1181 if (get_user_ifreq(&ifr
, &data
, argp
))
1183 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1184 if (!err
&& need_copyout
)
1185 if (put_user_ifreq(&ifr
, argp
))
1192 * With an ioctl, arg may well be a user mode pointer, but we don't know
1193 * what to do with it - that's up to the protocol still.
1196 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1198 struct socket
*sock
;
1200 void __user
*argp
= (void __user
*)arg
;
1204 sock
= file
->private_data
;
1207 if (unlikely(cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))) {
1211 if (get_user_ifreq(&ifr
, &data
, argp
))
1213 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1214 if (!err
&& need_copyout
)
1215 if (put_user_ifreq(&ifr
, argp
))
1218 #ifdef CONFIG_WEXT_CORE
1219 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1220 err
= wext_handle_ioctl(net
, cmd
, argp
);
1227 if (get_user(pid
, (int __user
*)argp
))
1229 err
= f_setown(sock
->file
, pid
, 1);
1233 err
= put_user(f_getown(sock
->file
),
1234 (int __user
*)argp
);
1240 err
= br_ioctl_call(net
, NULL
, cmd
, NULL
, argp
);
1245 if (!vlan_ioctl_hook
)
1246 request_module("8021q");
1248 mutex_lock(&vlan_ioctl_mutex
);
1249 if (vlan_ioctl_hook
)
1250 err
= vlan_ioctl_hook(net
, argp
);
1251 mutex_unlock(&vlan_ioctl_mutex
);
1255 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1258 err
= open_related_ns(&net
->ns
, get_net_ns
);
1260 case SIOCGSTAMP_OLD
:
1261 case SIOCGSTAMPNS_OLD
:
1262 if (!sock
->ops
->gettstamp
) {
1266 err
= sock
->ops
->gettstamp(sock
, argp
,
1267 cmd
== SIOCGSTAMP_OLD
,
1268 !IS_ENABLED(CONFIG_64BIT
));
1270 case SIOCGSTAMP_NEW
:
1271 case SIOCGSTAMPNS_NEW
:
1272 if (!sock
->ops
->gettstamp
) {
1276 err
= sock
->ops
->gettstamp(sock
, argp
,
1277 cmd
== SIOCGSTAMP_NEW
,
1282 err
= dev_ifconf(net
, argp
);
1286 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1293 * sock_create_lite - creates a socket
1294 * @family: protocol family (AF_INET, ...)
1295 * @type: communication type (SOCK_STREAM, ...)
1296 * @protocol: protocol (0, ...)
1299 * Creates a new socket and assigns it to @res, passing through LSM.
1300 * The new socket initialization is not complete, see kernel_accept().
1301 * Returns 0 or an error. On failure @res is set to %NULL.
1302 * This function internally uses GFP_KERNEL.
1305 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1308 struct socket
*sock
= NULL
;
1310 err
= security_socket_create(family
, type
, protocol
, 1);
1314 sock
= sock_alloc();
1321 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1333 EXPORT_SYMBOL(sock_create_lite
);
1335 /* No kernel lock held - perfect */
1336 static __poll_t
sock_poll(struct file
*file
, poll_table
*wait
)
1338 struct socket
*sock
= file
->private_data
;
1339 __poll_t events
= poll_requested_events(wait
), flag
= 0;
1341 if (!sock
->ops
->poll
)
1344 if (sk_can_busy_loop(sock
->sk
)) {
1345 /* poll once if requested by the syscall */
1346 if (events
& POLL_BUSY_LOOP
)
1347 sk_busy_loop(sock
->sk
, 1);
1349 /* if this socket can poll_ll, tell the system call */
1350 flag
= POLL_BUSY_LOOP
;
1353 return sock
->ops
->poll(file
, sock
, wait
) | flag
;
1356 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1358 struct socket
*sock
= file
->private_data
;
1360 return sock
->ops
->mmap(file
, sock
, vma
);
1363 static int sock_close(struct inode
*inode
, struct file
*filp
)
1365 __sock_release(SOCKET_I(inode
), inode
);
1370 * Update the socket async list
1372 * Fasync_list locking strategy.
1374 * 1. fasync_list is modified only under process context socket lock
1375 * i.e. under semaphore.
1376 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1377 * or under socket lock
1380 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1382 struct socket
*sock
= filp
->private_data
;
1383 struct sock
*sk
= sock
->sk
;
1384 struct socket_wq
*wq
= &sock
->wq
;
1390 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1392 if (!wq
->fasync_list
)
1393 sock_reset_flag(sk
, SOCK_FASYNC
);
1395 sock_set_flag(sk
, SOCK_FASYNC
);
1401 /* This function may be called only under rcu_lock */
1403 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1405 if (!wq
|| !wq
->fasync_list
)
1409 case SOCK_WAKE_WAITD
:
1410 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1413 case SOCK_WAKE_SPACE
:
1414 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1419 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1422 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1427 EXPORT_SYMBOL(sock_wake_async
);
1430 * __sock_create - creates a socket
1431 * @net: net namespace
1432 * @family: protocol family (AF_INET, ...)
1433 * @type: communication type (SOCK_STREAM, ...)
1434 * @protocol: protocol (0, ...)
1436 * @kern: boolean for kernel space sockets
1438 * Creates a new socket and assigns it to @res, passing through LSM.
1439 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1440 * be set to true if the socket resides in kernel space.
1441 * This function internally uses GFP_KERNEL.
1444 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1445 struct socket
**res
, int kern
)
1448 struct socket
*sock
;
1449 const struct net_proto_family
*pf
;
1452 * Check protocol is in range
1454 if (family
< 0 || family
>= NPROTO
)
1455 return -EAFNOSUPPORT
;
1456 if (type
< 0 || type
>= SOCK_MAX
)
1461 This uglymoron is moved from INET layer to here to avoid
1462 deadlock in module load.
1464 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1465 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1470 err
= security_socket_create(family
, type
, protocol
, kern
);
1475 * Allocate the socket and allow the family to set things up. if
1476 * the protocol is 0, the family is instructed to select an appropriate
1479 sock
= sock_alloc();
1481 net_warn_ratelimited("socket: no more sockets\n");
1482 return -ENFILE
; /* Not exactly a match, but its the
1483 closest posix thing */
1488 #ifdef CONFIG_MODULES
1489 /* Attempt to load a protocol module if the find failed.
1491 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1492 * requested real, full-featured networking support upon configuration.
1493 * Otherwise module support will break!
