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/net.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
|
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
;
760 int save_len
= msg
->msg_namelen
;
764 memcpy(&address
, msg
->msg_name
, msg
->msg_namelen
);
765 msg
->msg_name
= &address
;
768 ret
= __sock_sendmsg(sock
, msg
);
769 msg
->msg_name
= save_addr
;
770 msg
->msg_namelen
= save_len
;
774 EXPORT_SYMBOL(sock_sendmsg
);
777 * kernel_sendmsg - send a message through @sock (kernel-space)
779 * @msg: message header
781 * @num: vec array length
782 * @size: total message data size
784 * Builds the message data with @vec and sends it through @sock.
785 * Returns the number of bytes sent, or an error code.
788 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
789 struct kvec
*vec
, size_t num
, size_t size
)
791 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
792 return sock_sendmsg(sock
, msg
);
794 EXPORT_SYMBOL(kernel_sendmsg
);
797 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
799 * @msg: message header
800 * @vec: output s/g array
801 * @num: output s/g array length
802 * @size: total message data size
804 * Builds the message data with @vec and sends it through @sock.
805 * Returns the number of bytes sent, or an error code.
806 * Caller must hold @sk.
809 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
810 struct kvec
*vec
, size_t num
, size_t size
)
812 struct socket
*sock
= sk
->sk_socket
;
813 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
815 if (!ops
->sendmsg_locked
)
816 return sock_no_sendmsg_locked(sk
, msg
, size
);
818 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
820 return ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
822 EXPORT_SYMBOL(kernel_sendmsg_locked
);
824 static bool skb_is_err_queue(const struct sk_buff
*skb
)
826 /* pkt_type of skbs enqueued on the error queue are set to
827 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
828 * in recvmsg, since skbs received on a local socket will never
829 * have a pkt_type of PACKET_OUTGOING.
831 return skb
->pkt_type
== PACKET_OUTGOING
;
834 /* On transmit, software and hardware timestamps are returned independently.
835 * As the two skb clones share the hardware timestamp, which may be updated
836 * before the software timestamp is received, a hardware TX timestamp may be
837 * returned only if there is no software TX timestamp. Ignore false software
838 * timestamps, which may be made in the __sock_recv_timestamp() call when the
839 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
840 * hardware timestamp.
842 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
844 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
847 static ktime_t
get_timestamp(struct sock
*sk
, struct sk_buff
*skb
, int *if_index
)
849 bool cycles
= READ_ONCE(sk
->sk_tsflags
) & SOF_TIMESTAMPING_BIND_PHC
;
850 struct skb_shared_hwtstamps
*shhwtstamps
= skb_hwtstamps(skb
);
851 struct net_device
*orig_dev
;
855 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
857 *if_index
= orig_dev
->ifindex
;
858 hwtstamp
= netdev_get_tstamp(orig_dev
, shhwtstamps
, cycles
);
860 hwtstamp
= shhwtstamps
->hwtstamp
;
867 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
,
870 struct scm_ts_pktinfo ts_pktinfo
;
871 struct net_device
*orig_dev
;
873 if (!skb_mac_header_was_set(skb
))
876 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
880 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
882 if_index
= orig_dev
->ifindex
;
885 ts_pktinfo
.if_index
= if_index
;
887 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
888 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
889 sizeof(ts_pktinfo
), &ts_pktinfo
);
893 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
895 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
898 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
899 int new_tstamp
= sock_flag(sk
, SOCK_TSTAMP_NEW
);
900 struct scm_timestamping_internal tss
;
901 int empty
= 1, false_tstamp
= 0;
902 struct skb_shared_hwtstamps
*shhwtstamps
=
908 /* Race occurred between timestamp enabling and packet
909 receiving. Fill in the current time for now. */
910 if (need_software_tstamp
&& skb
->tstamp
== 0) {
911 __net_timestamp(skb
);
915 if (need_software_tstamp
) {
916 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
918 struct __kernel_sock_timeval tv
;
920 skb_get_new_timestamp(skb
, &tv
);
921 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_NEW
,
924 struct __kernel_old_timeval tv
;
926 skb_get_timestamp(skb
, &tv
);
927 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_OLD
,
932 struct __kernel_timespec ts
;
934 skb_get_new_timestampns(skb
, &ts
);
935 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_NEW
,
938 struct __kernel_old_timespec ts
;
940 skb_get_timestampns(skb
, &ts
);
941 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_OLD
,
947 memset(&tss
, 0, sizeof(tss
));
948 tsflags
= READ_ONCE(sk
->sk_tsflags
);
949 if ((tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
950 ktime_to_timespec64_cond(skb
->tstamp
, tss
.ts
+ 0))
953 (tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
954 !skb_is_swtx_tstamp(skb
, false_tstamp
)) {
956 if (skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP_NETDEV
)
957 hwtstamp
= get_timestamp(sk
, skb
, &if_index
);
959 hwtstamp
= shhwtstamps
->hwtstamp
;
961 if (tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
962 hwtstamp
= ptp_convert_timestamp(&hwtstamp
,
963 READ_ONCE(sk
->sk_bind_phc
));
965 if (ktime_to_timespec64_cond(hwtstamp
, tss
.ts
+ 2)) {
968 if ((tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
969 !skb_is_err_queue(skb
))
970 put_ts_pktinfo(msg
, skb
, if_index
);
974 if (sock_flag(sk
, SOCK_TSTAMP_NEW
))
975 put_cmsg_scm_timestamping64(msg
, &tss
);
977 put_cmsg_scm_timestamping(msg
, &tss
);
979 if (skb_is_err_queue(skb
) && skb
->len
&&
980 SKB_EXT_ERR(skb
)->opt_stats
)
981 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
982 skb
->len
, skb
->data
);
985 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
987 #ifdef CONFIG_WIRELESS
988 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
993 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
995 if (!skb
->wifi_acked_valid
)
998 ack
= skb
->wifi_acked
;
1000 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
1002 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
1005 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
1006 struct sk_buff
*skb
)
1008 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
1009 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
1010 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
1013 static void sock_recv_mark(struct msghdr
*msg
, struct sock
*sk
,
1014 struct sk_buff
*skb
)
1016 if (sock_flag(sk
, SOCK_RCVMARK
) && skb
) {
1017 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1018 __u32 mark
= skb
->mark
;
1020 put_cmsg(msg
, SOL_SOCKET
, SO_MARK
, sizeof(__u32
), &mark
);
1024 void __sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
1025 struct sk_buff
*skb
)
1027 sock_recv_timestamp(msg
, sk
, skb
);
1028 sock_recv_drops(msg
, sk
, skb
);
1029 sock_recv_mark(msg
, sk
, skb
);
1031 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs
);
1033 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket
*, struct msghdr
*,
1035 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket
*, struct msghdr
*,
1038 static noinline
void call_trace_sock_recv_length(struct sock
*sk
, int ret
, int flags
)
1040 trace_sock_recv_length(sk
, ret
, flags
);
1043 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
1046 int ret
= INDIRECT_CALL_INET(READ_ONCE(sock
->ops
)->recvmsg
,
1048 inet_recvmsg
, sock
, msg
,
1049 msg_data_left(msg
), flags
);
1050 if (trace_sock_recv_length_enabled())
1051 call_trace_sock_recv_length(sock
->sk
, ret
, flags
);
1056 * sock_recvmsg - receive a message from @sock
1058 * @msg: message to receive
1059 * @flags: message flags
1061 * Receives @msg from @sock, passing through LSM. Returns the total number
1062 * of bytes received, or an error.
