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/splice.h>
61 #include <linux/net.h>
62 #include <linux/interrupt.h>
63 #include <linux/thread_info.h>
64 #include <linux/rcupdate.h>
65 #include <linux/netdevice.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68 #include <linux/mutex.h>
69 #include <linux/if_bridge.h>
70 #include <linux/if_vlan.h>
71 #include <linux/ptp_classify.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
77 #include <linux/mount.h>
78 #include <linux/pseudo_fs.h>
79 #include <linux/security.h>
80 #include <linux/syscalls.h>
81 #include <linux/compat.h>
82 #include <linux/kmod.h>
83 #include <linux/audit.h>
84 #include <linux/wireless.h>
85 #include <linux/nsproxy.h>
86 #include <linux/magic.h>
87 #include <linux/slab.h>
88 #include <linux/xattr.h>
89 #include <linux/nospec.h>
90 #include <linux/indirect_call_wrapper.h>
91 #include <linux/io_uring.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/termios.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 #include <linux/ptp_clock_kernel.h>
111 #include <trace/events/sock.h>
113 #ifdef CONFIG_NET_RX_BUSY_POLL
114 unsigned int sysctl_net_busy_read __read_mostly
;
115 unsigned int sysctl_net_busy_poll __read_mostly
;
118 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
);
119 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
);
120 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
122 static int sock_close(struct inode
*inode
, struct file
*file
);
123 static __poll_t
sock_poll(struct file
*file
,
124 struct poll_table_struct
*wait
);
125 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
127 static long compat_sock_ioctl(struct file
*file
,
128 unsigned int cmd
, unsigned long arg
);
130 static int sock_fasync(int fd
, struct file
*filp
, int on
);
131 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
132 struct pipe_inode_info
*pipe
, size_t len
,
134 static void sock_splice_eof(struct file
*file
);
136 #ifdef CONFIG_PROC_FS
137 static void sock_show_fdinfo(struct seq_file
*m
, struct file
*f
)
139 struct socket
*sock
= f
->private_data
;
140 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
142 if (ops
->show_fdinfo
)
143 ops
->show_fdinfo(m
, sock
);
146 #define sock_show_fdinfo NULL
150 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151 * in the operation structures but are done directly via the socketcall() multiplexor.
154 static const struct file_operations socket_file_ops
= {
155 .owner
= THIS_MODULE
,
157 .read_iter
= sock_read_iter
,
158 .write_iter
= sock_write_iter
,
160 .unlocked_ioctl
= sock_ioctl
,
162 .compat_ioctl
= compat_sock_ioctl
,
164 .uring_cmd
= io_uring_cmd_sock
,
166 .release
= sock_close
,
167 .fasync
= sock_fasync
,
168 .splice_write
= splice_to_socket
,
169 .splice_read
= sock_splice_read
,
170 .splice_eof
= sock_splice_eof
,
171 .show_fdinfo
= sock_show_fdinfo
,
174 static const char * const pf_family_names
[] = {
175 [PF_UNSPEC
] = "PF_UNSPEC",
176 [PF_UNIX
] = "PF_UNIX/PF_LOCAL",
177 [PF_INET
] = "PF_INET",
178 [PF_AX25
] = "PF_AX25",
180 [PF_APPLETALK
] = "PF_APPLETALK",
181 [PF_NETROM
] = "PF_NETROM",
182 [PF_BRIDGE
] = "PF_BRIDGE",
183 [PF_ATMPVC
] = "PF_ATMPVC",
185 [PF_INET6
] = "PF_INET6",
186 [PF_ROSE
] = "PF_ROSE",
187 [PF_DECnet
] = "PF_DECnet",
188 [PF_NETBEUI
] = "PF_NETBEUI",
189 [PF_SECURITY
] = "PF_SECURITY",
191 [PF_NETLINK
] = "PF_NETLINK/PF_ROUTE",
192 [PF_PACKET
] = "PF_PACKET",
194 [PF_ECONET
] = "PF_ECONET",
195 [PF_ATMSVC
] = "PF_ATMSVC",
198 [PF_IRDA
] = "PF_IRDA",
199 [PF_PPPOX
] = "PF_PPPOX",
200 [PF_WANPIPE
] = "PF_WANPIPE",
203 [PF_MPLS
] = "PF_MPLS",
205 [PF_TIPC
] = "PF_TIPC",
206 [PF_BLUETOOTH
] = "PF_BLUETOOTH",
207 [PF_IUCV
] = "PF_IUCV",
208 [PF_RXRPC
] = "PF_RXRPC",
209 [PF_ISDN
] = "PF_ISDN",
210 [PF_PHONET
] = "PF_PHONET",
211 [PF_IEEE802154
] = "PF_IEEE802154",
212 [PF_CAIF
] = "PF_CAIF",
215 [PF_VSOCK
] = "PF_VSOCK",
217 [PF_QIPCRTR
] = "PF_QIPCRTR",
220 [PF_MCTP
] = "PF_MCTP",
224 * The protocol list. Each protocol is registered in here.
227 static DEFINE_SPINLOCK(net_family_lock
);
228 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
232 * Move socket addresses back and forth across the kernel/user
233 * divide and look after the messy bits.
237 * move_addr_to_kernel - copy a socket address into kernel space
238 * @uaddr: Address in user space
239 * @kaddr: Address in kernel space
240 * @ulen: Length in user space
242 * The address is copied into kernel space. If the provided address is
243 * too long an error code of -EINVAL is returned. If the copy gives
244 * invalid addresses -EFAULT is returned. On a success 0 is returned.
247 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
249 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
253 if (copy_from_user(kaddr
, uaddr
, ulen
))
255 return audit_sockaddr(ulen
, kaddr
);
259 * move_addr_to_user - copy an address to user space
260 * @kaddr: kernel space address
261 * @klen: length of address in kernel
262 * @uaddr: user space address
263 * @ulen: pointer to user length field
265 * The value pointed to by ulen on entry is the buffer length available.
266 * This is overwritten with the buffer space used. -EINVAL is returned
267 * if an overlong buffer is specified or a negative buffer size. -EFAULT
268 * is returned if either the buffer or the length field are not
270 * After copying the data up to the limit the user specifies, the true
271 * length of the data is written over the length limit the user
272 * specified. Zero is returned for a success.
275 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
276 void __user
*uaddr
, int __user
*ulen
)
281 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
282 err
= get_user(len
, ulen
);
290 if (audit_sockaddr(klen
, kaddr
))
292 if (copy_to_user(uaddr
, kaddr
, len
))
296 * "fromlen shall refer to the value before truncation.."
299 return __put_user(klen
, ulen
);
302 static struct kmem_cache
*sock_inode_cachep __ro_after_init
;
304 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
306 struct socket_alloc
*ei
;
308 ei
= alloc_inode_sb(sb
, sock_inode_cachep
, GFP_KERNEL
);
311 init_waitqueue_head(&ei
->socket
.wq
.wait
);
312 ei
->socket
.wq
.fasync_list
= NULL
;
313 ei
->socket
.wq
.flags
= 0;
315 ei
->socket
.state
= SS_UNCONNECTED
;
316 ei
->socket
.flags
= 0;
317 ei
->socket
.ops
= NULL
;
318 ei
->socket
.sk
= NULL
;
319 ei
->socket
.file
= NULL
;
321 return &ei
->vfs_inode
;
324 static void sock_free_inode(struct inode
*inode
)
326 struct socket_alloc
*ei
;
328 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
329 kmem_cache_free(sock_inode_cachep
, ei
);
332 static void init_once(void *foo
)
334 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
336 inode_init_once(&ei
->vfs_inode
);
339 static void init_inodecache(void)
341 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
342 sizeof(struct socket_alloc
),
344 (SLAB_HWCACHE_ALIGN
|
345 SLAB_RECLAIM_ACCOUNT
|
346 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
348 BUG_ON(sock_inode_cachep
== NULL
);
351 static const struct super_operations sockfs_ops
= {
352 .alloc_inode
= sock_alloc_inode
,
353 .free_inode
= sock_free_inode
,
354 .statfs
= simple_statfs
,
358 * sockfs_dname() is called from d_path().
360 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
362 return dynamic_dname(buffer
, buflen
, "socket:[%lu]",
363 d_inode(dentry
)->i_ino
);
366 static const struct dentry_operations sockfs_dentry_operations
= {
367 .d_dname
= sockfs_dname
,
370 static int sockfs_xattr_get(const struct xattr_handler
*handler
,
371 struct dentry
*dentry
, struct inode
*inode
,
372 const char *suffix
, void *value
, size_t size
)
375 if (dentry
->d_name
.len
+ 1 > size
)
377 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
379 return dentry
->d_name
.len
+ 1;
382 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
383 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
384 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
386 static const struct xattr_handler sockfs_xattr_handler
= {
387 .name
= XATTR_NAME_SOCKPROTONAME
,
388 .get
= sockfs_xattr_get
,
391 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
392 struct mnt_idmap
*idmap
,
393 struct dentry
*dentry
, struct inode
*inode
,
394 const char *suffix
, const void *value
,
395 size_t size
, int flags
)
397 /* Handled by LSM. */
401 static const struct xattr_handler sockfs_security_xattr_handler
= {
402 .prefix
= XATTR_SECURITY_PREFIX
,
403 .set
= sockfs_security_xattr_set
,
406 static const struct xattr_handler
* const sockfs_xattr_handlers
[] = {
407 &sockfs_xattr_handler
,
408 &sockfs_security_xattr_handler
,
412 static int sockfs_init_fs_context(struct fs_context
*fc
)
414 struct pseudo_fs_context
*ctx
= init_pseudo(fc
, SOCKFS_MAGIC
);
417 ctx
->ops
= &sockfs_ops
;
418 ctx
->dops
= &sockfs_dentry_operations
;
419 ctx
->xattr
= sockfs_xattr_handlers
;
423 static struct vfsmount
*sock_mnt __read_mostly
;
425 static struct file_system_type sock_fs_type
= {
427 .init_fs_context
= sockfs_init_fs_context
,
428 .kill_sb
= kill_anon_super
,
432 * Obtains the first available file descriptor and sets it up for use.