1495 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1496 request_module("net-pf-%d", family
);
1500 pf
= rcu_dereference(net_families
[family
]);
1501 err
= -EAFNOSUPPORT
;
1506 * We will call the ->create function, that possibly is in a loadable
1507 * module, so we have to bump that loadable module refcnt first.
1509 if (!try_module_get(pf
->owner
))
1512 /* Now protected by module ref count */
1515 err
= pf
->create(net
, sock
, protocol
, kern
);
1517 goto out_module_put
;
1520 * Now to bump the refcnt of the [loadable] module that owns this
1521 * socket at sock_release time we decrement its refcnt.
1523 if (!try_module_get(sock
->ops
->owner
))
1524 goto out_module_busy
;
1527 * Now that we're done with the ->create function, the [loadable]
1528 * module can have its refcnt decremented
1530 module_put(pf
->owner
);
1531 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1533 goto out_sock_release
;
1539 err
= -EAFNOSUPPORT
;
1542 module_put(pf
->owner
);
1549 goto out_sock_release
;
1551 EXPORT_SYMBOL(__sock_create
);
1554 * sock_create - creates a socket
1555 * @family: protocol family (AF_INET, ...)
1556 * @type: communication type (SOCK_STREAM, ...)
1557 * @protocol: protocol (0, ...)
1560 * A wrapper around __sock_create().
1561 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1564 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1566 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1568 EXPORT_SYMBOL(sock_create
);
1571 * sock_create_kern - creates a socket (kernel space)
1572 * @net: net namespace
1573 * @family: protocol family (AF_INET, ...)
1574 * @type: communication type (SOCK_STREAM, ...)
1575 * @protocol: protocol (0, ...)
1578 * A wrapper around __sock_create().
1579 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1582 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1584 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1586 EXPORT_SYMBOL(sock_create_kern
);
1588 static struct socket
*__sys_socket_create(int family
, int type
, int protocol
)
1590 struct socket
*sock
;
1593 /* Check the SOCK_* constants for consistency. */
1594 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1595 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1596 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1597 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1599 if ((type
& ~SOCK_TYPE_MASK
) & ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1600 return ERR_PTR(-EINVAL
);
1601 type
&= SOCK_TYPE_MASK
;
1603 retval
= sock_create(family
, type
, protocol
, &sock
);
1605 return ERR_PTR(retval
);
1610 struct file
*__sys_socket_file(int family
, int type
, int protocol
)
1612 struct socket
*sock
;
1616 sock
= __sys_socket_create(family
, type
, protocol
);
1618 return ERR_CAST(sock
);
1620 flags
= type
& ~SOCK_TYPE_MASK
;
1621 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1622 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1624 file
= sock_alloc_file(sock
, flags
, NULL
);
1631 int __sys_socket(int family
, int type
, int protocol
)
1633 struct socket
*sock
;
1636 sock
= __sys_socket_create(family
, type
, protocol
);
1638 return PTR_ERR(sock
);
1640 flags
= type
& ~SOCK_TYPE_MASK
;
1641 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1642 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1644 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1647 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1649 return __sys_socket(family
, type
, protocol
);
1653 * Create a pair of connected sockets.
1656 int __sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1658 struct socket
*sock1
, *sock2
;
1660 struct file
*newfile1
, *newfile2
;
1663 flags
= type
& ~SOCK_TYPE_MASK
;
1664 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1666 type
&= SOCK_TYPE_MASK
;
1668 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1669 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1672 * reserve descriptors and make sure we won't fail
1673 * to return them to userland.
1675 fd1
= get_unused_fd_flags(flags
);
1676 if (unlikely(fd1
< 0))
1679 fd2
= get_unused_fd_flags(flags
);
1680 if (unlikely(fd2
< 0)) {
1685 err
= put_user(fd1
, &usockvec
[0]);
1689 err
= put_user(fd2
, &usockvec
[1]);
1694 * Obtain the first socket and check if the underlying protocol
1695 * supports the socketpair call.
1698 err
= sock_create(family
, type
, protocol
, &sock1
);
1699 if (unlikely(err
< 0))
1702 err
= sock_create(family
, type
, protocol
, &sock2
);
1703 if (unlikely(err
< 0)) {
1704 sock_release(sock1
);
1708 err
= security_socket_socketpair(sock1
, sock2
);
1709 if (unlikely(err
)) {
1710 sock_release(sock2
);
1711 sock_release(sock1
);
1715 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1716 if (unlikely(err
< 0)) {
1717 sock_release(sock2
);
1718 sock_release(sock1
);
1722 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1723 if (IS_ERR(newfile1
)) {
1724 err
= PTR_ERR(newfile1
);
1725 sock_release(sock2
);
1729 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1730 if (IS_ERR(newfile2
)) {
1731 err
= PTR_ERR(newfile2
);
1736 audit_fd_pair(fd1
, fd2
);
1738 fd_install(fd1
, newfile1
);
1739 fd_install(fd2
, newfile2
);
1748 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1749 int __user
*, usockvec
)
1751 return __sys_socketpair(family
, type
, protocol
, usockvec
);
1755 * Bind a name to a socket. Nothing much to do here since it's
1756 * the protocol's responsibility to handle the local address.
1758 * We move the socket address to kernel space before we call
1759 * the protocol layer (having also checked the address is ok).
1762 int __sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1764 struct socket
*sock
;
1765 struct sockaddr_storage address
;
1766 int err
, fput_needed
;
1768 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1770 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1772 err
= security_socket_bind(sock
,
1773 (struct sockaddr
*)&address
,
1776 err
= sock
->ops
->bind(sock
,
1780 fput_light(sock
->file
, fput_needed
);
1785 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1787 return __sys_bind(fd
, umyaddr
, addrlen
);
1791 * Perform a listen. Basically, we allow the protocol to do anything
1792 * necessary for a listen, and if that works, we mark the socket as
1793 * ready for listening.