1064 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
1066 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
1068 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
1070 EXPORT_SYMBOL(sock_recvmsg
);
1073 * kernel_recvmsg - Receive a message from a socket (kernel space)
1074 * @sock: The socket to receive the message from
1075 * @msg: Received message
1076 * @vec: Input s/g array for message data
1077 * @num: Size of input s/g array
1078 * @size: Number of bytes to read
1079 * @flags: Message flags (MSG_DONTWAIT, etc...)
1081 * On return the msg structure contains the scatter/gather array passed in the
1082 * vec argument. The array is modified so that it consists of the unfilled
1083 * portion of the original array.
1085 * The returned value is the total number of bytes received, or an error.
1088 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
1089 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
1091 msg
->msg_control_is_user
= false;
1092 iov_iter_kvec(&msg
->msg_iter
, ITER_DEST
, vec
, num
, size
);
1093 return sock_recvmsg(sock
, msg
, flags
);
1095 EXPORT_SYMBOL(kernel_recvmsg
);
1097 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
1098 struct pipe_inode_info
*pipe
, size_t len
,
1101 struct socket
*sock
= file
->private_data
;
1102 const struct proto_ops
*ops
;
1104 ops
= READ_ONCE(sock
->ops
);
1105 if (unlikely(!ops
->splice_read
))
1106 return copy_splice_read(file
, ppos
, pipe
, len
, flags
);
1108 return ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
1111 static void sock_splice_eof(struct file
*file
)
1113 struct socket
*sock
= file
->private_data
;
1114 const struct proto_ops
*ops
;
1116 ops
= READ_ONCE(sock
->ops
);
1117 if (ops
->splice_eof
)
1118 ops
->splice_eof(sock
);
1121 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1123 struct file
*file
= iocb
->ki_filp
;
1124 struct socket
*sock
= file
->private_data
;
1125 struct msghdr msg
= {.msg_iter
= *to
,
1129 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1130 msg
.msg_flags
= MSG_DONTWAIT
;
1132 if (iocb
->ki_pos
!= 0)
1135 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
1138 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
1143 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1145 struct file
*file
= iocb
->ki_filp
;
1146 struct socket
*sock
= file
->private_data
;
1147 struct msghdr msg
= {.msg_iter
= *from
,
1151 if (iocb
->ki_pos
!= 0)
1154 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1155 msg
.msg_flags
= MSG_DONTWAIT
;
1157 if (sock
->type
== SOCK_SEQPACKET
)
1158 msg
.msg_flags
|= MSG_EOR
;
1160 res
= __sock_sendmsg(sock
, &msg
);
1161 *from
= msg
.msg_iter
;
1166 * Atomic setting of ioctl hooks to avoid race
1167 * with module unload.
1170 static DEFINE_MUTEX(br_ioctl_mutex
);
1171 static int (*br_ioctl_hook
)(struct net
*net
, struct net_bridge
*br
,
1172 unsigned int cmd
, struct ifreq
*ifr
,
1175 void brioctl_set(int (*hook
)(struct net
*net
, struct net_bridge
*br
,
1176 unsigned int cmd
, struct ifreq
*ifr
,
1179 mutex_lock(&br_ioctl_mutex
);
1180 br_ioctl_hook
= hook
;
1181 mutex_unlock(&br_ioctl_mutex
);
1183 EXPORT_SYMBOL(brioctl_set
);
1185 int br_ioctl_call(struct net
*net
, struct net_bridge
*br
, unsigned int cmd
,
1186 struct ifreq
*ifr
, void __user
*uarg
)
1191 request_module("bridge");
1193 mutex_lock(&br_ioctl_mutex
);
1195 err
= br_ioctl_hook(net
, br
, cmd
, ifr
, uarg
);
1196 mutex_unlock(&br_ioctl_mutex
);
1201 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1202 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1204 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1206 mutex_lock(&vlan_ioctl_mutex
);
1207 vlan_ioctl_hook
= hook
;
1208 mutex_unlock(&vlan_ioctl_mutex
);
1210 EXPORT_SYMBOL(vlan_ioctl_set
);
1212 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1213 unsigned int cmd
, unsigned long arg
)
1215 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
1219 void __user
*argp
= (void __user
*)arg
;
1222 err
= ops
->ioctl(sock
, cmd
, arg
);
1225 * If this ioctl is unknown try to hand it down
1226 * to the NIC driver.
1228 if (err
!= -ENOIOCTLCMD
)
1231 if (!is_socket_ioctl_cmd(cmd
))
1234 if (get_user_ifreq(&ifr
, &data
, argp
))
1236 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1237 if (!err
&& need_copyout
)
1238 if (put_user_ifreq(&ifr
, argp
))
1245 * With an ioctl, arg may well be a user mode pointer, but we don't know
1246 * what to do with it - that's up to the protocol still.