434 * These functions create file structures and maps them to fd space
435 * of the current process. On success it returns file descriptor
436 * and file struct implicitly stored in sock->file.
437 * Note that another thread may close file descriptor before we return
438 * from this function. We use the fact that now we do not refer
439 * to socket after mapping. If one day we will need it, this
440 * function will increment ref. count on file by 1.
442 * In any case returned fd MAY BE not valid!
443 * This race condition is unavoidable
444 * with shared fd spaces, we cannot solve it inside kernel,
445 * but we take care of internal coherence yet.
449 * sock_alloc_file - Bind a &socket to a &file
451 * @flags: file status flags
452 * @dname: protocol name
454 * Returns the &file bound with @sock, implicitly storing it
455 * in sock->file. If dname is %NULL, sets to "".
457 * On failure @sock is released, and an ERR pointer is returned.
459 * This function uses GFP_KERNEL internally.
462 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
467 dname
= sock
->sk
? sock
->sk
->sk_prot_creator
->name
: "";
469 file
= alloc_file_pseudo(SOCK_INODE(sock
), sock_mnt
, dname
,
470 O_RDWR
| (flags
& O_NONBLOCK
),
477 file
->f_mode
|= FMODE_NOWAIT
;
479 file
->private_data
= sock
;
480 stream_open(SOCK_INODE(sock
), file
);
483 EXPORT_SYMBOL(sock_alloc_file
);
485 static int sock_map_fd(struct socket
*sock
, int flags
)
487 struct file
*newfile
;
488 int fd
= get_unused_fd_flags(flags
);
489 if (unlikely(fd
< 0)) {
494 newfile
= sock_alloc_file(sock
, flags
, NULL
);
495 if (!IS_ERR(newfile
)) {
496 fd_install(fd
, newfile
);
501 return PTR_ERR(newfile
);
505 * sock_from_file - Return the &socket bounded to @file.
508 * On failure returns %NULL.
511 struct socket
*sock_from_file(struct file
*file
)
513 if (file
->f_op
== &socket_file_ops
)
514 return file
->private_data
; /* set in sock_alloc_file */
518 EXPORT_SYMBOL(sock_from_file
);
521 * sockfd_lookup - Go from a file number to its socket slot
523 * @err: pointer to an error code return
525 * The file handle passed in is locked and the socket it is bound
526 * to is returned. If an error occurs the err pointer is overwritten
527 * with a negative errno code and NULL is returned. The function checks
528 * for both invalid handles and passing a handle which is not a socket.
530 * On a success the socket object pointer is returned.
533 struct socket
*sockfd_lookup(int fd
, int *err
)
544 sock
= sock_from_file(file
);
551 EXPORT_SYMBOL(sockfd_lookup
);
553 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
555 struct fd f
= fdget(fd
);
560 sock
= sock_from_file(f
.file
);
562 *fput_needed
= f
.flags
& FDPUT_FPUT
;
571 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
577 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
587 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
592 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
599 static int sockfs_setattr(struct mnt_idmap
*idmap
,
600 struct dentry
*dentry
, struct iattr
*iattr
)
602 int err
= simple_setattr(&nop_mnt_idmap
, dentry
, iattr
);
604 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
605 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
608 sock
->sk
->sk_uid
= iattr
->ia_uid
;
616 static const struct inode_operations sockfs_inode_ops
= {
617 .listxattr
= sockfs_listxattr
,
618 .setattr
= sockfs_setattr
,
622 * sock_alloc - allocate a socket
624 * Allocate a new inode and socket object. The two are bound together
625 * and initialised. The socket is then returned. If we are out of inodes
626 * NULL is returned. This functions uses GFP_KERNEL internally.
629 struct socket
*sock_alloc(void)
634 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
638 sock
= SOCKET_I(inode
);
640 inode
->i_ino
= get_next_ino();
641 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
642 inode
->i_uid
= current_fsuid();
643 inode
->i_gid
= current_fsgid();
644 inode
->i_op
= &sockfs_inode_ops
;
648 EXPORT_SYMBOL(sock_alloc
);
650 static void __sock_release(struct socket
*sock
, struct inode
*inode
)
652 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
655 struct module
*owner
= ops
->owner
;
667 if (sock
->wq
.fasync_list
)
668 pr_err("%s: fasync list not empty!\n", __func__
);
671 iput(SOCK_INODE(sock
));
678 * sock_release - close a socket
679 * @sock: socket to close
681 * The socket is released from the protocol stack if it has a release
682 * callback, and the inode is then released if the socket is bound to
683 * an inode not a file.
685 void sock_release(struct socket
*sock
)
687 __sock_release(sock
, NULL
);
689 EXPORT_SYMBOL(sock_release
);
691 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
693 u8 flags
= *tx_flags
;
695 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
) {
696 flags
|= SKBTX_HW_TSTAMP
;
698 /* PTP hardware clocks can provide a free running cycle counter
699 * as a time base for virtual clocks. Tell driver to use the
700 * free running cycle counter for timestamp if socket is bound
703 if (tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
704 flags
|= SKBTX_HW_TSTAMP_USE_CYCLES
;
707 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
708 flags
|= SKBTX_SW_TSTAMP
;
710 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
711 flags
|= SKBTX_SCHED_TSTAMP
;
715 EXPORT_SYMBOL(__sock_tx_timestamp
);
717 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket
*, struct msghdr
*,
719 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket
*, struct msghdr
*,
722 static noinline
void call_trace_sock_send_length(struct sock
*sk
, int ret
,
725 trace_sock_send_length(sk
, ret
, 0);
728 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
730 int ret
= INDIRECT_CALL_INET(READ_ONCE(sock
->ops
)->sendmsg
, inet6_sendmsg
,
731 inet_sendmsg
, sock
, msg
,
733 BUG_ON(ret
== -EIOCBQUEUED
);
735 if (trace_sock_send_length_enabled())
736 call_trace_sock_send_length(sock
->sk
, ret
, 0);
740 static int __sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
742 int err
= security_socket_sendmsg(sock
, msg
,
745 return err
?: sock_sendmsg_nosec(sock
, msg
);
749 * sock_sendmsg - send a message through @sock
751 * @msg: message to send
753 * Sends @msg through @sock, passing through LSM.
754 * Returns the number of bytes sent, or an error code.
756 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
758 struct sockaddr_storage
*save_addr
= (struct sockaddr_storage
*)msg
->msg_name
;
759 struct sockaddr_storage address
;
763 memcpy(&address
, msg
->msg_name
, msg
->msg_namelen
);
764 msg
->msg_name
= &address
;
767 ret
= __sock_sendmsg(sock
, msg
);
768 msg
->msg_name
= save_addr
;
772 EXPORT_SYMBOL(sock_sendmsg
);
775 * kernel_sendmsg - send a message through @sock (kernel-space)
777 * @msg: message header
779 * @num: vec array length
780 * @size: total message data size
782 * Builds the message data with @vec and sends it through @sock.
783 * Returns the number of bytes sent, or an error code.
786 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
787 struct kvec
*vec
, size_t num
, size_t size
)
789 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
790 return sock_sendmsg(sock
, msg
);
792 EXPORT_SYMBOL(kernel_sendmsg
);
795 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
797 * @msg: message header
798 * @vec: output s/g array
799 * @num: output s/g array length
800 * @size: total message data size
802 * Builds the message data with @vec and sends it through @sock.
803 * Returns the number of bytes sent, or an error code.
804 * Caller must hold @sk.
807 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
808 struct kvec
*vec
, size_t num
, size_t size
)
810 struct socket
*sock
= sk
->sk_socket
;
811 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
813 if (!ops
->sendmsg_locked
)
814 return sock_no_sendmsg_locked(sk
, msg
, size
);
816 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
818 return ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
820 EXPORT_SYMBOL(kernel_sendmsg_locked
);
822 static bool skb_is_err_queue(const struct sk_buff
*skb
)
824 /* pkt_type of skbs enqueued on the error queue are set to
825 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
826 * in recvmsg, since skbs received on a local socket will never
827 * have a pkt_type of PACKET_OUTGOING.
829 return skb
->pkt_type
== PACKET_OUTGOING
;
832 /* On transmit, software and hardware timestamps are returned independently.
833 * As the two skb clones share the hardware timestamp, which may be updated
834 * before the software timestamp is received, a hardware TX timestamp may be
835 * returned only if there is no software TX timestamp. Ignore false software
836 * timestamps, which may be made in the __sock_recv_timestamp() call when the
837 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
838 * hardware timestamp.