1796 int __sys_listen(int fd
, int backlog
)
1798 struct socket
*sock
;
1799 int err
, fput_needed
;
1802 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1804 somaxconn
= READ_ONCE(sock_net(sock
->sk
)->core
.sysctl_somaxconn
);
1805 if ((unsigned int)backlog
> somaxconn
)
1806 backlog
= somaxconn
;
1808 err
= security_socket_listen(sock
, backlog
);
1810 err
= sock
->ops
->listen(sock
, backlog
);
1812 fput_light(sock
->file
, fput_needed
);
1817 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1819 return __sys_listen(fd
, backlog
);
1822 struct file
*do_accept(struct file
*file
, unsigned file_flags
,
1823 struct sockaddr __user
*upeer_sockaddr
,
1824 int __user
*upeer_addrlen
, int flags
)
1826 struct socket
*sock
, *newsock
;
1827 struct file
*newfile
;
1829 struct sockaddr_storage address
;
1831 sock
= sock_from_file(file
);
1833 return ERR_PTR(-ENOTSOCK
);
1835 newsock
= sock_alloc();
1837 return ERR_PTR(-ENFILE
);
1839 newsock
->type
= sock
->type
;
1840 newsock
->ops
= sock
->ops
;
1843 * We don't need try_module_get here, as the listening socket (sock)
1844 * has the protocol module (sock->ops->owner) held.
1846 __module_get(newsock
->ops
->owner
);
1848 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1849 if (IS_ERR(newfile
))
1852 err
= security_socket_accept(sock
, newsock
);
1856 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
| file_flags
,
1861 if (upeer_sockaddr
) {
1862 len
= newsock
->ops
->getname(newsock
,
1863 (struct sockaddr
*)&address
, 2);
1865 err
= -ECONNABORTED
;
1868 err
= move_addr_to_user(&address
,
1869 len
, upeer_sockaddr
, upeer_addrlen
);
1874 /* File flags are not inherited via accept() unlike another OSes. */
1878 return ERR_PTR(err
);
1881 static int __sys_accept4_file(struct file
*file
, struct sockaddr __user
*upeer_sockaddr
,
1882 int __user
*upeer_addrlen
, int flags
)
1884 struct file
*newfile
;
1887 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1890 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1891 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1893 newfd
= get_unused_fd_flags(flags
);
1894 if (unlikely(newfd
< 0))
1897 newfile
= do_accept(file
, 0, upeer_sockaddr
, upeer_addrlen
,
1899 if (IS_ERR(newfile
)) {
1900 put_unused_fd(newfd
);
1901 return PTR_ERR(newfile
);
1903 fd_install(newfd
, newfile
);
1908 * For accept, we attempt to create a new socket, set up the link
1909 * with the client, wake up the client, then return the new
1910 * connected fd. We collect the address of the connector in kernel
1911 * space and move it to user at the very end. This is unclean because
1912 * we open the socket then return an error.
1914 * 1003.1g adds the ability to recvmsg() to query connection pending
1915 * status to recvmsg. We need to add that support in a way thats
1916 * clean when we restructure accept also.
1919 int __sys_accept4(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1920 int __user
*upeer_addrlen
, int flags
)
1927 ret
= __sys_accept4_file(f
.file
, upeer_sockaddr
,
1928 upeer_addrlen
, flags
);
1935 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1936 int __user
*, upeer_addrlen
, int, flags
)
1938 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, flags
);
1941 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1942 int __user
*, upeer_addrlen
)
1944 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1948 * Attempt to connect to a socket with the server address. The address
1949 * is in user space so we verify it is OK and move it to kernel space.
1951 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1954 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1955 * other SEQPACKET protocols that take time to connect() as it doesn't
1956 * include the -EINPROGRESS status for such sockets.
1959 int __sys_connect_file(struct file
*file
, struct sockaddr_storage
*address
,
1960 int addrlen
, int file_flags
)
1962 struct socket
*sock
;
1965 sock
= sock_from_file(file
);
1972 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1976 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
1977 sock
->file
->f_flags
| file_flags
);
1982 int __sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1989 struct sockaddr_storage address
;
1991 ret
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1993 ret
= __sys_connect_file(f
.file
, &address
, addrlen
, 0);
2000 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
2003 return __sys_connect(fd
, uservaddr
, addrlen
);
2007 * Get the local address ('name') of a socket object. Move the obtained
2008 * name to user space.
2011 int __sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
2012 int __user
*usockaddr_len
)
2014 struct socket
*sock
;
2015 struct sockaddr_storage address
;
2016 int err
, fput_needed
;
2018 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2022 err
= security_socket_getsockname(sock
);
2026 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 0);
2029 /* "err" is actually length in this case */
2030 err
= move_addr_to_user(&address
, err
, usockaddr
, usockaddr_len
);
2033 fput_light(sock
->file
, fput_needed
);
2038 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
2039 int __user
*, usockaddr_len
)
2041 return __sys_getsockname(fd
, usockaddr
, usockaddr_len
);
2045 * Get the remote address ('name') of a socket object. Move the obtained
2046 * name to user space.
2049 int __sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
2050 int __user
*usockaddr_len
)
2052 struct socket
*sock
;
2053 struct sockaddr_storage address
;
2054 int err
, fput_needed
;
2056 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2058 err
= security_socket_getpeername(sock
);
2060 fput_light(sock
->file
, fput_needed
);
2064 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 1);
2066 /* "err" is actually length in this case */
2067 err
= move_addr_to_user(&address
, err
, usockaddr
,
2069 fput_light(sock
->file
, fput_needed
);
2074 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
2075 int __user
*, usockaddr_len
)
2077 return __sys_getpeername(fd
, usockaddr
, usockaddr_len
);
2081 * Send a datagram to a given address. We move the address into kernel
2082 * space and check the user space data area is readable before invoking
2085 int __sys_sendto(int fd
, void __user
*buff
, size_t len
, unsigned int flags
,
2086 struct sockaddr __user
*addr
, int addr_len
)
2088 struct socket
*sock
;
2089 struct sockaddr_storage address
;
2095 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
2098 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2102 msg
.msg_name
= NULL
;
2103 msg
.msg_control
= NULL
;
2104 msg
.msg_controllen
= 0;
2105 msg
.msg_namelen
= 0;
2106 msg
.msg_ubuf
= NULL
;
2108 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
2111 msg
.msg_name
= (struct sockaddr
*)&address
;
2112 msg
.msg_namelen
= addr_len
;
2114 if (sock
->file
->f_flags
& O_NONBLOCK
)
2115 flags
|= MSG_DONTWAIT
;
2116 msg
.msg_flags
= flags
;
2117 err
= sock_sendmsg(sock
, &msg
);
2120 fput_light(sock
->file
, fput_needed
);
2125 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
2126 unsigned int, flags
, struct sockaddr __user
*, addr
,
2129 return __sys_sendto(fd
, buff
, len
, flags
, addr
, addr_len
);
2133 * Send a datagram down a socket.