1249 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1251 const struct proto_ops
*ops
;
1252 struct socket
*sock
;
1254 void __user
*argp
= (void __user
*)arg
;
1258 sock
= file
->private_data
;
1259 ops
= READ_ONCE(sock
->ops
);
1262 if (unlikely(cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))) {
1266 if (get_user_ifreq(&ifr
, &data
, argp
))
1268 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1269 if (!err
&& need_copyout
)
1270 if (put_user_ifreq(&ifr
, argp
))
1273 #ifdef CONFIG_WEXT_CORE
1274 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1275 err
= wext_handle_ioctl(net
, cmd
, argp
);
1282 if (get_user(pid
, (int __user
*)argp
))
1284 err
= f_setown(sock
->file
, pid
, 1);
1288 err
= put_user(f_getown(sock
->file
),
1289 (int __user
*)argp
);
1295 err
= br_ioctl_call(net
, NULL
, cmd
, NULL
, argp
);
1300 if (!vlan_ioctl_hook
)
1301 request_module("8021q");
1303 mutex_lock(&vlan_ioctl_mutex
);
1304 if (vlan_ioctl_hook
)
1305 err
= vlan_ioctl_hook(net
, argp
);
1306 mutex_unlock(&vlan_ioctl_mutex
);
1310 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1313 err
= open_related_ns(&net
->ns
, get_net_ns
);
1315 case SIOCGSTAMP_OLD
:
1316 case SIOCGSTAMPNS_OLD
:
1317 if (!ops
->gettstamp
) {
1321 err
= ops
->gettstamp(sock
, argp
,
1322 cmd
== SIOCGSTAMP_OLD
,
1323 !IS_ENABLED(CONFIG_64BIT
));
1325 case SIOCGSTAMP_NEW
:
1326 case SIOCGSTAMPNS_NEW
:
1327 if (!ops
->gettstamp
) {
1331 err
= ops
->gettstamp(sock
, argp
,
1332 cmd
== SIOCGSTAMP_NEW
,
1337 err
= dev_ifconf(net
, argp
);
1341 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1348 * sock_create_lite - creates a socket
1349 * @family: protocol family (AF_INET, ...)
1350 * @type: communication type (SOCK_STREAM, ...)
1351 * @protocol: protocol (0, ...)
1354 * Creates a new socket and assigns it to @res, passing through LSM.
1355 * The new socket initialization is not complete, see kernel_accept().
1356 * Returns 0 or an error. On failure @res is set to %NULL.
1357 * This function internally uses GFP_KERNEL.
1360 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1363 struct socket
*sock
= NULL
;
1365 err
= security_socket_create(family
, type
, protocol
, 1);
1369 sock
= sock_alloc();
1376 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1388 EXPORT_SYMBOL(sock_create_lite
);
1390 /* No kernel lock held - perfect */
1391 static __poll_t
sock_poll(struct file
*file
, poll_table
*wait
)
1393 struct socket
*sock
= file
->private_data
;
1394 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
1395 __poll_t events
= poll_requested_events(wait
), flag
= 0;
1400 if (sk_can_busy_loop(sock
->sk
)) {
1401 /* poll once if requested by the syscall */
1402 if (events
& POLL_BUSY_LOOP
)
1403 sk_busy_loop(sock
->sk
, 1);
1405 /* if this socket can poll_ll, tell the system call */
1406 flag
= POLL_BUSY_LOOP
;
1409 return ops
->poll(file
, sock
, wait
) | flag
;
1412 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1414 struct socket
*sock
= file
->private_data
;
1416 return READ_ONCE(sock
->ops
)->mmap(file
, sock
, vma
);
1419 static int sock_close(struct inode
*inode
, struct file
*filp
)
1421 __sock_release(SOCKET_I(inode
), inode
);
1426 * Update the socket async list
1428 * Fasync_list locking strategy.
1430 * 1. fasync_list is modified only under process context socket lock
1431 * i.e. under semaphore.
1432 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1433 * or under socket lock
1436 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1438 struct socket
*sock
= filp
->private_data
;
1439 struct sock
*sk
= sock
->sk
;
1440 struct socket_wq
*wq
= &sock
->wq
;
1446 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1448 if (!wq
->fasync_list
)
1449 sock_reset_flag(sk
, SOCK_FASYNC
);
1451 sock_set_flag(sk
, SOCK_FASYNC
);
1457 /* This function may be called only under rcu_lock */
1459 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1461 if (!wq
|| !wq
->fasync_list
)
1465 case SOCK_WAKE_WAITD
:
1466 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1469 case SOCK_WAKE_SPACE
:
1470 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1475 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1478 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1483 EXPORT_SYMBOL(sock_wake_async
);
1486 * __sock_create - creates a socket
1487 * @net: net namespace
1488 * @family: protocol family (AF_INET, ...)
1489 * @type: communication type (SOCK_STREAM, ...)
1490 * @protocol: protocol (0, ...)
1492 * @kern: boolean for kernel space sockets
1494 * Creates a new socket and assigns it to @res, passing through LSM.
1495 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1496 * be set to true if the socket resides in kernel space.
1497 * This function internally uses GFP_KERNEL.
1500 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1501 struct socket
**res
, int kern
)
1504 struct socket
*sock
;
1505 const struct net_proto_family
*pf
;
1508 * Check protocol is in range
1510 if (family
< 0 || family
>= NPROTO
)
1511 return -EAFNOSUPPORT
;
1512 if (type
< 0 || type
>= SOCK_MAX
)
1517 This uglymoron is moved from INET layer to here to avoid
1518 deadlock in module load.
1520 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1521 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1526 err
= security_socket_create(family
, type
, protocol
, kern
);
1531 * Allocate the socket and allow the family to set things up. if
1532 * the protocol is 0, the family is instructed to select an appropriate
1535 sock
= sock_alloc();
1537 net_warn_ratelimited("socket: no more sockets\n");
1538 return -ENFILE
; /* Not exactly a match, but its the
1539 closest posix thing */
1544 #ifdef CONFIG_MODULES
1545 /* Attempt to load a protocol module if the find failed.
1547 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1548 * requested real, full-featured networking support upon configuration.
1549 * Otherwise module support will break!
1551 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1552 request_module("net-pf-%d", family
);
1556 pf
= rcu_dereference(net_families
[family
]);
1557 err
= -EAFNOSUPPORT
;
1562 * We will call the ->create function, that possibly is in a loadable
1563 * module, so we have to bump that loadable module refcnt first.
1565 if (!try_module_get(pf
->owner
))
1568 /* Now protected by module ref count */
1571 err
= pf
->create(net
, sock
, protocol
, kern
);
1573 goto out_module_put
;
1576 * Now to bump the refcnt of the [loadable] module that owns this
1577 * socket at sock_release time we decrement its refcnt.
1579 if (!try_module_get(sock
->ops
->owner
))
1580 goto out_module_busy
;
1583 * Now that we're done with the ->create function, the [loadable]
1584 * module can have its refcnt decremented
1586 module_put(pf
->owner
);
1587 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1589 goto out_sock_release
;
1595 err
= -EAFNOSUPPORT
;
1598 module_put(pf
->owner
);
1605 goto out_sock_release
;
1607 EXPORT_SYMBOL(__sock_create
);
1610 * sock_create - creates a socket
1611 * @family: protocol family (AF_INET, ...)
1612 * @type: communication type (SOCK_STREAM, ...)