840 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
842 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
845 static ktime_t
get_timestamp(struct sock
*sk
, struct sk_buff
*skb
, int *if_index
)
847 bool cycles
= READ_ONCE(sk
->sk_tsflags
) & SOF_TIMESTAMPING_BIND_PHC
;
848 struct skb_shared_hwtstamps
*shhwtstamps
= skb_hwtstamps(skb
);
849 struct net_device
*orig_dev
;
853 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
855 *if_index
= orig_dev
->ifindex
;
856 hwtstamp
= netdev_get_tstamp(orig_dev
, shhwtstamps
, cycles
);
858 hwtstamp
= shhwtstamps
->hwtstamp
;
865 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
,
868 struct scm_ts_pktinfo ts_pktinfo
;
869 struct net_device
*orig_dev
;
871 if (!skb_mac_header_was_set(skb
))
874 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
878 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
880 if_index
= orig_dev
->ifindex
;
883 ts_pktinfo
.if_index
= if_index
;
885 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
886 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
887 sizeof(ts_pktinfo
), &ts_pktinfo
);
891 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
893 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
896 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
897 int new_tstamp
= sock_flag(sk
, SOCK_TSTAMP_NEW
);
898 struct scm_timestamping_internal tss
;
899 int empty
= 1, false_tstamp
= 0;
900 struct skb_shared_hwtstamps
*shhwtstamps
=
906 /* Race occurred between timestamp enabling and packet
907 receiving. Fill in the current time for now. */
908 if (need_software_tstamp
&& skb
->tstamp
== 0) {
909 __net_timestamp(skb
);
913 if (need_software_tstamp
) {
914 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
916 struct __kernel_sock_timeval tv
;
918 skb_get_new_timestamp(skb
, &tv
);
919 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_NEW
,
922 struct __kernel_old_timeval tv
;
924 skb_get_timestamp(skb
, &tv
);
925 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_OLD
,
930 struct __kernel_timespec ts
;
932 skb_get_new_timestampns(skb
, &ts
);
933 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_NEW
,
936 struct __kernel_old_timespec ts
;
938 skb_get_timestampns(skb
, &ts
);
939 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_OLD
,
945 memset(&tss
, 0, sizeof(tss
));
946 tsflags
= READ_ONCE(sk
->sk_tsflags
);
947 if ((tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
948 ktime_to_timespec64_cond(skb
->tstamp
, tss
.ts
+ 0))
951 (tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
952 !skb_is_swtx_tstamp(skb
, false_tstamp
)) {
954 if (skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP_NETDEV
)
955 hwtstamp
= get_timestamp(sk
, skb
, &if_index
);
957 hwtstamp
= shhwtstamps
->hwtstamp
;
959 if (tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
960 hwtstamp
= ptp_convert_timestamp(&hwtstamp
,
961 READ_ONCE(sk
->sk_bind_phc
));
963 if (ktime_to_timespec64_cond(hwtstamp
, tss
.ts
+ 2)) {
966 if ((tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
967 !skb_is_err_queue(skb
))
968 put_ts_pktinfo(msg
, skb
, if_index
);
972 if (sock_flag(sk
, SOCK_TSTAMP_NEW
))
973 put_cmsg_scm_timestamping64(msg
, &tss
);
975 put_cmsg_scm_timestamping(msg
, &tss
);
977 if (skb_is_err_queue(skb
) && skb
->len
&&
978 SKB_EXT_ERR(skb
)->opt_stats
)
979 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
980 skb
->len
, skb
->data
);
983 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
985 #ifdef CONFIG_WIRELESS
986 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
991 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
993 if (!skb
->wifi_acked_valid
)
996 ack
= skb
->wifi_acked
;
998 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
1000 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
1003 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
1004 struct sk_buff
*skb
)
1006 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
1007 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
1008 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
1011 static void sock_recv_mark(struct msghdr
*msg
, struct sock
*sk
,
1012 struct sk_buff
*skb
)
1014 if (sock_flag(sk
, SOCK_RCVMARK
) && skb
) {
1015 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1016 __u32 mark
= skb
->mark
;
1018 put_cmsg(msg
, SOL_SOCKET
, SO_MARK
, sizeof(__u32
), &mark
);
1022 void __sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
1023 struct sk_buff
*skb
)
1025 sock_recv_timestamp(msg
, sk
, skb
);
1026 sock_recv_drops(msg
, sk
, skb
);
1027 sock_recv_mark(msg
, sk
, skb
);
1029 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs
);
1031 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket
*, struct msghdr
*,
1033 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket
*, struct msghdr
*,
1036 static noinline
void call_trace_sock_recv_length(struct sock
*sk
, int ret
, int flags
)
1038 trace_sock_recv_length(sk
, ret
, flags
);
1041 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
1044 int ret
= INDIRECT_CALL_INET(READ_ONCE(sock
->ops
)->recvmsg
,
1046 inet_recvmsg
, sock
, msg
,
1047 msg_data_left(msg
), flags
);
1048 if (trace_sock_recv_length_enabled())
1049 call_trace_sock_recv_length(sock
->sk
, ret
, flags
);
1054 * sock_recvmsg - receive a message from @sock
1056 * @msg: message to receive
1057 * @flags: message flags
1059 * Receives @msg from @sock, passing through LSM. Returns the total number
1060 * of bytes received, or an error.
1062 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
1064 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
1066 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
1068 EXPORT_SYMBOL(sock_recvmsg
);
1071 * kernel_recvmsg - Receive a message from a socket (kernel space)
1072 * @sock: The socket to receive the message from
1073 * @msg: Received message
1074 * @vec: Input s/g array for message data
1075 * @num: Size of input s/g array
1076 * @size: Number of bytes to read
1077 * @flags: Message flags (MSG_DONTWAIT, etc...)
1079 * On return the msg structure contains the scatter/gather array passed in the
1080 * vec argument. The array is modified so that it consists of the unfilled
1081 * portion of the original array.
1083 * The returned value is the total number of bytes received, or an error.
1086 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
1087 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
1089 msg
->msg_control_is_user
= false;
1090 iov_iter_kvec(&msg
->msg_iter
, ITER_DEST
, vec
, num
, size
);
1091 return sock_recvmsg(sock
, msg
, flags
);
1093 EXPORT_SYMBOL(kernel_recvmsg
);
1095 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
1096 struct pipe_inode_info
*pipe
, size_t len
,
1099 struct socket
*sock
= file
->private_data
;
1100 const struct proto_ops
*ops
;
1102 ops
= READ_ONCE(sock
->ops
);
1103 if (unlikely(!ops
->splice_read
))
1104 return copy_splice_read(file
, ppos
, pipe
, len
, flags
);
1106 return ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
1109 static void sock_splice_eof(struct file
*file
)
1111 struct socket
*sock
= file
->private_data
;
1112 const struct proto_ops
*ops
;
1114 ops
= READ_ONCE(sock
->ops
);
1115 if (ops
->splice_eof
)
1116 ops
->splice_eof(sock
);
1119 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1121 struct file
*file
= iocb
->ki_filp
;
1122 struct socket
*sock
= file
->private_data
;
1123 struct msghdr msg
= {.msg_iter
= *to
,
1127 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1128 msg
.msg_flags
= MSG_DONTWAIT
;
1130 if (iocb
->ki_pos
!= 0)
1133 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
1136 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
1141 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1143 struct file
*file
= iocb
->ki_filp
;
1144 struct socket
*sock
= file
->private_data
;
1145 struct msghdr msg
= {.msg_iter
= *from
,
1149 if (iocb
->ki_pos
!= 0)
1152 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1153 msg
.msg_flags
= MSG_DONTWAIT
;
1155 if (sock
->type
== SOCK_SEQPACKET
)
1156 msg
.msg_flags
|= MSG_EOR
;
1158 res
= __sock_sendmsg(sock
, &msg
);
1159 *from
= msg
.msg_iter
;
1164 * Atomic setting of ioctl hooks to avoid race
1165 * with module unload.
1168 static DEFINE_MUTEX(br_ioctl_mutex
);
1169 static int (*br_ioctl_hook
)(struct net
*net
, struct net_bridge
*br
,
1170 unsigned int cmd
, struct ifreq
*ifr
,
1173 void brioctl_set(int (*hook
)(struct net
*net
, struct net_bridge
*br
,
1174 unsigned int cmd
, struct ifreq
*ifr
,
1177 mutex_lock(&br_ioctl_mutex
);
1178 br_ioctl_hook
= hook
;
1179 mutex_unlock(&br_ioctl_mutex
);
1181 EXPORT_SYMBOL(brioctl_set
);
1183 int br_ioctl_call(struct net
*net
, struct net_bridge
*br
, unsigned int cmd
,
1184 struct ifreq
*ifr
, void __user
*uarg
)
1189 request_module("bridge");
1191 mutex_lock(&br_ioctl_mutex
);
1193 err
= br_ioctl_hook(net
, br
, cmd
, ifr
, uarg
);
1194 mutex_unlock(&br_ioctl_mutex
);
1199 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1200 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1202 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1204 mutex_lock(&vlan_ioctl_mutex
);
1205 vlan_ioctl_hook
= hook
;
1206 mutex_unlock(&vlan_ioctl_mutex
);
1208 EXPORT_SYMBOL(vlan_ioctl_set
);
1210 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1211 unsigned int cmd
, unsigned long arg
)
1213 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
1217 void __user
*argp
= (void __user
*)arg
;
1220 err
= ops
->ioctl(sock
, cmd
, arg
);
1223 * If this ioctl is unknown try to hand it down
1224 * to the NIC driver.
1226 if (err
!= -ENOIOCTLCMD
)
1229 if (!is_socket_ioctl_cmd(cmd
))
1232 if (get_user_ifreq(&ifr
, &data
, argp
))
1234 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1235 if (!err
&& need_copyout
)
1236 if (put_user_ifreq(&ifr
, argp
))
1243 * With an ioctl, arg may well be a user mode pointer, but we don't know
1244 * what to do with it - that's up to the protocol still.
1247 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1249 const struct proto_ops
*ops
;
1250 struct socket
*sock
;
1252 void __user
*argp
= (void __user
*)arg
;
1256 sock
= file
->private_data
;
1257 ops
= READ_ONCE(sock
->ops
);
1260 if (unlikely(cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))) {
1264 if (get_user_ifreq(&ifr
, &data
, argp
))
1266 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1267 if (!err
&& need_copyout
)
1268 if (put_user_ifreq(&ifr
, argp
))
1271 #ifdef CONFIG_WEXT_CORE
1272 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1273 err
= wext_handle_ioctl(net
, cmd
, argp
);
1280 if (get_user(pid
, (int __user
*)argp
))
1282 err
= f_setown(sock
->file
, pid
, 1);
1286 err
= put_user(f_getown(sock
->file
),
1287 (int __user
*)argp
);
1293 err
= br_ioctl_call(net
, NULL
, cmd
, NULL
, argp
);
1298 if (!vlan_ioctl_hook
)
1299 request_module("8021q");
1301 mutex_lock(&vlan_ioctl_mutex
);
1302 if (vlan_ioctl_hook
)
1303 err
= vlan_ioctl_hook(net
, argp
);
1304 mutex_unlock(&vlan_ioctl_mutex
);
1308 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1311 err
= open_related_ns(&net
->ns
, get_net_ns
);
1313 case SIOCGSTAMP_OLD
:
1314 case SIOCGSTAMPNS_OLD
:
1315 if (!ops
->gettstamp
) {
1319 err
= ops
->gettstamp(sock
, argp
,
1320 cmd
== SIOCGSTAMP_OLD
,
1321 !IS_ENABLED(CONFIG_64BIT
));
1323 case SIOCGSTAMP_NEW
:
1324 case SIOCGSTAMPNS_NEW
:
1325 if (!ops
->gettstamp
) {
1329 err
= ops
->gettstamp(sock
, argp
,
1330 cmd
== SIOCGSTAMP_NEW
,
1335 err
= dev_ifconf(net
, argp
);
1339 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1346 * sock_create_lite - creates a socket
1347 * @family: protocol family (AF_INET, ...)