2136 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
2137 unsigned int, flags
)
2139 return __sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
2143 * Receive a frame from the socket and optionally record the address of the
2144 * sender. We verify the buffers are writable and if needed move the
2145 * sender address from kernel to user space.
2147 int __sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
, unsigned int flags
,
2148 struct sockaddr __user
*addr
, int __user
*addr_len
)
2150 struct sockaddr_storage address
;
2151 struct msghdr msg
= {
2152 /* Save some cycles and don't copy the address if not needed */
2153 .msg_name
= addr
? (struct sockaddr
*)&address
: NULL
,
2155 struct socket
*sock
;
2160 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
2163 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2167 if (sock
->file
->f_flags
& O_NONBLOCK
)
2168 flags
|= MSG_DONTWAIT
;
2169 err
= sock_recvmsg(sock
, &msg
, flags
);
2171 if (err
>= 0 && addr
!= NULL
) {
2172 err2
= move_addr_to_user(&address
,
2173 msg
.msg_namelen
, addr
, addr_len
);
2178 fput_light(sock
->file
, fput_needed
);
2183 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
2184 unsigned int, flags
, struct sockaddr __user
*, addr
,
2185 int __user
*, addr_len
)
2187 return __sys_recvfrom(fd
, ubuf
, size
, flags
, addr
, addr_len
);
2191 * Receive a datagram from a socket.
2194 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
2195 unsigned int, flags
)
2197 return __sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
2200 static bool sock_use_custom_sol_socket(const struct socket
*sock
)
2202 const struct sock
*sk
= sock
->sk
;
2204 /* Use sock->ops->setsockopt() for MPTCP */
2205 return IS_ENABLED(CONFIG_MPTCP
) &&
2206 sk
->sk_protocol
== IPPROTO_MPTCP
&&
2207 sk
->sk_type
== SOCK_STREAM
&&
2208 (sk
->sk_family
== AF_INET
|| sk
->sk_family
== AF_INET6
);
2212 * Set a socket option. Because we don't know the option lengths we have
2213 * to pass the user mode parameter for the protocols to sort out.
2215 int __sys_setsockopt(int fd
, int level
, int optname
, char __user
*user_optval
,
2218 sockptr_t optval
= USER_SOCKPTR(user_optval
);
2219 char *kernel_optval
= NULL
;
2220 int err
, fput_needed
;
2221 struct socket
*sock
;
2226 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2230 err
= security_socket_setsockopt(sock
, level
, optname
);
2234 if (!in_compat_syscall())
2235 err
= BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock
->sk
, &level
, &optname
,
2236 user_optval
, &optlen
,
2246 optval
= KERNEL_SOCKPTR(kernel_optval
);
2247 if (level
== SOL_SOCKET
&& !sock_use_custom_sol_socket(sock
))
2248 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2249 else if (unlikely(!sock
->ops
->setsockopt
))
2252 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2254 kfree(kernel_optval
);
2256 fput_light(sock
->file
, fput_needed
);
2260 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
2261 char __user
*, optval
, int, optlen
)
2263 return __sys_setsockopt(fd
, level
, optname
, optval
, optlen
);
2266 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level
,
2270 * Get a socket option. Because we don't know the option lengths we have
2271 * to pass a user mode parameter for the protocols to sort out.
2273 int __sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
,
2276 int err
, fput_needed
;
2277 struct socket
*sock
;
2280 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2284 err
= security_socket_getsockopt(sock
, level
, optname
);
2288 if (!in_compat_syscall())
2289 max_optlen
= BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen
);
2291 if (level
== SOL_SOCKET
)
2292 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2293 else if (unlikely(!sock
->ops
->getsockopt
))
2296 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2299 if (!in_compat_syscall())
2300 err
= BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock
->sk
, level
, optname
,
2301 optval
, optlen
, max_optlen
,
2304 fput_light(sock
->file
, fput_needed
);
2308 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
2309 char __user
*, optval
, int __user
*, optlen
)
2311 return __sys_getsockopt(fd
, level
, optname
, optval
, optlen
);
2315 * Shutdown a socket.
2318 int __sys_shutdown_sock(struct socket
*sock
, int how
)
2322 err
= security_socket_shutdown(sock
, how
);
2324 err
= sock
->ops
->shutdown(sock
, how
);
2329 int __sys_shutdown(int fd
, int how
)
2331 int err
, fput_needed
;
2332 struct socket
*sock
;
2334 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2336 err
= __sys_shutdown_sock(sock
, how
);
2337 fput_light(sock
->file
, fput_needed
);
2342 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
2344 return __sys_shutdown(fd
, how
);
2347 /* A couple of helpful macros for getting the address of the 32/64 bit
2348 * fields which are the same type (int / unsigned) on our platforms.
2350 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2351 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2352 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2354 struct used_address
{
2355 struct sockaddr_storage name
;
2356 unsigned int name_len
;
2359 int __copy_msghdr(struct msghdr
*kmsg
,
2360 struct user_msghdr
*msg
,
2361 struct sockaddr __user
**save_addr
)
2365 kmsg
->msg_control_is_user
= true;
2366 kmsg
->msg_get_inq
= 0;
2367 kmsg
->msg_control_user
= msg
->msg_control
;
2368 kmsg
->msg_controllen
= msg
->msg_controllen
;
2369 kmsg
->msg_flags
= msg
->msg_flags
;
2371 kmsg
->msg_namelen
= msg
->msg_namelen
;
2373 kmsg
->msg_namelen
= 0;
2375 if (kmsg
->msg_namelen
< 0)
2378 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2379 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2382 *save_addr
= msg
->msg_name
;
2384 if (msg
->msg_name
&& kmsg
->msg_namelen
) {
2386 err
= move_addr_to_kernel(msg
->msg_name
,
2393 kmsg
->msg_name
= NULL
;
2394 kmsg
->msg_namelen
= 0;
2397 if (msg
->msg_iovlen
> UIO_MAXIOV
)
2400 kmsg
->msg_iocb
= NULL
;
2401 kmsg
->msg_ubuf
= NULL
;
2405 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
2406 struct user_msghdr __user
*umsg
,
2407 struct sockaddr __user
**save_addr
,
2410 struct user_msghdr msg
;
2413 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
2416 err
= __copy_msghdr(kmsg
, &msg
, save_addr
);
2420 err
= import_iovec(save_addr
? READ
: WRITE
,
2421 msg
.msg_iov
, msg
.msg_iovlen
,
2422 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
2423 return err
< 0 ? err
: 0;
2426 static int ____sys_sendmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2427 unsigned int flags
, struct used_address
*used_address
,
2428 unsigned int allowed_msghdr_flags
)
2430 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2431 __aligned(sizeof(__kernel_size_t
));
2432 /* 20 is size of ipv6_pktinfo */
2433 unsigned char *ctl_buf
= ctl
;
2439 if (msg_sys
->msg_controllen
> INT_MAX
)
2441 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
2442 ctl_len
= msg_sys
->msg_controllen
;
2443 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2445 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2449 ctl_buf
= msg_sys
->msg_control
;
2450 ctl_len
= msg_sys
->msg_controllen
;
2451 } else if (ctl_len
) {
2452 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2453 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2454 if (ctl_len
> sizeof(ctl
)) {
2455 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2456 if (ctl_buf
== NULL
)
2460 if (copy_from_user(ctl_buf
, msg_sys
->msg_control_user
, ctl_len
))
2462 msg_sys
->msg_control
= ctl_buf
;
2463 msg_sys
->msg_control_is_user
= false;
2465 msg_sys
->msg_flags
= flags
;
2467 if (sock
->file
->f_flags
& O_NONBLOCK
)
2468 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2470 * If this is sendmmsg() and current destination address is same as
2471 * previously succeeded address, omit asking LSM's decision.