1613 * @protocol: protocol (0, ...)
1616 * A wrapper around __sock_create().
1617 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1620 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1622 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1624 EXPORT_SYMBOL(sock_create
);
1627 * sock_create_kern - creates a socket (kernel space)
1628 * @net: net namespace
1629 * @family: protocol family (AF_INET, ...)
1630 * @type: communication type (SOCK_STREAM, ...)
1631 * @protocol: protocol (0, ...)
1634 * A wrapper around __sock_create().
1635 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1638 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1640 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1642 EXPORT_SYMBOL(sock_create_kern
);
1644 static struct socket
*__sys_socket_create(int family
, int type
, int protocol
)
1646 struct socket
*sock
;
1649 /* Check the SOCK_* constants for consistency. */
1650 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1651 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1652 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1653 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1655 if ((type
& ~SOCK_TYPE_MASK
) & ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1656 return ERR_PTR(-EINVAL
);
1657 type
&= SOCK_TYPE_MASK
;
1659 retval
= sock_create(family
, type
, protocol
, &sock
);
1661 return ERR_PTR(retval
);
1666 struct file
*__sys_socket_file(int family
, int type
, int protocol
)
1668 struct socket
*sock
;
1671 sock
= __sys_socket_create(family
, type
, protocol
);
1673 return ERR_CAST(sock
);
1675 flags
= type
& ~SOCK_TYPE_MASK
;
1676 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1677 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1679 return sock_alloc_file(sock
, flags
, NULL
);
1682 /* A hook for bpf progs to attach to and update socket protocol.
1684 * A static noinline declaration here could cause the compiler to
1685 * optimize away the function. A global noinline declaration will
1686 * keep the definition, but may optimize away the callsite.
1687 * Therefore, __weak is needed to ensure that the call is still
1688 * emitted, by telling the compiler that we don't know what the
1689 * function might eventually be.
1694 __weak noinline
int update_socket_protocol(int family
, int type
, int protocol
)
1701 int __sys_socket(int family
, int type
, int protocol
)
1703 struct socket
*sock
;
1706 sock
= __sys_socket_create(family
, type
,
1707 update_socket_protocol(family
, type
, protocol
));
1709 return PTR_ERR(sock
);
1711 flags
= type
& ~SOCK_TYPE_MASK
;
1712 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1713 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1715 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1718 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1720 return __sys_socket(family
, type
, protocol
);
1724 * Create a pair of connected sockets.
1727 int __sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1729 struct socket
*sock1
, *sock2
;
1731 struct file
*newfile1
, *newfile2
;
1734 flags
= type
& ~SOCK_TYPE_MASK
;
1735 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1737 type
&= SOCK_TYPE_MASK
;
1739 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1740 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1743 * reserve descriptors and make sure we won't fail
1744 * to return them to userland.
1746 fd1
= get_unused_fd_flags(flags
);
1747 if (unlikely(fd1
< 0))
1750 fd2
= get_unused_fd_flags(flags
);
1751 if (unlikely(fd2
< 0)) {
1756 err
= put_user(fd1
, &usockvec
[0]);
1760 err
= put_user(fd2
, &usockvec
[1]);
1765 * Obtain the first socket and check if the underlying protocol
1766 * supports the socketpair call.
1769 err
= sock_create(family
, type
, protocol
, &sock1
);
1770 if (unlikely(err
< 0))
1773 err
= sock_create(family
, type
, protocol
, &sock2
);
1774 if (unlikely(err
< 0)) {
1775 sock_release(sock1
);
1779 err
= security_socket_socketpair(sock1
, sock2
);
1780 if (unlikely(err
)) {
1781 sock_release(sock2
);
1782 sock_release(sock1
);
1786 err
= READ_ONCE(sock1
->ops
)->socketpair(sock1
, sock2
);
1787 if (unlikely(err
< 0)) {
1788 sock_release(sock2
);
1789 sock_release(sock1
);
1793 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1794 if (IS_ERR(newfile1
)) {
1795 err
= PTR_ERR(newfile1
);
1796 sock_release(sock2
);
1800 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1801 if (IS_ERR(newfile2
)) {
1802 err
= PTR_ERR(newfile2
);
1807 audit_fd_pair(fd1
, fd2
);
1809 fd_install(fd1
, newfile1
);
1810 fd_install(fd2
, newfile2
);
1819 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1820 int __user
*, usockvec
)
1822 return __sys_socketpair(family
, type
, protocol
, usockvec
);
1825 int __sys_bind_socket(struct socket
*sock
, struct sockaddr_storage
*address
,
1830 err
= security_socket_bind(sock
, (struct sockaddr
*)address
,
1833 err
= READ_ONCE(sock
->ops
)->bind(sock
,
1834 (struct sockaddr
*)address
,
1840 * Bind a name to a socket. Nothing much to do here since it's
1841 * the protocol's responsibility to handle the local address.
1843 * We move the socket address to kernel space before we call
1844 * the protocol layer (having also checked the address is ok).
1847 int __sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1849 struct socket
*sock
;
1850 struct sockaddr_storage address
;
1851 int err
, fput_needed
;
1853 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1855 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1857 err
= __sys_bind_socket(sock
, &address
, addrlen
);
1858 fput_light(sock
->file
, fput_needed
);
1863 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1865 return __sys_bind(fd
, umyaddr
, addrlen
);
1869 * Perform a listen. Basically, we allow the protocol to do anything
1870 * necessary for a listen, and if that works, we mark the socket as
1871 * ready for listening.
1873 int __sys_listen_socket(struct socket
*sock
, int backlog
)
1877 somaxconn
= READ_ONCE(sock_net(sock
->sk
)->core
.sysctl_somaxconn
);
1878 if ((unsigned int)backlog
> somaxconn
)
1879 backlog
= somaxconn
;
1881 err
= security_socket_listen(sock
, backlog
);
1883 err
= READ_ONCE(sock
->ops
)->listen(sock
, backlog
);
1887 int __sys_listen(int fd
, int backlog
)
1889 struct socket
*sock
;
1890 int err
, fput_needed
;
1892 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1894 err
= __sys_listen_socket(sock
, backlog
);
1895 fput_light(sock
->file
, fput_needed
);
1900 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1902 return __sys_listen(fd
, backlog
);
1905 struct file
*do_accept(struct file
*file
, struct proto_accept_arg
*arg
,
1906 struct sockaddr __user
*upeer_sockaddr
,
1907 int __user
*upeer_addrlen
, int flags
)
1909 struct socket
*sock
, *newsock
;
1910 struct file
*newfile
;
1912 struct sockaddr_storage address
;
1913 const struct proto_ops
*ops
;
1915 sock
= sock_from_file(file
);
1917 return ERR_PTR(-ENOTSOCK
);
1919 newsock
= sock_alloc();
1921 return ERR_PTR(-ENFILE
);
1922 ops
= READ_ONCE(sock
->ops
);
1924 newsock
->type
= sock
->type
;
1928 * We don't need try_module_get here, as the listening socket (sock)
1929 * has the protocol module (sock->ops->owner) held.