1348 * @type: communication type (SOCK_STREAM, ...)
1349 * @protocol: protocol (0, ...)
1352 * Creates a new socket and assigns it to @res, passing through LSM.
1353 * The new socket initialization is not complete, see kernel_accept().
1354 * Returns 0 or an error. On failure @res is set to %NULL.
1355 * This function internally uses GFP_KERNEL.
1358 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1361 struct socket
*sock
= NULL
;
1363 err
= security_socket_create(family
, type
, protocol
, 1);
1367 sock
= sock_alloc();
1374 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1386 EXPORT_SYMBOL(sock_create_lite
);
1388 /* No kernel lock held - perfect */
1389 static __poll_t
sock_poll(struct file
*file
, poll_table
*wait
)
1391 struct socket
*sock
= file
->private_data
;
1392 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
1393 __poll_t events
= poll_requested_events(wait
), flag
= 0;
1398 if (sk_can_busy_loop(sock
->sk
)) {
1399 /* poll once if requested by the syscall */
1400 if (events
& POLL_BUSY_LOOP
)
1401 sk_busy_loop(sock
->sk
, 1);
1403 /* if this socket can poll_ll, tell the system call */
1404 flag
= POLL_BUSY_LOOP
;
1407 return ops
->poll(file
, sock
, wait
) | flag
;
1410 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1412 struct socket
*sock
= file
->private_data
;
1414 return READ_ONCE(sock
->ops
)->mmap(file
, sock
, vma
);
1417 static int sock_close(struct inode
*inode
, struct file
*filp
)
1419 __sock_release(SOCKET_I(inode
), inode
);
1424 * Update the socket async list
1426 * Fasync_list locking strategy.
1428 * 1. fasync_list is modified only under process context socket lock
1429 * i.e. under semaphore.
1430 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1431 * or under socket lock
1434 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1436 struct socket
*sock
= filp
->private_data
;
1437 struct sock
*sk
= sock
->sk
;
1438 struct socket_wq
*wq
= &sock
->wq
;
1444 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1446 if (!wq
->fasync_list
)
1447 sock_reset_flag(sk
, SOCK_FASYNC
);
1449 sock_set_flag(sk
, SOCK_FASYNC
);
1455 /* This function may be called only under rcu_lock */
1457 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1459 if (!wq
|| !wq
->fasync_list
)
1463 case SOCK_WAKE_WAITD
:
1464 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1467 case SOCK_WAKE_SPACE
:
1468 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1473 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1476 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1481 EXPORT_SYMBOL(sock_wake_async
);
1484 * __sock_create - creates a socket
1485 * @net: net namespace
1486 * @family: protocol family (AF_INET, ...)
1487 * @type: communication type (SOCK_STREAM, ...)
1488 * @protocol: protocol (0, ...)
1490 * @kern: boolean for kernel space sockets
1492 * Creates a new socket and assigns it to @res, passing through LSM.
1493 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1494 * be set to true if the socket resides in kernel space.
1495 * This function internally uses GFP_KERNEL.
1498 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1499 struct socket
**res
, int kern
)
1502 struct socket
*sock
;
1503 const struct net_proto_family
*pf
;
1506 * Check protocol is in range
1508 if (family
< 0 || family
>= NPROTO
)
1509 return -EAFNOSUPPORT
;
1510 if (type
< 0 || type
>= SOCK_MAX
)
1515 This uglymoron is moved from INET layer to here to avoid
1516 deadlock in module load.
1518 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1519 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1524 err
= security_socket_create(family
, type
, protocol
, kern
);
1529 * Allocate the socket and allow the family to set things up. if
1530 * the protocol is 0, the family is instructed to select an appropriate
1533 sock
= sock_alloc();
1535 net_warn_ratelimited("socket: no more sockets\n");
1536 return -ENFILE
; /* Not exactly a match, but its the
1537 closest posix thing */
1542 #ifdef CONFIG_MODULES
1543 /* Attempt to load a protocol module if the find failed.
1545 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1546 * requested real, full-featured networking support upon configuration.
1547 * Otherwise module support will break!
1549 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1550 request_module("net-pf-%d", family
);
1554 pf
= rcu_dereference(net_families
[family
]);
1555 err
= -EAFNOSUPPORT
;
1560 * We will call the ->create function, that possibly is in a loadable
1561 * module, so we have to bump that loadable module refcnt first.
1563 if (!try_module_get(pf
->owner
))
1566 /* Now protected by module ref count */
1569 err
= pf
->create(net
, sock
, protocol
, kern
);
1571 goto out_module_put
;
1574 * Now to bump the refcnt of the [loadable] module that owns this
1575 * socket at sock_release time we decrement its refcnt.
1577 if (!try_module_get(sock
->ops
->owner
))
1578 goto out_module_busy
;
1581 * Now that we're done with the ->create function, the [loadable]
1582 * module can have its refcnt decremented
1584 module_put(pf
->owner
);
1585 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1587 goto out_sock_release
;
1593 err
= -EAFNOSUPPORT
;
1596 module_put(pf
->owner
);
1603 goto out_sock_release
;
1605 EXPORT_SYMBOL(__sock_create
);
1608 * sock_create - creates a socket
1609 * @family: protocol family (AF_INET, ...)
1610 * @type: communication type (SOCK_STREAM, ...)
1611 * @protocol: protocol (0, ...)
1614 * A wrapper around __sock_create().
1615 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1618 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1620 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1622 EXPORT_SYMBOL(sock_create
);
1625 * sock_create_kern - creates a socket (kernel space)
1626 * @net: net namespace
1627 * @family: protocol family (AF_INET, ...)
1628 * @type: communication type (SOCK_STREAM, ...)
1629 * @protocol: protocol (0, ...)
1632 * A wrapper around __sock_create().
1633 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1636 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1638 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1640 EXPORT_SYMBOL(sock_create_kern
);
1642 static struct socket
*__sys_socket_create(int family
, int type
, int protocol
)
1644 struct socket
*sock
;
1647 /* Check the SOCK_* constants for consistency. */
1648 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1649 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1650 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1651 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1653 if ((type
& ~SOCK_TYPE_MASK
) & ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1654 return ERR_PTR(-EINVAL
);
1655 type
&= SOCK_TYPE_MASK
;
1657 retval
= sock_create(family
, type
, protocol
, &sock
);
1659 return ERR_PTR(retval
);
1664 struct file
*__sys_socket_file(int family
, int type
, int protocol
)
1666 struct socket
*sock
;
1669 sock
= __sys_socket_create(family
, type
, protocol
);
1671 return ERR_CAST(sock
);
1673 flags
= type
& ~SOCK_TYPE_MASK
;
1674 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1675 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1677 return sock_alloc_file(sock
, flags
, NULL
);
1680 /* A hook for bpf progs to attach to and update socket protocol.
1682 * A static noinline declaration here could cause the compiler to
1683 * optimize away the function. A global noinline declaration will
1684 * keep the definition, but may optimize away the callsite.
1685 * Therefore, __weak is needed to ensure that the call is still
1686 * emitted, by telling the compiler that we don't know what the
1687 * function might eventually be.
1692 __weak noinline
int update_socket_protocol(int family
, int type
, int protocol
)
1699 int __sys_socket(int family
, int type
, int protocol
)
1701 struct socket
*sock
;
1704 sock
= __sys_socket_create(family
, type
,
1705 update_socket_protocol(family
, type
, protocol
));
1707 return PTR_ERR(sock
);
1709 flags
= type
& ~SOCK_TYPE_MASK
;
1710 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1711 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1713 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1716 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1718 return __sys_socket(family
, type
, protocol
);
1722 * Create a pair of connected sockets.
1725 int __sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1727 struct socket
*sock1
, *sock2
;
1729 struct file
*newfile1
, *newfile2
;
1732 flags
= type
& ~SOCK_TYPE_MASK
;
1733 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1735 type
&= SOCK_TYPE_MASK
;
1737 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1738 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1741 * reserve descriptors and make sure we won't fail
1742 * to return them to userland.
1744 fd1
= get_unused_fd_flags(flags
);
1745 if (unlikely(fd1
< 0))
1748 fd2
= get_unused_fd_flags(flags
);
1749 if (unlikely(fd2
< 0)) {
1754 err
= put_user(fd1
, &usockvec
[0]);
1758 err
= put_user(fd2
, &usockvec
[1]);
1763 * Obtain the first socket and check if the underlying protocol
1764 * supports the socketpair call.
1767 err
= sock_create(family
, type
, protocol
, &sock1
);
1768 if (unlikely(err
< 0))
1771 err
= sock_create(family
, type
, protocol
, &sock2
);
1772 if (unlikely(err
< 0)) {
1773 sock_release(sock1
);
1777 err
= security_socket_socketpair(sock1
, sock2
);
1778 if (unlikely(err
)) {
1779 sock_release(sock2
);
1780 sock_release(sock1
);
1784 err
= READ_ONCE(sock1
->ops
)->socketpair(sock1
, sock2
);
1785 if (unlikely(err
< 0)) {
1786 sock_release(sock2
);
1787 sock_release(sock1
);
1791 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1792 if (IS_ERR(newfile1
)) {
1793 err
= PTR_ERR(newfile1
);
1794 sock_release(sock2
);
1798 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1799 if (IS_ERR(newfile2
)) {
1800 err
= PTR_ERR(newfile2
);
1805 audit_fd_pair(fd1
, fd2
);
1807 fd_install(fd1
, newfile1
);
1808 fd_install(fd2
, newfile2
);
1817 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1818 int __user
*, usockvec
)
1820 return __sys_socketpair(family
, type
, protocol
, usockvec
);
1824 * Bind a name to a socket. Nothing much to do here since it's
1825 * the protocol's responsibility to handle the local address.
1827 * We move the socket address to kernel space before we call
1828 * the protocol layer (having also checked the address is ok).