2472 * used_address->name_len is initialized to UINT_MAX so that the first
2473 * destination address never matches.
2475 if (used_address
&& msg_sys
->msg_name
&&
2476 used_address
->name_len
== msg_sys
->msg_namelen
&&
2477 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2478 used_address
->name_len
)) {
2479 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2482 err
= sock_sendmsg(sock
, msg_sys
);
2484 * If this is sendmmsg() and sending to current destination address was
2485 * successful, remember it.
2487 if (used_address
&& err
>= 0) {
2488 used_address
->name_len
= msg_sys
->msg_namelen
;
2489 if (msg_sys
->msg_name
)
2490 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2491 used_address
->name_len
);
2496 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2501 int sendmsg_copy_msghdr(struct msghdr
*msg
,
2502 struct user_msghdr __user
*umsg
, unsigned flags
,
2507 if (flags
& MSG_CMSG_COMPAT
) {
2508 struct compat_msghdr __user
*msg_compat
;
2510 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2511 err
= get_compat_msghdr(msg
, msg_compat
, NULL
, iov
);
2513 err
= copy_msghdr_from_user(msg
, umsg
, NULL
, iov
);
2521 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2522 struct msghdr
*msg_sys
, unsigned int flags
,
2523 struct used_address
*used_address
,
2524 unsigned int allowed_msghdr_flags
)
2526 struct sockaddr_storage address
;
2527 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2530 msg_sys
->msg_name
= &address
;
2532 err
= sendmsg_copy_msghdr(msg_sys
, msg
, flags
, &iov
);
2536 err
= ____sys_sendmsg(sock
, msg_sys
, flags
, used_address
,
2537 allowed_msghdr_flags
);
2543 * BSD sendmsg interface
2545 long __sys_sendmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2548 return ____sys_sendmsg(sock
, msg
, flags
, NULL
, 0);
2551 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2552 bool forbid_cmsg_compat
)
2554 int fput_needed
, err
;
2555 struct msghdr msg_sys
;
2556 struct socket
*sock
;
2558 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2561 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2565 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2567 fput_light(sock
->file
, fput_needed
);
2572 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2574 return __sys_sendmsg(fd
, msg
, flags
, true);
2578 * Linux sendmmsg interface
2581 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2582 unsigned int flags
, bool forbid_cmsg_compat
)
2584 int fput_needed
, err
, datagrams
;
2585 struct socket
*sock
;
2586 struct mmsghdr __user
*entry
;
2587 struct compat_mmsghdr __user
*compat_entry
;
2588 struct msghdr msg_sys
;
2589 struct used_address used_address
;
2590 unsigned int oflags
= flags
;
2592 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2595 if (vlen
> UIO_MAXIOV
)
2600 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2604 used_address
.name_len
= UINT_MAX
;
2606 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2610 while (datagrams
< vlen
) {
2611 if (datagrams
== vlen
- 1)
2614 if (MSG_CMSG_COMPAT
& flags
) {
2615 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2616 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2619 err
= __put_user(err
, &compat_entry
->msg_len
);
2622 err
= ___sys_sendmsg(sock
,
2623 (struct user_msghdr __user
*)entry
,
2624 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2627 err
= put_user(err
, &entry
->msg_len
);
2634 if (msg_data_left(&msg_sys
))
2639 fput_light(sock
->file
, fput_needed
);
2641 /* We only return an error if no datagrams were able to be sent */
2648 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2649 unsigned int, vlen
, unsigned int, flags
)
2651 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
, true);
2654 int recvmsg_copy_msghdr(struct msghdr
*msg
,
2655 struct user_msghdr __user
*umsg
, unsigned flags
,
2656 struct sockaddr __user
**uaddr
,
2661 if (MSG_CMSG_COMPAT
& flags
) {
2662 struct compat_msghdr __user
*msg_compat
;
2664 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2665 err
= get_compat_msghdr(msg
, msg_compat
, uaddr
, iov
);
2667 err
= copy_msghdr_from_user(msg
, umsg
, uaddr
, iov
);
2675 static int ____sys_recvmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2676 struct user_msghdr __user
*msg
,
2677 struct sockaddr __user
*uaddr
,
2678 unsigned int flags
, int nosec
)
2680 struct compat_msghdr __user
*msg_compat
=
2681 (struct compat_msghdr __user
*) msg
;
2682 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2683 struct sockaddr_storage addr
;
2684 unsigned long cmsg_ptr
;
2688 msg_sys
->msg_name
= &addr
;
2689 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2690 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2692 /* We assume all kernel code knows the size of sockaddr_storage */
2693 msg_sys
->msg_namelen
= 0;
2695 if (sock
->file
->f_flags
& O_NONBLOCK
)
2696 flags
|= MSG_DONTWAIT
;
2698 if (unlikely(nosec
))
2699 err
= sock_recvmsg_nosec(sock
, msg_sys
, flags
);
2701 err
= sock_recvmsg(sock
, msg_sys
, flags
);
2707 if (uaddr
!= NULL
) {
2708 err
= move_addr_to_user(&addr
,
2709 msg_sys
->msg_namelen
, uaddr
,
2714 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2718 if (MSG_CMSG_COMPAT
& flags
)
2719 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2720 &msg_compat
->msg_controllen
);
2722 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2723 &msg
->msg_controllen
);
2731 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2732 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2734 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2735 /* user mode address pointers */
2736 struct sockaddr __user
*uaddr
;
2739 err
= recvmsg_copy_msghdr(msg_sys
, msg
, flags
, &uaddr
, &iov
);
2743 err
= ____sys_recvmsg(sock
, msg_sys
, msg
, uaddr
, flags
, nosec
);
2749 * BSD recvmsg interface
2752 long __sys_recvmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2753 struct user_msghdr __user
*umsg
,
2754 struct sockaddr __user
*uaddr
, unsigned int flags
)
2756 return ____sys_recvmsg(sock
, msg
, umsg
, uaddr
, flags
, 0);
2759 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2760 bool forbid_cmsg_compat
)
2762 int fput_needed
, err
;
2763 struct msghdr msg_sys
;
2764 struct socket
*sock
;
2766 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2769 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2773 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2775 fput_light(sock
->file
, fput_needed
);
2780 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2781 unsigned int, flags
)
2783 return __sys_recvmsg(fd
, msg
, flags
, true);
2787 * Linux recvmmsg interface
2790 static int do_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2791 unsigned int vlen
, unsigned int flags
,
2792 struct timespec64
*timeout
)
2794 int fput_needed
, err
, datagrams
;
2795 struct socket
*sock
;
2796 struct mmsghdr __user
*entry
;
2797 struct compat_mmsghdr __user
*compat_entry
;
2798 struct msghdr msg_sys
;
2799 struct timespec64 end_time
;
2800 struct timespec64 timeout64
;
2803 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2809 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2813 if (likely(!(flags
& MSG_ERRQUEUE
))) {
2814 err
= sock_error(sock
->sk
);
2822 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2824 while (datagrams
< vlen
) {
2826 * No need to ask LSM for more than the first datagram.