1931 __module_get(ops
->owner
);
1933 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1934 if (IS_ERR(newfile
))
1937 err
= security_socket_accept(sock
, newsock
);
1941 arg
->flags
|= sock
->file
->f_flags
;
1942 err
= ops
->accept(sock
, newsock
, arg
);
1946 if (upeer_sockaddr
) {
1947 len
= ops
->getname(newsock
, (struct sockaddr
*)&address
, 2);
1949 err
= -ECONNABORTED
;
1952 err
= move_addr_to_user(&address
,
1953 len
, upeer_sockaddr
, upeer_addrlen
);
1958 /* File flags are not inherited via accept() unlike another OSes. */
1962 return ERR_PTR(err
);
1965 static int __sys_accept4_file(struct file
*file
, struct sockaddr __user
*upeer_sockaddr
,
1966 int __user
*upeer_addrlen
, int flags
)
1968 struct proto_accept_arg arg
= { };
1969 struct file
*newfile
;
1972 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1975 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1976 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1978 newfd
= get_unused_fd_flags(flags
);
1979 if (unlikely(newfd
< 0))
1982 newfile
= do_accept(file
, &arg
, upeer_sockaddr
, upeer_addrlen
,
1984 if (IS_ERR(newfile
)) {
1985 put_unused_fd(newfd
);
1986 return PTR_ERR(newfile
);
1988 fd_install(newfd
, newfile
);
1993 * For accept, we attempt to create a new socket, set up the link
1994 * with the client, wake up the client, then return the new
1995 * connected fd. We collect the address of the connector in kernel
1996 * space and move it to user at the very end. This is unclean because
1997 * we open the socket then return an error.
1999 * 1003.1g adds the ability to recvmsg() to query connection pending
2000 * status to recvmsg. We need to add that support in a way thats
2001 * clean when we restructure accept also.
2004 int __sys_accept4(int fd
, struct sockaddr __user
*upeer_sockaddr
,
2005 int __user
*upeer_addrlen
, int flags
)
2012 ret
= __sys_accept4_file(f
.file
, upeer_sockaddr
,
2013 upeer_addrlen
, flags
);
2020 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
2021 int __user
*, upeer_addrlen
, int, flags
)
2023 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, flags
);
2026 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
2027 int __user
*, upeer_addrlen
)
2029 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
2033 * Attempt to connect to a socket with the server address. The address
2034 * is in user space so we verify it is OK and move it to kernel space.
2036 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2039 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2040 * other SEQPACKET protocols that take time to connect() as it doesn't
2041 * include the -EINPROGRESS status for such sockets.
2044 int __sys_connect_file(struct file
*file
, struct sockaddr_storage
*address
,
2045 int addrlen
, int file_flags
)
2047 struct socket
*sock
;
2050 sock
= sock_from_file(file
);
2057 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
2061 err
= READ_ONCE(sock
->ops
)->connect(sock
, (struct sockaddr
*)address
,
2062 addrlen
, sock
->file
->f_flags
| file_flags
);
2067 int __sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
2074 struct sockaddr_storage address
;
2076 ret
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
2078 ret
= __sys_connect_file(f
.file
, &address
, addrlen
, 0);
2085 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
2088 return __sys_connect(fd
, uservaddr
, addrlen
);
2092 * Get the local address ('name') of a socket object. Move the obtained
2093 * name to user space.
2096 int __sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
2097 int __user
*usockaddr_len
)
2099 struct socket
*sock
;
2100 struct sockaddr_storage address
;
2101 int err
, fput_needed
;
2103 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2107 err
= security_socket_getsockname(sock
);
2111 err
= READ_ONCE(sock
->ops
)->getname(sock
, (struct sockaddr
*)&address
, 0);
2114 /* "err" is actually length in this case */
2115 err
= move_addr_to_user(&address
, err
, usockaddr
, usockaddr_len
);
2118 fput_light(sock
->file
, fput_needed
);
2123 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
2124 int __user
*, usockaddr_len
)
2126 return __sys_getsockname(fd
, usockaddr
, usockaddr_len
);
2130 * Get the remote address ('name') of a socket object. Move the obtained
2131 * name to user space.
2134 int __sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
2135 int __user
*usockaddr_len
)
2137 struct socket
*sock
;
2138 struct sockaddr_storage address
;
2139 int err
, fput_needed
;
2141 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2143 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
2145 err
= security_socket_getpeername(sock
);
2147 fput_light(sock
->file
, fput_needed
);
2151 err
= ops
->getname(sock
, (struct sockaddr
*)&address
, 1);
2153 /* "err" is actually length in this case */
2154 err
= move_addr_to_user(&address
, err
, usockaddr
,
2156 fput_light(sock
->file
, fput_needed
);
2161 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
2162 int __user
*, usockaddr_len
)
2164 return __sys_getpeername(fd
, usockaddr
, usockaddr_len
);
2168 * Send a datagram to a given address. We move the address into kernel
2169 * space and check the user space data area is readable before invoking
2172 int __sys_sendto(int fd
, void __user
*buff
, size_t len
, unsigned int flags
,
2173 struct sockaddr __user
*addr
, int addr_len
)
2175 struct socket
*sock
;
2176 struct sockaddr_storage address
;
2181 err
= import_ubuf(ITER_SOURCE
, buff
, len
, &msg
.msg_iter
);
2184 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2188 msg
.msg_name
= NULL
;
2189 msg
.msg_control
= NULL
;
2190 msg
.msg_controllen
= 0;
2191 msg
.msg_namelen
= 0;
2192 msg
.msg_ubuf
= NULL
;
2194 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
2197 msg
.msg_name
= (struct sockaddr
*)&address
;
2198 msg
.msg_namelen
= addr_len
;
2200 flags
&= ~MSG_INTERNAL_SENDMSG_FLAGS
;
2201 if (sock
->file
->f_flags
& O_NONBLOCK
)
2202 flags
|= MSG_DONTWAIT
;
2203 msg
.msg_flags
= flags
;
2204 err
= __sock_sendmsg(sock
, &msg
);
2207 fput_light(sock
->file
, fput_needed
);
2212 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
2213 unsigned int, flags
, struct sockaddr __user
*, addr
,
2216 return __sys_sendto(fd
, buff
, len
, flags
, addr
, addr_len
);
2220 * Send a datagram down a socket.