1831 int __sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1833 struct socket
*sock
;
1834 struct sockaddr_storage address
;
1835 int err
, fput_needed
;
1837 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1839 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1841 err
= security_socket_bind(sock
,
1842 (struct sockaddr
*)&address
,
1845 err
= READ_ONCE(sock
->ops
)->bind(sock
,
1849 fput_light(sock
->file
, fput_needed
);
1854 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1856 return __sys_bind(fd
, umyaddr
, addrlen
);
1860 * Perform a listen. Basically, we allow the protocol to do anything
1861 * necessary for a listen, and if that works, we mark the socket as
1862 * ready for listening.
1865 int __sys_listen(int fd
, int backlog
)
1867 struct socket
*sock
;
1868 int err
, fput_needed
;
1871 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1873 somaxconn
= READ_ONCE(sock_net(sock
->sk
)->core
.sysctl_somaxconn
);
1874 if ((unsigned int)backlog
> somaxconn
)
1875 backlog
= somaxconn
;
1877 err
= security_socket_listen(sock
, backlog
);
1879 err
= READ_ONCE(sock
->ops
)->listen(sock
, backlog
);
1881 fput_light(sock
->file
, fput_needed
);
1886 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1888 return __sys_listen(fd
, backlog
);
1891 struct file
*do_accept(struct file
*file
, unsigned file_flags
,
1892 struct sockaddr __user
*upeer_sockaddr
,
1893 int __user
*upeer_addrlen
, int flags
)
1895 struct socket
*sock
, *newsock
;
1896 struct file
*newfile
;
1898 struct sockaddr_storage address
;
1899 const struct proto_ops
*ops
;
1901 sock
= sock_from_file(file
);
1903 return ERR_PTR(-ENOTSOCK
);
1905 newsock
= sock_alloc();
1907 return ERR_PTR(-ENFILE
);
1908 ops
= READ_ONCE(sock
->ops
);
1910 newsock
->type
= sock
->type
;
1914 * We don't need try_module_get here, as the listening socket (sock)
1915 * has the protocol module (sock->ops->owner) held.
1917 __module_get(ops
->owner
);
1919 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1920 if (IS_ERR(newfile
))
1923 err
= security_socket_accept(sock
, newsock
);
1927 err
= ops
->accept(sock
, newsock
, sock
->file
->f_flags
| file_flags
,
1932 if (upeer_sockaddr
) {
1933 len
= ops
->getname(newsock
, (struct sockaddr
*)&address
, 2);
1935 err
= -ECONNABORTED
;
1938 err
= move_addr_to_user(&address
,
1939 len
, upeer_sockaddr
, upeer_addrlen
);
1944 /* File flags are not inherited via accept() unlike another OSes. */
1948 return ERR_PTR(err
);
1951 static int __sys_accept4_file(struct file
*file
, struct sockaddr __user
*upeer_sockaddr
,
1952 int __user
*upeer_addrlen
, int flags
)
1954 struct file
*newfile
;
1957 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1960 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1961 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1963 newfd
= get_unused_fd_flags(flags
);
1964 if (unlikely(newfd
< 0))
1967 newfile
= do_accept(file
, 0, upeer_sockaddr
, upeer_addrlen
,
1969 if (IS_ERR(newfile
)) {
1970 put_unused_fd(newfd
);
1971 return PTR_ERR(newfile
);
1973 fd_install(newfd
, newfile
);
1978 * For accept, we attempt to create a new socket, set up the link
1979 * with the client, wake up the client, then return the new
1980 * connected fd. We collect the address of the connector in kernel
1981 * space and move it to user at the very end. This is unclean because
1982 * we open the socket then return an error.
1984 * 1003.1g adds the ability to recvmsg() to query connection pending
1985 * status to recvmsg. We need to add that support in a way thats
1986 * clean when we restructure accept also.
1989 int __sys_accept4(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1990 int __user
*upeer_addrlen
, int flags
)
1997 ret
= __sys_accept4_file(f
.file
, upeer_sockaddr
,
1998 upeer_addrlen
, flags
);
2005 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
2006 int __user
*, upeer_addrlen
, int, flags
)
2008 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, flags
);
2011 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
2012 int __user
*, upeer_addrlen
)
2014 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
2018 * Attempt to connect to a socket with the server address. The address
2019 * is in user space so we verify it is OK and move it to kernel space.
2021 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2024 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2025 * other SEQPACKET protocols that take time to connect() as it doesn't
2026 * include the -EINPROGRESS status for such sockets.
2029 int __sys_connect_file(struct file
*file
, struct sockaddr_storage
*address
,
2030 int addrlen
, int file_flags
)
2032 struct socket
*sock
;
2035 sock
= sock_from_file(file
);
2042 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
2046 err
= READ_ONCE(sock
->ops
)->connect(sock
, (struct sockaddr
*)address
,
2047 addrlen
, sock
->file
->f_flags
| file_flags
);
2052 int __sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
2059 struct sockaddr_storage address
;
2061 ret
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
2063 ret
= __sys_connect_file(f
.file
, &address
, addrlen
, 0);
2070 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
2073 return __sys_connect(fd
, uservaddr
, addrlen
);
2077 * Get the local address ('name') of a socket object. Move the obtained
2078 * name to user space.
2081 int __sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
2082 int __user
*usockaddr_len
)
2084 struct socket
*sock
;
2085 struct sockaddr_storage address
;
2086 int err
, fput_needed
;
2088 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2092 err
= security_socket_getsockname(sock
);
2096 err
= READ_ONCE(sock
->ops
)->getname(sock
, (struct sockaddr
*)&address
, 0);
2099 /* "err" is actually length in this case */
2100 err
= move_addr_to_user(&address
, err
, usockaddr
, usockaddr_len
);
2103 fput_light(sock
->file
, fput_needed
);
2108 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
2109 int __user
*, usockaddr_len
)
2111 return __sys_getsockname(fd
, usockaddr
, usockaddr_len
);
2115 * Get the remote address ('name') of a socket object. Move the obtained
2116 * name to user space.
2119 int __sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
2120 int __user
*usockaddr_len
)
2122 struct socket
*sock
;
2123 struct sockaddr_storage address
;
2124 int err
, fput_needed
;
2126 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2128 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
2130 err
= security_socket_getpeername(sock
);
2132 fput_light(sock
->file
, fput_needed
);
2136 err
= ops
->getname(sock
, (struct sockaddr
*)&address
, 1);
2138 /* "err" is actually length in this case */
2139 err
= move_addr_to_user(&address
, err
, usockaddr
,
2141 fput_light(sock
->file
, fput_needed
);
2146 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
2147 int __user
*, usockaddr_len
)
2149 return __sys_getpeername(fd
, usockaddr
, usockaddr_len
);
2153 * Send a datagram to a given address. We move the address into kernel
2154 * space and check the user space data area is readable before invoking
2157 int __sys_sendto(int fd
, void __user
*buff
, size_t len
, unsigned int flags
,
2158 struct sockaddr __user
*addr
, int addr_len
)
2160 struct socket
*sock
;
2161 struct sockaddr_storage address
;
2167 err
= import_single_range(ITER_SOURCE
, buff
, len
, &iov
, &msg
.msg_iter
);
2170 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2174 msg
.msg_name
= NULL
;
2175 msg
.msg_control
= NULL
;
2176 msg
.msg_controllen
= 0;
2177 msg
.msg_namelen
= 0;
2178 msg
.msg_ubuf
= NULL
;
2180 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
2183 msg
.msg_name
= (struct sockaddr
*)&address
;
2184 msg
.msg_namelen
= addr_len
;
2186 flags
&= ~MSG_INTERNAL_SENDMSG_FLAGS
;
2187 if (sock
->file
->f_flags
& O_NONBLOCK
)
2188 flags
|= MSG_DONTWAIT
;
2189 msg
.msg_flags
= flags
;
2190 err
= __sock_sendmsg(sock
, &msg
);
2193 fput_light(sock
->file
, fput_needed
);
2198 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
2199 unsigned int, flags
, struct sockaddr __user
*, addr
,
2202 return __sys_sendto(fd
, buff
, len
, flags
, addr
, addr_len
);
2206 * Send a datagram down a socket.
2209 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
2210 unsigned int, flags
)
2212 return __sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
2216 * Receive a frame from the socket and optionally record the address of the
2217 * sender. We verify the buffers are writable and if needed move the
2218 * sender address from kernel to user space.
2220 int __sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
, unsigned int flags
,
2221 struct sockaddr __user
*addr
, int __user
*addr_len
)
2223 struct sockaddr_storage address
;
2224 struct msghdr msg
= {
2225 /* Save some cycles and don't copy the address if not needed */
2226 .msg_name
= addr
? (struct sockaddr
*)&address
: NULL
,
2228 struct socket
*sock
;
2233 err
= import_single_range(ITER_DEST
, ubuf
, size
, &iov
, &msg
.msg_iter
);
2236 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2240 if (sock
->file
->f_flags
& O_NONBLOCK
)
2241 flags
|= MSG_DONTWAIT
;
2242 err
= sock_recvmsg(sock
, &msg
, flags
);
2244 if (err
>= 0 && addr
!= NULL
) {
2245 err2
= move_addr_to_user(&address
,
2246 msg
.msg_namelen
, addr
, addr_len
);
2251 fput_light(sock
->file
, fput_needed
);
2256 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
2257 unsigned int, flags
, struct sockaddr __user
*, addr
,
2258 int __user
*, addr_len
)
2260 return __sys_recvfrom(fd
, ubuf
, size
, flags
, addr
, addr_len
);
2264 * Receive a datagram from a socket.