2828 if (MSG_CMSG_COMPAT
& flags
) {
2829 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2830 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2834 err
= __put_user(err
, &compat_entry
->msg_len
);
2837 err
= ___sys_recvmsg(sock
,
2838 (struct user_msghdr __user
*)entry
,
2839 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2843 err
= put_user(err
, &entry
->msg_len
);
2851 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2852 if (flags
& MSG_WAITFORONE
)
2853 flags
|= MSG_DONTWAIT
;
2856 ktime_get_ts64(&timeout64
);
2857 *timeout
= timespec64_sub(end_time
, timeout64
);
2858 if (timeout
->tv_sec
< 0) {
2859 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2863 /* Timeout, return less than vlen datagrams */
2864 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2868 /* Out of band data, return right away */
2869 if (msg_sys
.msg_flags
& MSG_OOB
)
2877 if (datagrams
== 0) {
2883 * We may return less entries than requested (vlen) if the
2884 * sock is non block and there aren't enough datagrams...
2886 if (err
!= -EAGAIN
) {
2888 * ... or if recvmsg returns an error after we
2889 * received some datagrams, where we record the
2890 * error to return on the next call or if the
2891 * app asks about it using getsockopt(SO_ERROR).
2893 sock
->sk
->sk_err
= -err
;
2896 fput_light(sock
->file
, fput_needed
);
2901 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2902 unsigned int vlen
, unsigned int flags
,
2903 struct __kernel_timespec __user
*timeout
,
2904 struct old_timespec32 __user
*timeout32
)
2907 struct timespec64 timeout_sys
;
2909 if (timeout
&& get_timespec64(&timeout_sys
, timeout
))
2912 if (timeout32
&& get_old_timespec32(&timeout_sys
, timeout32
))
2915 if (!timeout
&& !timeout32
)
2916 return do_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2918 datagrams
= do_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2923 if (timeout
&& put_timespec64(&timeout_sys
, timeout
))
2924 datagrams
= -EFAULT
;
2926 if (timeout32
&& put_old_timespec32(&timeout_sys
, timeout32
))
2927 datagrams
= -EFAULT
;
2932 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2933 unsigned int, vlen
, unsigned int, flags
,
2934 struct __kernel_timespec __user
*, timeout
)
2936 if (flags
& MSG_CMSG_COMPAT
)
2939 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, timeout
, NULL
);
2942 #ifdef CONFIG_COMPAT_32BIT_TIME
2943 SYSCALL_DEFINE5(recvmmsg_time32
, int, fd
, struct mmsghdr __user
*, mmsg
,
2944 unsigned int, vlen
, unsigned int, flags
,
2945 struct old_timespec32 __user
*, timeout
)
2947 if (flags
& MSG_CMSG_COMPAT
)
2950 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
, timeout
);
2954 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2955 /* Argument list sizes for sys_socketcall */
2956 #define AL(x) ((x) * sizeof(unsigned long))
2957 static const unsigned char nargs
[21] = {
2958 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2959 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2960 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2967 * System call vectors.
2969 * Argument checking cleaned up. Saved 20% in size.
2970 * This function doesn't need to set the kernel lock because
2971 * it is set by the callees.
2974 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2976 unsigned long a
[AUDITSC_ARGS
];
2977 unsigned long a0
, a1
;
2981 if (call
< 1 || call
> SYS_SENDMMSG
)
2983 call
= array_index_nospec(call
, SYS_SENDMMSG
+ 1);
2986 if (len
> sizeof(a
))
2989 /* copy_from_user should be SMP safe. */
2990 if (copy_from_user(a
, args
, len
))
2993 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
3002 err
= __sys_socket(a0
, a1
, a
[2]);
3005 err
= __sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3008 err
= __sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3011 err
= __sys_listen(a0
, a1
);
3014 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3015 (int __user
*)a
[2], 0);
3017 case SYS_GETSOCKNAME
:
3019 __sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
3020 (int __user
*)a
[2]);
3022 case SYS_GETPEERNAME
:
3024 __sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
3025 (int __user
*)a
[2]);
3027 case SYS_SOCKETPAIR
:
3028 err
= __sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
3031 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3035 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3036 (struct sockaddr __user
*)a
[4], a
[5]);
3039 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3043 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3044 (struct sockaddr __user
*)a
[4],
3045 (int __user
*)a
[5]);
3048 err
= __sys_shutdown(a0
, a1
);
3050 case SYS_SETSOCKOPT
:
3051 err
= __sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3054 case SYS_GETSOCKOPT
:
3056 __sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3057 (int __user
*)a
[4]);
3060 err
= __sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
,
3064 err
= __sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2],
3068 err
= __sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
,
3072 if (IS_ENABLED(CONFIG_64BIT
))
3073 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3075 (struct __kernel_timespec __user
*)a
[4],
3078 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3080 (struct old_timespec32 __user
*)a
[4]);
3083 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3084 (int __user
*)a
[2], a
[3]);
3093 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3096 * sock_register - add a socket protocol handler
3097 * @ops: description of protocol
3099 * This function is called by a protocol handler that wants to
3100 * advertise its address family, and have it linked into the
3101 * socket interface. The value ops->family corresponds to the
3102 * socket system call protocol family.