2223 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
2224 unsigned int, flags
)
2226 return __sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
2230 * Receive a frame from the socket and optionally record the address of the
2231 * sender. We verify the buffers are writable and if needed move the
2232 * sender address from kernel to user space.
2234 int __sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
, unsigned int flags
,
2235 struct sockaddr __user
*addr
, int __user
*addr_len
)
2237 struct sockaddr_storage address
;
2238 struct msghdr msg
= {
2239 /* Save some cycles and don't copy the address if not needed */
2240 .msg_name
= addr
? (struct sockaddr
*)&address
: NULL
,
2242 struct socket
*sock
;
2246 err
= import_ubuf(ITER_DEST
, ubuf
, size
, &msg
.msg_iter
);
2249 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2253 if (sock
->file
->f_flags
& O_NONBLOCK
)
2254 flags
|= MSG_DONTWAIT
;
2255 err
= sock_recvmsg(sock
, &msg
, flags
);
2257 if (err
>= 0 && addr
!= NULL
) {
2258 err2
= move_addr_to_user(&address
,
2259 msg
.msg_namelen
, addr
, addr_len
);
2264 fput_light(sock
->file
, fput_needed
);
2269 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
2270 unsigned int, flags
, struct sockaddr __user
*, addr
,
2271 int __user
*, addr_len
)
2273 return __sys_recvfrom(fd
, ubuf
, size
, flags
, addr
, addr_len
);
2277 * Receive a datagram from a socket.
2280 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
2281 unsigned int, flags
)
2283 return __sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
2286 static bool sock_use_custom_sol_socket(const struct socket
*sock
)
2288 return test_bit(SOCK_CUSTOM_SOCKOPT
, &sock
->flags
);
2291 int do_sock_setsockopt(struct socket
*sock
, bool compat
, int level
,
2292 int optname
, sockptr_t optval
, int optlen
)
2294 const struct proto_ops
*ops
;
2295 char *kernel_optval
= NULL
;
2301 err
= security_socket_setsockopt(sock
, level
, optname
);
2306 err
= BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock
->sk
, &level
, &optname
,
2317 optval
= KERNEL_SOCKPTR(kernel_optval
);
2318 ops
= READ_ONCE(sock
->ops
);
2319 if (level
== SOL_SOCKET
&& !sock_use_custom_sol_socket(sock
))
2320 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2321 else if (unlikely(!ops
->setsockopt
))
2324 err
= ops
->setsockopt(sock
, level
, optname
, optval
,
2326 kfree(kernel_optval
);
2330 EXPORT_SYMBOL(do_sock_setsockopt
);
2332 /* Set a socket option. Because we don't know the option lengths we have
2333 * to pass the user mode parameter for the protocols to sort out.
2335 int __sys_setsockopt(int fd
, int level
, int optname
, char __user
*user_optval
,
2338 sockptr_t optval
= USER_SOCKPTR(user_optval
);
2339 bool compat
= in_compat_syscall();
2340 int err
, fput_needed
;
2341 struct socket
*sock
;
2343 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2347 err
= do_sock_setsockopt(sock
, compat
, level
, optname
, optval
, optlen
);
2349 fput_light(sock
->file
, fput_needed
);
2353 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
2354 char __user
*, optval
, int, optlen
)
2356 return __sys_setsockopt(fd
, level
, optname
, optval
, optlen
);
2359 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level
,
2362 int do_sock_getsockopt(struct socket
*sock
, bool compat
, int level
,
2363 int optname
, sockptr_t optval
, sockptr_t optlen
)
2365 int max_optlen __maybe_unused
;
2366 const struct proto_ops
*ops
;
2369 err
= security_socket_getsockopt(sock
, level
, optname
);
2374 max_optlen
= BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen
);
2376 ops
= READ_ONCE(sock
->ops
);
2377 if (level
== SOL_SOCKET
) {
2378 err
= sk_getsockopt(sock
->sk
, level
, optname
, optval
, optlen
);
2379 } else if (unlikely(!ops
->getsockopt
)) {
2382 if (WARN_ONCE(optval
.is_kernel
|| optlen
.is_kernel
,
2383 "Invalid argument type"))
2386 err
= ops
->getsockopt(sock
, level
, optname
, optval
.user
,
2391 err
= BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock
->sk
, level
, optname
,
2392 optval
, optlen
, max_optlen
,
2397 EXPORT_SYMBOL(do_sock_getsockopt
);
2400 * Get a socket option. Because we don't know the option lengths we have
2401 * to pass a user mode parameter for the protocols to sort out.
2403 int __sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
,
2406 int err
, fput_needed
;
2407 struct socket
*sock
;
2410 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2414 compat
= in_compat_syscall();
2415 err
= do_sock_getsockopt(sock
, compat
, level
, optname
,
2416 USER_SOCKPTR(optval
), USER_SOCKPTR(optlen
));
2418 fput_light(sock
->file
, fput_needed
);
2422 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
2423 char __user
*, optval
, int __user
*, optlen
)
2425 return __sys_getsockopt(fd
, level
, optname
, optval
, optlen
);
2429 * Shutdown a socket.
2432 int __sys_shutdown_sock(struct socket
*sock
, int how
)
2436 err
= security_socket_shutdown(sock
, how
);
2438 err
= READ_ONCE(sock
->ops
)->shutdown(sock
, how
);
2443 int __sys_shutdown(int fd
, int how
)
2445 int err
, fput_needed
;
2446 struct socket
*sock
;
2448 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2450 err
= __sys_shutdown_sock(sock
, how
);
2451 fput_light(sock
->file
, fput_needed
);
2456 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
2458 return __sys_shutdown(fd
, how
);
2461 /* A couple of helpful macros for getting the address of the 32/64 bit
2462 * fields which are the same type (int / unsigned) on our platforms.