2267 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
2268 unsigned int, flags
)
2270 return __sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
2273 static bool sock_use_custom_sol_socket(const struct socket
*sock
)
2275 return test_bit(SOCK_CUSTOM_SOCKOPT
, &sock
->flags
);
2278 int do_sock_setsockopt(struct socket
*sock
, bool compat
, int level
,
2279 int optname
, sockptr_t optval
, int optlen
)
2281 const struct proto_ops
*ops
;
2282 char *kernel_optval
= NULL
;
2288 err
= security_socket_setsockopt(sock
, level
, optname
);
2293 err
= BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock
->sk
, &level
, &optname
,
2304 optval
= KERNEL_SOCKPTR(kernel_optval
);
2305 ops
= READ_ONCE(sock
->ops
);
2306 if (level
== SOL_SOCKET
&& !sock_use_custom_sol_socket(sock
))
2307 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2308 else if (unlikely(!ops
->setsockopt
))
2311 err
= ops
->setsockopt(sock
, level
, optname
, optval
,
2313 kfree(kernel_optval
);
2317 EXPORT_SYMBOL(do_sock_setsockopt
);
2319 /* Set a socket option. Because we don't know the option lengths we have
2320 * to pass the user mode parameter for the protocols to sort out.
2322 int __sys_setsockopt(int fd
, int level
, int optname
, char __user
*user_optval
,
2325 sockptr_t optval
= USER_SOCKPTR(user_optval
);
2326 bool compat
= in_compat_syscall();
2327 int err
, fput_needed
;
2328 struct socket
*sock
;
2330 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2334 err
= do_sock_setsockopt(sock
, compat
, level
, optname
, optval
, optlen
);
2336 fput_light(sock
->file
, fput_needed
);
2340 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
2341 char __user
*, optval
, int, optlen
)
2343 return __sys_setsockopt(fd
, level
, optname
, optval
, optlen
);
2346 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level
,
2349 int do_sock_getsockopt(struct socket
*sock
, bool compat
, int level
,
2350 int optname
, sockptr_t optval
, sockptr_t optlen
)
2352 int max_optlen __maybe_unused
;
2353 const struct proto_ops
*ops
;
2356 err
= security_socket_getsockopt(sock
, level
, optname
);
2361 max_optlen
= BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen
);
2363 ops
= READ_ONCE(sock
->ops
);
2364 if (level
== SOL_SOCKET
) {
2365 err
= sk_getsockopt(sock
->sk
, level
, optname
, optval
, optlen
);
2366 } else if (unlikely(!ops
->getsockopt
)) {
2369 if (WARN_ONCE(optval
.is_kernel
|| optlen
.is_kernel
,
2370 "Invalid argument type"))
2373 err
= ops
->getsockopt(sock
, level
, optname
, optval
.user
,
2378 err
= BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock
->sk
, level
, optname
,
2379 optval
, optlen
, max_optlen
,
2384 EXPORT_SYMBOL(do_sock_getsockopt
);
2387 * Get a socket option. Because we don't know the option lengths we have
2388 * to pass a user mode parameter for the protocols to sort out.
2390 int __sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
,
2393 int err
, fput_needed
;
2394 struct socket
*sock
;
2397 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2401 compat
= in_compat_syscall();
2402 err
= do_sock_getsockopt(sock
, compat
, level
, optname
,
2403 USER_SOCKPTR(optval
), USER_SOCKPTR(optlen
));
2405 fput_light(sock
->file
, fput_needed
);
2409 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
2410 char __user
*, optval
, int __user
*, optlen
)
2412 return __sys_getsockopt(fd
, level
, optname
, optval
, optlen
);
2416 * Shutdown a socket.
2419 int __sys_shutdown_sock(struct socket
*sock
, int how
)
2423 err
= security_socket_shutdown(sock
, how
);
2425 err
= READ_ONCE(sock
->ops
)->shutdown(sock
, how
);
2430 int __sys_shutdown(int fd
, int how
)
2432 int err
, fput_needed
;
2433 struct socket
*sock
;
2435 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2437 err
= __sys_shutdown_sock(sock
, how
);
2438 fput_light(sock
->file
, fput_needed
);
2443 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
2445 return __sys_shutdown(fd
, how
);
2448 /* A couple of helpful macros for getting the address of the 32/64 bit
2449 * fields which are the same type (int / unsigned) on our platforms.
2451 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2452 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2453 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2455 struct used_address
{
2456 struct sockaddr_storage name
;
2457 unsigned int name_len
;
2460 int __copy_msghdr(struct msghdr
*kmsg
,
2461 struct user_msghdr
*msg
,
2462 struct sockaddr __user
**save_addr
)
2466 kmsg
->msg_control_is_user
= true;
2467 kmsg
->msg_get_inq
= 0;
2468 kmsg
->msg_control_user
= msg
->msg_control
;
2469 kmsg
->msg_controllen
= msg
->msg_controllen
;
2470 kmsg
->msg_flags
= msg
->msg_flags
;
2472 kmsg
->msg_namelen
= msg
->msg_namelen
;
2474 kmsg
->msg_namelen
= 0;
2476 if (kmsg
->msg_namelen
< 0)
2479 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2480 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2483 *save_addr
= msg
->msg_name
;
2485 if (msg
->msg_name
&& kmsg
->msg_namelen
) {
2487 err
= move_addr_to_kernel(msg
->msg_name
,
2494 kmsg
->msg_name
= NULL
;
2495 kmsg
->msg_namelen
= 0;
2498 if (msg
->msg_iovlen
> UIO_MAXIOV
)
2501 kmsg
->msg_iocb
= NULL
;
2502 kmsg
->msg_ubuf
= NULL
;
2506 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
2507 struct user_msghdr __user
*umsg
,
2508 struct sockaddr __user
**save_addr
,
2511 struct user_msghdr msg
;
2514 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
2517 err
= __copy_msghdr(kmsg
, &msg
, save_addr
);
2521 err
= import_iovec(save_addr
? ITER_DEST
: ITER_SOURCE
,
2522 msg
.msg_iov
, msg
.msg_iovlen
,
2523 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
2524 return err
< 0 ? err
: 0;
2527 static int ____sys_sendmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2528 unsigned int flags
, struct used_address
*used_address
,
2529 unsigned int allowed_msghdr_flags
)
2531 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2532 __aligned(sizeof(__kernel_size_t
));
2533 /* 20 is size of ipv6_pktinfo */
2534 unsigned char *ctl_buf
= ctl
;
2540 if (msg_sys
->msg_controllen
> INT_MAX
)
2542 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
2543 ctl_len
= msg_sys
->msg_controllen
;
2544 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2546 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2550 ctl_buf
= msg_sys
->msg_control
;
2551 ctl_len
= msg_sys
->msg_controllen
;
2552 } else if (ctl_len
) {
2553 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2554 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2555 if (ctl_len
> sizeof(ctl
)) {
2556 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2557 if (ctl_buf
== NULL
)
2561 if (copy_from_user(ctl_buf
, msg_sys
->msg_control_user
, ctl_len
))
2563 msg_sys
->msg_control
= ctl_buf
;
2564 msg_sys
->msg_control_is_user
= false;
2566 flags
&= ~MSG_INTERNAL_SENDMSG_FLAGS
;
2567 msg_sys
->msg_flags
= flags
;
2569 if (sock
->file
->f_flags
& O_NONBLOCK
)
2570 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2572 * If this is sendmmsg() and current destination address is same as
2573 * previously succeeded address, omit asking LSM's decision.
2574 * used_address->name_len is initialized to UINT_MAX so that the first
2575 * destination address never matches.
2577 if (used_address
&& msg_sys
->msg_name
&&
2578 used_address
->name_len
== msg_sys
->msg_namelen
&&
2579 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2580 used_address
->name_len
)) {
2581 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2584 err
= __sock_sendmsg(sock
, msg_sys
);
2586 * If this is sendmmsg() and sending to current destination address was
2587 * successful, remember it.