3104 int sock_register(const struct net_proto_family
*ops
)
3108 if (ops
->family
>= NPROTO
) {
3109 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
3113 spin_lock(&net_family_lock
);
3114 if (rcu_dereference_protected(net_families
[ops
->family
],
3115 lockdep_is_held(&net_family_lock
)))
3118 rcu_assign_pointer(net_families
[ops
->family
], ops
);
3121 spin_unlock(&net_family_lock
);
3123 pr_info("NET: Registered %s protocol family\n", pf_family_names
[ops
->family
]);
3126 EXPORT_SYMBOL(sock_register
);
3129 * sock_unregister - remove a protocol handler
3130 * @family: protocol family to remove
3132 * This function is called by a protocol handler that wants to
3133 * remove its address family, and have it unlinked from the
3134 * new socket creation.
3136 * If protocol handler is a module, then it can use module reference
3137 * counts to protect against new references. If protocol handler is not
3138 * a module then it needs to provide its own protection in
3139 * the ops->create routine.
3141 void sock_unregister(int family
)
3143 BUG_ON(family
< 0 || family
>= NPROTO
);
3145 spin_lock(&net_family_lock
);
3146 RCU_INIT_POINTER(net_families
[family
], NULL
);
3147 spin_unlock(&net_family_lock
);
3151 pr_info("NET: Unregistered %s protocol family\n", pf_family_names
[family
]);
3153 EXPORT_SYMBOL(sock_unregister
);
3155 bool sock_is_registered(int family
)
3157 return family
< NPROTO
&& rcu_access_pointer(net_families
[family
]);
3160 static int __init
sock_init(void)
3164 * Initialize the network sysctl infrastructure.
3166 err
= net_sysctl_init();
3171 * Initialize skbuff SLAB cache
3176 * Initialize the protocols module.
3181 err
= register_filesystem(&sock_fs_type
);
3184 sock_mnt
= kern_mount(&sock_fs_type
);
3185 if (IS_ERR(sock_mnt
)) {
3186 err
= PTR_ERR(sock_mnt
);
3190 /* The real protocol initialization is performed in later initcalls.
3193 #ifdef CONFIG_NETFILTER
3194 err
= netfilter_init();
3199 ptp_classifier_init();
3205 unregister_filesystem(&sock_fs_type
);
3209 core_initcall(sock_init
); /* early initcall */
3211 #ifdef CONFIG_PROC_FS
3212 void socket_seq_show(struct seq_file
*seq
)
3214 seq_printf(seq
, "sockets: used %d\n",
3215 sock_inuse_get(seq
->private));
3217 #endif /* CONFIG_PROC_FS */
3219 /* Handle the fact that while struct ifreq has the same *layout* on
3220 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3221 * which are handled elsewhere, it still has different *size* due to
3222 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3223 * resulting in struct ifreq being 32 and 40 bytes respectively).
3224 * As a result, if the struct happens to be at the end of a page and
3225 * the next page isn't readable/writable, we get a fault. To prevent
3226 * that, copy back and forth to the full size.
3228 int get_user_ifreq(struct ifreq
*ifr
, void __user
**ifrdata
, void __user
*arg
)
3230 if (in_compat_syscall()) {
3231 struct compat_ifreq
*ifr32
= (struct compat_ifreq
*)ifr
;
3233 memset(ifr
, 0, sizeof(*ifr
));
3234 if (copy_from_user(ifr32
, arg
, sizeof(*ifr32
)))
3238 *ifrdata
= compat_ptr(ifr32
->ifr_data
);
3243 if (copy_from_user(ifr
, arg
, sizeof(*ifr
)))
3247 *ifrdata
= ifr
->ifr_data
;
3251 EXPORT_SYMBOL(get_user_ifreq
);
3253 int put_user_ifreq(struct ifreq
*ifr
, void __user
*arg
)
3255 size_t size
= sizeof(*ifr
);
3257 if (in_compat_syscall())
3258 size
= sizeof(struct compat_ifreq
);
3260 if (copy_to_user(arg
, ifr
, size
))
3265 EXPORT_SYMBOL(put_user_ifreq
);
3267 #ifdef CONFIG_COMPAT
3268 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3270 compat_uptr_t uptr32
;
3275 if (get_user_ifreq(&ifr
, NULL
, uifr32
))
3278 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
3281 saved
= ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
;
3282 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= compat_ptr(uptr32
);
3284 err
= dev_ioctl(net
, SIOCWANDEV
, &ifr
, NULL
, NULL
);
3286 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= saved
;
3287 if (put_user_ifreq(&ifr
, uifr32
))
3293 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3294 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3295 struct compat_ifreq __user
*u_ifreq32
)
3300 if (!is_socket_ioctl_cmd(cmd
))
3302 if (get_user_ifreq(&ifreq
, &data
, u_ifreq32
))
3304 ifreq
.ifr_data
= data
;
3306 return dev_ioctl(net
, cmd
, &ifreq
, data
, NULL
);
3309 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3310 unsigned int cmd
, unsigned long arg
)
3312 void __user
*argp
= compat_ptr(arg
);
3313 struct sock
*sk
= sock
->sk
;
3314 struct net
*net
= sock_net(sk
);
3316 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3317 return sock_ioctl(file
, cmd
, (unsigned long)argp
);
3321 return compat_siocwandev(net
, argp
);
3322 case SIOCGSTAMP_OLD
:
3323 case SIOCGSTAMPNS_OLD
:
3324 if (!sock
->ops
->gettstamp
)
3325 return -ENOIOCTLCMD
;
3326 return sock
->ops
->gettstamp(sock
, argp
, cmd
== SIOCGSTAMP_OLD
,
3327 !COMPAT_USE_64BIT_TIME
);
3330 case SIOCBONDSLAVEINFOQUERY
:
3331 case SIOCBONDINFOQUERY
:
3334 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3345 case SIOCGSTAMP_NEW
:
3346 case SIOCGSTAMPNS_NEW
:
3350 return sock_ioctl(file
, cmd
, arg
);
3369 case SIOCSIFHWBROADCAST
:
3371 case SIOCGIFBRDADDR
:
3372 case SIOCSIFBRDADDR
:
3373 case SIOCGIFDSTADDR
:
3374 case SIOCSIFDSTADDR
:
3375 case SIOCGIFNETMASK
:
3376 case SIOCSIFNETMASK
:
3388 case SIOCBONDENSLAVE
:
3389 case SIOCBONDRELEASE
:
3390 case SIOCBONDSETHWADDR
:
3391 case SIOCBONDCHANGEACTIVE
:
3398 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3401 return -ENOIOCTLCMD
;
3404 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3407 struct socket
*sock
= file
->private_data
;
3408 int ret
= -ENOIOCTLCMD
;
3415 if (sock
->ops
->compat_ioctl
)
3416 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3418 if (ret
== -ENOIOCTLCMD
&&
3419 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3420 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3422 if (ret
== -ENOIOCTLCMD
)
3423 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3430 * kernel_bind - bind an address to a socket (kernel space)
3433 * @addrlen: length of address
3435 * Returns 0 or an error.