2464 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2465 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2466 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2468 struct used_address
{
2469 struct sockaddr_storage name
;
2470 unsigned int name_len
;
2473 int __copy_msghdr(struct msghdr
*kmsg
,
2474 struct user_msghdr
*msg
,
2475 struct sockaddr __user
**save_addr
)
2479 kmsg
->msg_control_is_user
= true;
2480 kmsg
->msg_get_inq
= 0;
2481 kmsg
->msg_control_user
= msg
->msg_control
;
2482 kmsg
->msg_controllen
= msg
->msg_controllen
;
2483 kmsg
->msg_flags
= msg
->msg_flags
;
2485 kmsg
->msg_namelen
= msg
->msg_namelen
;
2487 kmsg
->msg_namelen
= 0;
2489 if (kmsg
->msg_namelen
< 0)
2492 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2493 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2496 *save_addr
= msg
->msg_name
;
2498 if (msg
->msg_name
&& kmsg
->msg_namelen
) {
2500 err
= move_addr_to_kernel(msg
->msg_name
,
2507 kmsg
->msg_name
= NULL
;
2508 kmsg
->msg_namelen
= 0;
2511 if (msg
->msg_iovlen
> UIO_MAXIOV
)
2514 kmsg
->msg_iocb
= NULL
;
2515 kmsg
->msg_ubuf
= NULL
;
2519 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
2520 struct user_msghdr __user
*umsg
,
2521 struct sockaddr __user
**save_addr
,
2524 struct user_msghdr msg
;
2527 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
2530 err
= __copy_msghdr(kmsg
, &msg
, save_addr
);
2534 err
= import_iovec(save_addr
? ITER_DEST
: ITER_SOURCE
,
2535 msg
.msg_iov
, msg
.msg_iovlen
,
2536 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
2537 return err
< 0 ? err
: 0;
2540 static int ____sys_sendmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2541 unsigned int flags
, struct used_address
*used_address
,
2542 unsigned int allowed_msghdr_flags
)
2544 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2545 __aligned(sizeof(__kernel_size_t
));
2546 /* 20 is size of ipv6_pktinfo */
2547 unsigned char *ctl_buf
= ctl
;
2553 if (msg_sys
->msg_controllen
> INT_MAX
)
2555 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
2556 ctl_len
= msg_sys
->msg_controllen
;
2557 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2559 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2563 ctl_buf
= msg_sys
->msg_control
;
2564 ctl_len
= msg_sys
->msg_controllen
;
2565 } else if (ctl_len
) {
2566 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2567 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2568 if (ctl_len
> sizeof(ctl
)) {
2569 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2570 if (ctl_buf
== NULL
)
2574 if (copy_from_user(ctl_buf
, msg_sys
->msg_control_user
, ctl_len
))
2576 msg_sys
->msg_control
= ctl_buf
;
2577 msg_sys
->msg_control_is_user
= false;
2579 flags
&= ~MSG_INTERNAL_SENDMSG_FLAGS
;
2580 msg_sys
->msg_flags
= flags
;
2582 if (sock
->file
->f_flags
& O_NONBLOCK
)
2583 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2585 * If this is sendmmsg() and current destination address is same as
2586 * previously succeeded address, omit asking LSM's decision.
2587 * used_address->name_len is initialized to UINT_MAX so that the first
2588 * destination address never matches.
2590 if (used_address
&& msg_sys
->msg_name
&&
2591 used_address
->name_len
== msg_sys
->msg_namelen
&&
2592 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2593 used_address
->name_len
)) {
2594 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2597 err
= __sock_sendmsg(sock
, msg_sys
);
2599 * If this is sendmmsg() and sending to current destination address was
2600 * successful, remember it.
2602 if (used_address
&& err
>= 0) {
2603 used_address
->name_len
= msg_sys
->msg_namelen
;
2604 if (msg_sys
->msg_name
)
2605 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2606 used_address
->name_len
);
2611 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2616 static int sendmsg_copy_msghdr(struct msghdr
*msg
,
2617 struct user_msghdr __user
*umsg
, unsigned flags
,
2622 if (flags
& MSG_CMSG_COMPAT
) {
2623 struct compat_msghdr __user
*msg_compat
;
2625 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2626 err
= get_compat_msghdr(msg
, msg_compat
, NULL
, iov
);
2628 err
= copy_msghdr_from_user(msg
, umsg
, NULL
, iov
);
2636 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2637 struct msghdr
*msg_sys
, unsigned int flags
,
2638 struct used_address
*used_address
,
2639 unsigned int allowed_msghdr_flags
)
2641 struct sockaddr_storage address
;
2642 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2645 msg_sys
->msg_name
= &address
;
2647 err
= sendmsg_copy_msghdr(msg_sys
, msg
, flags
, &iov
);
2651 err
= ____sys_sendmsg(sock
, msg_sys
, flags
, used_address
,
2652 allowed_msghdr_flags
);
2658 * BSD sendmsg interface
2660 long __sys_sendmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2663 return ____sys_sendmsg(sock
, msg
, flags
, NULL
, 0);
2666 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2667 bool forbid_cmsg_compat
)
2669 int fput_needed
, err
;
2670 struct msghdr msg_sys
;
2671 struct socket
*sock
;
2673 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2676 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2680 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2682 fput_light(sock
->file
, fput_needed
);
2687 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2689 return __sys_sendmsg(fd
, msg
, flags
, true);
2693 * Linux sendmmsg interface
2696 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2697 unsigned int flags
, bool forbid_cmsg_compat
)
2699 int fput_needed
, err
, datagrams
;
2700 struct socket
*sock
;
2701 struct mmsghdr __user
*entry
;
2702 struct compat_mmsghdr __user
*compat_entry
;
2703 struct msghdr msg_sys
;
2704 struct used_address used_address
;
2705 unsigned int oflags
= flags
;
2707 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2710 if (vlen
> UIO_MAXIOV
)
2715 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2719 used_address
.name_len
= UINT_MAX
;
2721 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2725 while (datagrams
< vlen
) {
2726 if (datagrams
== vlen
- 1)
2729 if (MSG_CMSG_COMPAT
& flags
) {
2730 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2731 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2734 err
= __put_user(err
, &compat_entry
->msg_len
);
2737 err
= ___sys_sendmsg(sock
,
2738 (struct user_msghdr __user
*)entry
,
2739 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2742 err
= put_user(err
, &entry
->msg_len
);
2749 if (msg_data_left(&msg_sys
))
2754 fput_light(sock
->file
, fput_needed
);
2756 /* We only return an error if no datagrams were able to be sent */
2763 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2764 unsigned int, vlen
, unsigned int, flags
)
2766 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
, true);
2769 static int recvmsg_copy_msghdr(struct msghdr
*msg
,
2770 struct user_msghdr __user
*umsg
, unsigned flags
,
2771 struct sockaddr __user
**uaddr
,
2776 if (MSG_CMSG_COMPAT