2589 if (used_address
&& err
>= 0) {
2590 used_address
->name_len
= msg_sys
->msg_namelen
;
2591 if (msg_sys
->msg_name
)
2592 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2593 used_address
->name_len
);
2598 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2603 int sendmsg_copy_msghdr(struct msghdr
*msg
,
2604 struct user_msghdr __user
*umsg
, unsigned flags
,
2609 if (flags
& MSG_CMSG_COMPAT
) {
2610 struct compat_msghdr __user
*msg_compat
;
2612 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2613 err
= get_compat_msghdr(msg
, msg_compat
, NULL
, iov
);
2615 err
= copy_msghdr_from_user(msg
, umsg
, NULL
, iov
);
2623 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2624 struct msghdr
*msg_sys
, unsigned int flags
,
2625 struct used_address
*used_address
,
2626 unsigned int allowed_msghdr_flags
)
2628 struct sockaddr_storage address
;
2629 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2632 msg_sys
->msg_name
= &address
;
2634 err
= sendmsg_copy_msghdr(msg_sys
, msg
, flags
, &iov
);
2638 err
= ____sys_sendmsg(sock
, msg_sys
, flags
, used_address
,
2639 allowed_msghdr_flags
);
2645 * BSD sendmsg interface
2647 long __sys_sendmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2650 return ____sys_sendmsg(sock
, msg
, flags
, NULL
, 0);
2653 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2654 bool forbid_cmsg_compat
)
2656 int fput_needed
, err
;
2657 struct msghdr msg_sys
;
2658 struct socket
*sock
;
2660 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2663 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2667 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2669 fput_light(sock
->file
, fput_needed
);
2674 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2676 return __sys_sendmsg(fd
, msg
, flags
, true);
2680 * Linux sendmmsg interface
2683 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2684 unsigned int flags
, bool forbid_cmsg_compat
)
2686 int fput_needed
, err
, datagrams
;
2687 struct socket
*sock
;
2688 struct mmsghdr __user
*entry
;
2689 struct compat_mmsghdr __user
*compat_entry
;
2690 struct msghdr msg_sys
;
2691 struct used_address used_address
;
2692 unsigned int oflags
= flags
;
2694 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2697 if (vlen
> UIO_MAXIOV
)
2702 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2706 used_address
.name_len
= UINT_MAX
;
2708 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2712 while (datagrams
< vlen
) {
2713 if (datagrams
== vlen
- 1)
2716 if (MSG_CMSG_COMPAT
& flags
) {
2717 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2718 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2721 err
= __put_user(err
, &compat_entry
->msg_len
);
2724 err
= ___sys_sendmsg(sock
,
2725 (struct user_msghdr __user
*)entry
,
2726 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2729 err
= put_user(err
, &entry
->msg_len
);
2736 if (msg_data_left(&msg_sys
))
2741 fput_light(sock
->file
, fput_needed
);
2743 /* We only return an error if no datagrams were able to be sent */
2750 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2751 unsigned int, vlen
, unsigned int, flags
)
2753 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
, true);
2756 int recvmsg_copy_msghdr(struct msghdr
*msg
,
2757 struct user_msghdr __user
*umsg
, unsigned flags
,
2758 struct sockaddr __user
**uaddr
,
2763 if (MSG_CMSG_COMPAT
& flags
) {
2764 struct compat_msghdr __user
*msg_compat
;
2766 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2767 err
= get_compat_msghdr(msg
, msg_compat
, uaddr
, iov
);
2769 err
= copy_msghdr_from_user(msg
, umsg
, uaddr
, iov
);
2777 static int ____sys_recvmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2778 struct user_msghdr __user
*msg
,
2779 struct sockaddr __user
*uaddr
,
2780 unsigned int flags
, int nosec
)
2782 struct compat_msghdr __user
*msg_compat
=
2783 (struct compat_msghdr __user
*) msg
;
2784 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2785 struct sockaddr_storage addr
;
2786 unsigned long cmsg_ptr
;
2790 msg_sys
->msg_name
= &addr
;
2791 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2792 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2794 /* We assume all kernel code knows the size of sockaddr_storage */
2795 msg_sys
->msg_namelen
= 0;
2797 if (sock
->file
->f_flags
& O_NONBLOCK
)
2798 flags
|= MSG_DONTWAIT
;
2800 if (unlikely(nosec
))
2801 err
= sock_recvmsg_nosec(sock
, msg_sys
, flags
);
2803 err
= sock_recvmsg(sock
, msg_sys
, flags
);
2809 if (uaddr
!= NULL
) {
2810 err
= move_addr_to_user(&addr
,
2811 msg_sys
->msg_namelen
, uaddr
,
2816 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2820 if (MSG_CMSG_COMPAT
& flags
)
2821 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2822 &msg_compat
->msg_controllen
);
2824 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2825 &msg
->msg_controllen
);
2833 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2834 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2836 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2837 /* user mode address pointers */
2838 struct sockaddr __user
*uaddr
;
2841 err
= recvmsg_copy_msghdr(msg_sys
, msg
, flags
, &uaddr
, &iov
);
2845 err
= ____sys_recvmsg(sock
, msg_sys
, msg
, uaddr
, flags
, nosec
);
2851 * BSD recvmsg interface
2854 long __sys_recvmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2855 struct user_msghdr __user
*umsg
,
2856 struct sockaddr __user
*uaddr
, unsigned int flags
)
2858 return ____sys_recvmsg(sock
, msg
, umsg
, uaddr
, flags
, 0);
2861 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2862 bool forbid_cmsg_compat
)
2864 int fput_needed
, err
;
2865 struct msghdr msg_sys
;
2866 struct socket
*sock
;
2868 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2871 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2875 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2877 fput_light(sock
->file
, fput_needed
);
2882 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2883 unsigned int, flags
)
2885 return __sys_recvmsg(fd
, msg
, flags
, true);
2889 * Linux recvmmsg interface
2892 static int do_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2893 unsigned int vlen
, unsigned int flags
,
2894 struct timespec64
*timeout
)
2896 int fput_needed
, err
, datagrams
;
2897 struct socket
*sock
;
2898 struct mmsghdr __user
*entry
;
2899 struct compat_mmsghdr __user
*compat_entry
;
2900 struct msghdr msg_sys
;
2901 struct timespec64 end_time
;
2902 struct timespec64 timeout64
;
2905 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2911 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2915 if (likely(!(flags
& MSG_ERRQUEUE
))) {
2916 err
= sock_error(sock
->sk
);
2924 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2926 while (datagrams
< vlen
) {
2928 * No need to ask LSM for more than the first datagram.
2930 if (MSG_CMSG_COMPAT
& flags
) {
2931 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2932 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2936 err
= __put_user(err
, &compat_entry
->msg_len
);
2939 err
= ___sys_recvmsg(sock
,
2940 (struct user_msghdr __user
*)entry
,
2941 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2945 err
= put_user(err
, &entry
->msg_len
);
2953 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2954 if (flags
& MSG_WAITFORONE
)
2955 flags
|= MSG_DONTWAIT
;
2958 ktime_get_ts64(&timeout64
);
2959 *timeout
= timespec64_sub(end_time
, timeout64
);
2960 if (timeout
->tv_sec
< 0) {
2961 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2965 /* Timeout, return less than vlen datagrams */
2966 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2970 /* Out of band data, return right away */
2971 if (msg_sys
.msg_flags
& MSG_OOB
)
2979 if (datagrams
== 0) {
2985 * We may return less entries than requested (vlen) if the
2986 * sock is non block and there aren't enough datagrams...
2988 if (err
!= -EAGAIN
) {
2990 * ... or if recvmsg returns an error after we
2991 * received some datagrams, where we record the
2992 * error to return on the next call or if the
2993 * app asks about it using getsockopt(SO_ERROR).
2995 WRITE_ONCE(sock
->sk
->sk_err
, -err
);
2998 fput_light(sock
->file
, fput_needed
);
3003 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
3004 unsigned int vlen
, unsigned int flags
,
3005 struct __kernel_timespec __user
*timeout
,
3006 struct old_timespec32 __user
*timeout32
)
3009 struct timespec64 timeout_sys
;
3011 if (timeout
&& get_timespec64(&timeout_sys
, timeout
))
3014 if (timeout32
&& get_old_timespec32(&timeout_sys
, timeout32
))
3017 if (!timeout
&& !timeout32
)
3018 return do_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
3020 datagrams
= do_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
3025 if (timeout
&& put_timespec64(&timeout_sys
, timeout
))
3026 datagrams
= -EFAULT
;
3028 if (timeout32
&& put_old_timespec32(&timeout_sys
, timeout32
))
3029 datagrams
= -EFAULT
;
3034 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
3035 unsigned int, vlen
, unsigned int, flags
,
3036 struct __kernel_timespec __user
*, timeout
)
3038 if (flags
& MSG_CMSG_COMPAT
)
3041 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, timeout
, NULL
);
3044 #ifdef CONFIG_COMPAT_32BIT_TIME
3045 SYSCALL_DEFINE5(recvmmsg_time32
, int, fd
, struct mmsghdr __user
*, mmsg
,
3046 unsigned int, vlen
, unsigned int, flags
,
3047 struct old_timespec32 __user
*, timeout
)
3049 if (flags
& MSG_CMSG_COMPAT
)
3052 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
, timeout
);
3056 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3057 /* Argument list sizes for sys_socketcall */
3058 #define AL(x) ((x) * sizeof(unsigned long))
3059 static const unsigned char nargs
[21] = {
3060 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3061 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3062 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3069 * System call vectors.
3071 * Argument checking cleaned up. Saved 20% in size.
3072 * This function doesn't need to set the kernel lock because
3073 * it is set by the callees.
3076 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
3078 unsigned long a
[AUDITSC_ARGS
];
3079 unsigned long a0
, a1
;
3083 if (call
< 1 || call
> SYS_SENDMMSG
)
3085 call
= array_index_nospec(call
, SYS_SENDMMSG
+ 1);
3088 if (len
> sizeof(a
))
3091 /* copy_from_user should be SMP safe. */
3092 if (copy_from_user(a
, args
, len
))
3095 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
3104 err
= __sys_socket(a0
, a1
, a
[2]);
3107 err
= __sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3110 err
= __sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3113 err
= __sys_listen(a0
, a1
);
3116 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3117 (int __user
*)a
[2], 0);
3119 case SYS_GETSOCKNAME
:
3121 __sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
3122 (int __user
*)a
[2]);
3124 case SYS_GETPEERNAME
:
3126 __sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
3127 (int __user
*)a
[2]);
3129 case SYS_SOCKETPAIR
:
3130 err
= __sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
3133 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3137 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3138 (struct sockaddr __user
*)a
[4], a
[5]);
3141 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3145 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3146 (struct sockaddr __user
*)a
[4],
3147 (int __user
*)a
[5]);
3150 err
= __sys_shutdown(a0
, a1
);
3152 case SYS_SETSOCKOPT
:
3153 err
= __sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3156 case SYS_GETSOCKOPT
:
3158 __sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3159 (int __user
*)a
[4]);
3162 err
= __sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
,
3166 err
= __sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2],
3170 err
= __sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
,
3174 if (IS_ENABLED(CONFIG_64BIT
))
3175 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3177 (struct __kernel_timespec __user
*)a
[4],
3180 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3182 (struct old_timespec32 __user
*)a
[4]);
3185 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3186 (int __user
*)a
[2], a
[3]);
3195 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3198 * sock_register - add a socket protocol handler
3199 * @ops: description of protocol
3201 * This function is called by a protocol handler that wants to
3202 * advertise its address family, and have it linked into the
3203 * socket interface. The value ops->family corresponds to the
3204 * socket system call protocol family.
3206 int sock_register(const struct net_proto_family
*ops
)
3210 if (ops
->family
>= NPROTO
) {
3211 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
3215 spin_lock(&net_family_lock
);
3216 if (rcu_dereference_protected(net_families
[ops
->family
],
3217 lockdep_is_held(&net_family_lock
)))
3220 rcu_assign_pointer(net_families
[ops
->family
], ops
);
3223 spin_unlock(&net_family_lock
);
3225 pr_info("NET: Registered %s protocol family\n", pf_family_names
[ops
->family
]);
3228 EXPORT_SYMBOL(sock_register
);
3231 * sock_unregister - remove a protocol handler
3232 * @family: protocol family to remove
3234 * This function is called by a protocol handler that wants to
3235 * remove its address family, and have it unlinked from the
3236 * new socket creation.
3238 * If protocol handler is a module, then it can use module reference
3239 * counts to protect against new references. If protocol handler is not
3240 * a module then it needs to provide its own protection in
3241 * the ops->create routine.