3438 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3440 return sock
->ops
->bind(sock
, addr
, addrlen
);
3442 EXPORT_SYMBOL(kernel_bind
);
3445 * kernel_listen - move socket to listening state (kernel space)
3447 * @backlog: pending connections queue size
3449 * Returns 0 or an error.
3452 int kernel_listen(struct socket
*sock
, int backlog
)
3454 return sock
->ops
->listen(sock
, backlog
);
3456 EXPORT_SYMBOL(kernel_listen
);
3459 * kernel_accept - accept a connection (kernel space)
3460 * @sock: listening socket
3461 * @newsock: new connected socket
3464 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3465 * If it fails, @newsock is guaranteed to be %NULL.
3466 * Returns 0 or an error.
3469 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3471 struct sock
*sk
= sock
->sk
;
3474 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3479 err
= sock
->ops
->accept(sock
, *newsock
, flags
, true);
3481 sock_release(*newsock
);
3486 (*newsock
)->ops
= sock
->ops
;
3487 __module_get((*newsock
)->ops
->owner
);
3492 EXPORT_SYMBOL(kernel_accept
);
3495 * kernel_connect - connect a socket (kernel space)
3498 * @addrlen: address length
3499 * @flags: flags (O_NONBLOCK, ...)
3501 * For datagram sockets, @addr is the address to which datagrams are sent
3502 * by default, and the only address from which datagrams are received.
3503 * For stream sockets, attempts to connect to @addr.
3504 * Returns 0 or an error code.
3507 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3510 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3512 EXPORT_SYMBOL(kernel_connect
);
3515 * kernel_getsockname - get the address which the socket is bound (kernel space)
3517 * @addr: address holder
3519 * Fills the @addr pointer with the address which the socket is bound.
3520 * Returns the length of the address in bytes or an error code.
3523 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
)
3525 return sock
->ops
->getname(sock
, addr
, 0);
3527 EXPORT_SYMBOL(kernel_getsockname
);
3530 * kernel_getpeername - get the address which the socket is connected (kernel space)
3532 * @addr: address holder
3534 * Fills the @addr pointer with the address which the socket is connected.
3535 * Returns the length of the address in bytes or an error code.
3538 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
)
3540 return sock
->ops
->getname(sock
, addr
, 1);
3542 EXPORT_SYMBOL(kernel_getpeername
);
3545 * kernel_sendpage - send a &page through a socket (kernel space)
3548 * @offset: page offset
3549 * @size: total size in bytes
3550 * @flags: flags (MSG_DONTWAIT, ...)
3552 * Returns the total amount sent in bytes or an error.
3555 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3556 size_t size
, int flags
)
3558 if (sock
->ops
->sendpage
) {
3559 /* Warn in case the improper page to zero-copy send */
3560 WARN_ONCE(!sendpage_ok(page
), "improper page for zero-copy send");
3561 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3563 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3565 EXPORT_SYMBOL(kernel_sendpage
);
3568 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3571 * @offset: page offset
3572 * @size: total size in bytes
3573 * @flags: flags (MSG_DONTWAIT, ...)
3575 * Returns the total amount sent in bytes or an error.
3576 * Caller must hold @sk.
3579 int kernel_sendpage_locked(struct sock
*sk
, struct page
*page
, int offset
,
3580 size_t size
, int flags
)
3582 struct socket
*sock
= sk
->sk_socket
;
3584 if (sock
->ops
->sendpage_locked
)
3585 return sock
->ops
->sendpage_locked(sk
, page
, offset
, size
,
3588 return sock_no_sendpage_locked(sk
, page
, offset
, size
, flags
);
3590 EXPORT_SYMBOL(kernel_sendpage_locked
);
3593 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3595 * @how: connection part
3597 * Returns 0 or an error.
3600 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3602 return sock
->ops
->shutdown(sock
, how
);
3604 EXPORT_SYMBOL(kernel_sock_shutdown
);
3607 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3610 * This routine returns the IP overhead imposed by a socket i.e.
3611 * the length of the underlying IP header, depending on whether
3612 * this is an IPv4 or IPv6 socket and the length from IP options turned
3613 * on at the socket. Assumes that the caller has a lock on the socket.
3616 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3618 struct inet_sock
*inet
;
3619 struct ip_options_rcu
*opt
;
3621 #if IS_ENABLED(CONFIG_IPV6)
3622 struct ipv6_pinfo
*np
;
3623 struct ipv6_txoptions
*optv6
= NULL
;
3624 #endif /* IS_ENABLED(CONFIG_IPV6) */
3629 switch (sk
->sk_family
) {
3632 overhead
+= sizeof(struct iphdr
);
3633 opt
= rcu_dereference_protected(inet
->inet_opt
,
3634 sock_owned_by_user(sk
));
3636 overhead
+= opt
->opt
.optlen
;
3638 #if IS_ENABLED(CONFIG_IPV6)
3641 overhead
+= sizeof(struct ipv6hdr
);
3643 optv6
= rcu_dereference_protected(np
->opt
,
3644 sock_owned_by_user(sk
));
3646 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3648 #endif /* IS_ENABLED(CONFIG_IPV6) */
3649 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3653 EXPORT_SYMBOL(kernel_sock_ip_overhead
);