3243 void sock_unregister(int family
)
3245 BUG_ON(family
< 0 || family
>= NPROTO
);
3247 spin_lock(&net_family_lock
);
3248 RCU_INIT_POINTER(net_families
[family
], NULL
);
3249 spin_unlock(&net_family_lock
);
3253 pr_info("NET: Unregistered %s protocol family\n", pf_family_names
[family
]);
3255 EXPORT_SYMBOL(sock_unregister
);
3257 bool sock_is_registered(int family
)
3259 return family
< NPROTO
&& rcu_access_pointer(net_families
[family
]);
3262 static int __init
sock_init(void)
3266 * Initialize the network sysctl infrastructure.
3268 err
= net_sysctl_init();
3273 * Initialize skbuff SLAB cache
3278 * Initialize the protocols module.
3283 err
= register_filesystem(&sock_fs_type
);
3286 sock_mnt
= kern_mount(&sock_fs_type
);
3287 if (IS_ERR(sock_mnt
)) {
3288 err
= PTR_ERR(sock_mnt
);
3292 /* The real protocol initialization is performed in later initcalls.
3295 #ifdef CONFIG_NETFILTER
3296 err
= netfilter_init();
3301 ptp_classifier_init();
3307 unregister_filesystem(&sock_fs_type
);
3311 core_initcall(sock_init
); /* early initcall */
3313 #ifdef CONFIG_PROC_FS
3314 void socket_seq_show(struct seq_file
*seq
)
3316 seq_printf(seq
, "sockets: used %d\n",
3317 sock_inuse_get(seq
->private));
3319 #endif /* CONFIG_PROC_FS */
3321 /* Handle the fact that while struct ifreq has the same *layout* on
3322 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3323 * which are handled elsewhere, it still has different *size* due to
3324 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3325 * resulting in struct ifreq being 32 and 40 bytes respectively).
3326 * As a result, if the struct happens to be at the end of a page and
3327 * the next page isn't readable/writable, we get a fault. To prevent
3328 * that, copy back and forth to the full size.
3330 int get_user_ifreq(struct ifreq
*ifr
, void __user
**ifrdata
, void __user
*arg
)
3332 if (in_compat_syscall()) {
3333 struct compat_ifreq
*ifr32
= (struct compat_ifreq
*)ifr
;
3335 memset(ifr
, 0, sizeof(*ifr
));
3336 if (copy_from_user(ifr32
, arg
, sizeof(*ifr32
)))
3340 *ifrdata
= compat_ptr(ifr32
->ifr_data
);
3345 if (copy_from_user(ifr
, arg
, sizeof(*ifr
)))
3349 *ifrdata
= ifr
->ifr_data
;
3353 EXPORT_SYMBOL(get_user_ifreq
);
3355 int put_user_ifreq(struct ifreq
*ifr
, void __user
*arg
)
3357 size_t size
= sizeof(*ifr
);
3359 if (in_compat_syscall())
3360 size
= sizeof(struct compat_ifreq
);
3362 if (copy_to_user(arg
, ifr
, size
))
3367 EXPORT_SYMBOL(put_user_ifreq
);
3369 #ifdef CONFIG_COMPAT
3370 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3372 compat_uptr_t uptr32
;
3377 if (get_user_ifreq(&ifr
, NULL
, uifr32
))
3380 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
3383 saved
= ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
;
3384 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= compat_ptr(uptr32
);
3386 err
= dev_ioctl(net
, SIOCWANDEV
, &ifr
, NULL
, NULL
);
3388 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= saved
;
3389 if (put_user_ifreq(&ifr
, uifr32
))
3395 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3396 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3397 struct compat_ifreq __user
*u_ifreq32
)
3402 if (!is_socket_ioctl_cmd(cmd
))
3404 if (get_user_ifreq(&ifreq
, &data
, u_ifreq32
))
3406 ifreq
.ifr_data
= data
;
3408 return dev_ioctl(net
, cmd
, &ifreq
, data
, NULL
);
3411 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3412 unsigned int cmd
, unsigned long arg
)
3414 void __user
*argp
= compat_ptr(arg
);
3415 struct sock
*sk
= sock
->sk
;
3416 struct net
*net
= sock_net(sk
);
3417 const struct proto_ops
*ops
;
3419 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3420 return sock_ioctl(file
, cmd
, (unsigned long)argp
);
3424 return compat_siocwandev(net
, argp
);
3425 case SIOCGSTAMP_OLD
:
3426 case SIOCGSTAMPNS_OLD
:
3427 ops
= READ_ONCE(sock
->ops
);
3428 if (!ops
->gettstamp
)
3429 return -ENOIOCTLCMD
;
3430 return ops
->gettstamp(sock
, argp
, cmd
== SIOCGSTAMP_OLD
,
3431 !COMPAT_USE_64BIT_TIME
);
3434 case SIOCBONDSLAVEINFOQUERY
:
3435 case SIOCBONDINFOQUERY
:
3438 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3449 case SIOCGSTAMP_NEW
:
3450 case SIOCGSTAMPNS_NEW
:
3454 return sock_ioctl(file
, cmd
, arg
);
3473 case SIOCSIFHWBROADCAST
:
3475 case SIOCGIFBRDADDR
:
3476 case SIOCSIFBRDADDR
:
3477 case SIOCGIFDSTADDR
:
3478 case SIOCSIFDSTADDR
:
3479 case SIOCGIFNETMASK
:
3480 case SIOCSIFNETMASK
:
3492 case SIOCBONDENSLAVE
:
3493 case SIOCBONDRELEASE
:
3494 case SIOCBONDSETHWADDR
:
3495 case SIOCBONDCHANGEACTIVE
:
3502 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3505 return -ENOIOCTLCMD
;
3508 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3511 struct socket
*sock
= file
->private_data
;
3512 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
3513 int ret
= -ENOIOCTLCMD
;
3520 if (ops
->compat_ioctl
)
3521 ret
= ops
->compat_ioctl(sock
, cmd
, arg
);
3523 if (ret
== -ENOIOCTLCMD
&&
3524 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3525 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3527 if (ret
== -ENOIOCTLCMD
)
3528 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3535 * kernel_bind - bind an address to a socket (kernel space)
3538 * @addrlen: length of address
3540 * Returns 0 or an error.
3543 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3545 struct sockaddr_storage address
;
3547 memcpy(&address
, addr
, addrlen
);
3549 return READ_ONCE(sock
->ops
)->bind(sock
, (struct sockaddr
*)&address
,
3552 EXPORT_SYMBOL(kernel_bind
);
3555 * kernel_listen - move socket to listening state (kernel space)
3557 * @backlog: pending connections queue size
3559 * Returns 0 or an error.
3562 int kernel_listen(struct socket
*sock
, int backlog
)
3564 return READ_ONCE(sock
->ops
)->listen(sock
, backlog
);
3566 EXPORT_SYMBOL(kernel_listen
);
3569 * kernel_accept - accept a connection (kernel space)
3570 * @sock: listening socket
3571 * @newsock: new connected socket
3574 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3575 * If it fails, @newsock is guaranteed to be %NULL.
3576 * Returns 0 or an error.
3579 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3581 struct sock
*sk
= sock
->sk
;
3582 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
3585 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3590 err
= ops
->accept(sock
, *newsock
, flags
, true);
3592 sock_release(*newsock
);
3597 (*newsock
)->ops
= ops
;
3598 __module_get(ops
->owner
);
3603 EXPORT_SYMBOL(kernel_accept
);
3606 * kernel_connect - connect a socket (kernel space)
3609 * @addrlen: address length
3610 * @flags: flags (O_NONBLOCK, ...)
3612 * For datagram sockets, @addr is the address to which datagrams are sent
3613 * by default, and the only address from which datagrams are received.
3614 * For stream sockets, attempts to connect to @addr.
3615 * Returns 0 or an error code.
3618 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3621 struct sockaddr_storage address
;
3623 memcpy(&address
, addr
, addrlen
);
3625 return READ_ONCE(sock
->ops
)->connect(sock
, (struct sockaddr
*)&address
,
3628 EXPORT_SYMBOL(kernel_connect
);
3631 * kernel_getsockname - get the address which the socket is bound (kernel space)
3633 * @addr: address holder
3635 * Fills the @addr pointer with the address which the socket is bound.
3636 * Returns the length of the address in bytes or an error code.
3639 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
)
3641 return READ_ONCE(sock
->ops
)->getname(sock
, addr
, 0);
3643 EXPORT_SYMBOL(kernel_getsockname
);
3646 * kernel_getpeername - get the address which the socket is connected (kernel space)
3648 * @addr: address holder
3650 * Fills the @addr pointer with the address which the socket is connected.
3651 * Returns the length of the address in bytes or an error code.
3654 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
)
3656 return READ_ONCE(sock
->ops
)->getname(sock
, addr
, 1);
3658 EXPORT_SYMBOL(kernel_getpeername
);
3661 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3663 * @how: connection part
3665 * Returns 0 or an error.
3668 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3670 return READ_ONCE(sock
->ops
)->shutdown(sock
, how
);
3672 EXPORT_SYMBOL(kernel_sock_shutdown
);
3675 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3678 * This routine returns the IP overhead imposed by a socket i.e.
3679 * the length of the underlying IP header, depending on whether
3680 * this is an IPv4 or IPv6 socket and the length from IP options turned
3681 * on at the socket. Assumes that the caller has a lock on the socket.
3684 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3686 struct inet_sock
*inet
;
3687 struct ip_options_rcu
*opt
;
3689 #if IS_ENABLED(CONFIG_IPV6)
3690 struct ipv6_pinfo
*np
;
3691 struct ipv6_txoptions
*optv6
= NULL
;
3692 #endif /* IS_ENABLED(CONFIG_IPV6) */
3697 switch (sk
->sk_family
) {
3700 overhead
+= sizeof(struct iphdr
);
3701 opt
= rcu_dereference_protected(inet
->inet_opt
,
3702 sock_owned_by_user(sk
));
3704 overhead
+= opt
->opt
.optlen
;
3706 #if IS_ENABLED(CONFIG_IPV6)
3709 overhead
+= sizeof(struct ipv6hdr
);
3711 optv6
= rcu_dereference_protected(np
->opt
,
3712 sock_owned_by_user(sk
));
3714 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3716 #endif /* IS_ENABLED(CONFIG_IPV6) */
3717 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3721 EXPORT_SYMBOL(kernel_sock_ip_overhead
);