& flags
) {
2777 struct compat_msghdr __user
*msg_compat
;
2779 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2780 err
= get_compat_msghdr(msg
, msg_compat
, uaddr
, iov
);
2782 err
= copy_msghdr_from_user(msg
, umsg
, uaddr
, iov
);
2790 static int ____sys_recvmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2791 struct user_msghdr __user
*msg
,
2792 struct sockaddr __user
*uaddr
,
2793 unsigned int flags
, int nosec
)
2795 struct compat_msghdr __user
*msg_compat
=
2796 (struct compat_msghdr __user
*) msg
;
2797 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2798 struct sockaddr_storage addr
;
2799 unsigned long cmsg_ptr
;
2803 msg_sys
->msg_name
= &addr
;
2804 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2805 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2807 /* We assume all kernel code knows the size of sockaddr_storage */
2808 msg_sys
->msg_namelen
= 0;
2810 if (sock
->file
->f_flags
& O_NONBLOCK
)
2811 flags
|= MSG_DONTWAIT
;
2813 if (unlikely(nosec
))
2814 err
= sock_recvmsg_nosec(sock
, msg_sys
, flags
);
2816 err
= sock_recvmsg(sock
, msg_sys
, flags
);
2822 if (uaddr
!= NULL
) {
2823 err
= move_addr_to_user(&addr
,
2824 msg_sys
->msg_namelen
, uaddr
,
2829 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2833 if (MSG_CMSG_COMPAT
& flags
)
2834 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2835 &msg_compat
->msg_controllen
);
2837 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2838 &msg
->msg_controllen
);
2846 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2847 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2849 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2850 /* user mode address pointers */
2851 struct sockaddr __user
*uaddr
;
2854 err
= recvmsg_copy_msghdr(msg_sys
, msg
, flags
, &uaddr
, &iov
);
2858 err
= ____sys_recvmsg(sock
, msg_sys
, msg
, uaddr
, flags
, nosec
);
2864 * BSD recvmsg interface
2867 long __sys_recvmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2868 struct user_msghdr __user
*umsg
,
2869 struct sockaddr __user
*uaddr
, unsigned int flags
)
2871 return ____sys_recvmsg(sock
, msg
, umsg
, uaddr
, flags
, 0);
2874 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2875 bool forbid_cmsg_compat
)
2877 int fput_needed
, err
;
2878 struct msghdr msg_sys
;
2879 struct socket
*sock
;
2881 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2884 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2888 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2890 fput_light(sock
->file
, fput_needed
);
2895 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2896 unsigned int, flags
)
2898 return __sys_recvmsg(fd
, msg
, flags
, true);
2902 * Linux recvmmsg interface
2905 static int do_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2906 unsigned int vlen
, unsigned int flags
,
2907 struct timespec64
*timeout
)
2909 int fput_needed
, err
, datagrams
;
2910 struct socket
*sock
;
2911 struct mmsghdr __user
*entry
;
2912 struct compat_mmsghdr __user
*compat_entry
;
2913 struct msghdr msg_sys
;
2914 struct timespec64 end_time
;
2915 struct timespec64 timeout64
;
2918 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2924 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2928 if (likely(!(flags
& MSG_ERRQUEUE
))) {
2929 err
= sock_error(sock
->sk
);
2937 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2939 while (datagrams
< vlen
) {
2941 * No need to ask LSM for more than the first datagram.
2943 if (MSG_CMSG_COMPAT
& flags
) {
2944 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2945 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2949 err
= __put_user(err
, &compat_entry
->msg_len
);
2952 err
= ___sys_recvmsg(sock
,
2953 (struct user_msghdr __user
*)entry
,
2954 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2958 err
= put_user(err
, &entry
->msg_len
);
2966 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2967 if (flags
& MSG_WAITFORONE
)
2968 flags
|= MSG_DONTWAIT
;
2971 ktime_get_ts64(&timeout64
);
2972 *timeout
= timespec64_sub(end_time
, timeout64
);
2973 if (timeout
->tv_sec
< 0) {
2974 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2978 /* Timeout, return less than vlen datagrams */
2979 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2983 /* Out of band data, return right away */
2984 if (msg_sys
.msg_flags
& MSG_OOB
)
2992 if (datagrams
== 0) {
2998 * We may return less entries than requested (vlen) if the
2999 * sock is non block and there aren't enough datagrams...
3001 if (err
!= -EAGAIN
) {
3003 * ... or if recvmsg returns an error after we
3004 * received some datagrams, where we record the
3005 * error to return on the next call or if the
3006 * app asks about it using getsockopt(SO_ERROR).
3008 WRITE_ONCE(sock
->sk
->sk_err
, -err
);
3011 fput_light(sock
->file
, fput_needed
);
3016 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
3017 unsigned int vlen
, unsigned int flags
,
3018 struct __kernel_timespec __user
*timeout
,
3019 struct old_timespec32 __user
*timeout32
)
3022 struct timespec64 timeout_sys
;
3024 if (timeout
&& get_timespec64(&timeout_sys
, timeout
))
3027 if (timeout32
&& get_old_timespec32(&timeout_sys
, timeout32
))
3030 if (!timeout
&& !timeout32
)
3031 return do_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
3033 datagrams
= do_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
3038 if (timeout
&& put_timespec64(&timeout_sys
, timeout
))
3039 datagrams
= -EFAULT
;
3041 if (timeout32
&& put_old_timespec32(&timeout_sys
, timeout32
))
3042 datagrams
= -EFAULT
;
3047 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
3048 unsigned int, vlen
, unsigned int, flags
,
3049 struct __kernel_timespec __user
*, timeout
)
3051 if (flags
& MSG_CMSG_COMPAT
)
3054 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, timeout
, NULL
);
3057 #ifdef CONFIG_COMPAT_32BIT_TIME
3058 SYSCALL_DEFINE5(recvmmsg_time32
, int, fd
, struct mmsghdr __user
*, mmsg
,
3059 unsigned int, vlen
, unsigned int, flags
,
3060 struct old_timespec32 __user
*, timeout
)
3062 if (flags
& MSG_CMSG_COMPAT
)
3065 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
, timeout
);
3069 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3070 /* Argument list sizes for sys_socketcall */
3071 #define AL(x) ((x) * sizeof(unsigned long))
3072 static const unsigned char nargs
[21] = {
3073 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3074 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3075 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3082 * System call vectors.
3084 * Argument checking cleaned up. Saved 20% in size.
3085 * This function doesn't need to set the kernel lock because
3086 * it is set by the callees.
3089 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
3091 unsigned long a
[AUDITSC_ARGS
];
3092 unsigned long a0
, a1
;
3096 if (call
< 1 || call
> SYS_SENDMMSG
)
3098 call
= array_index_nospec(call
, SYS_SENDMMSG
+ 1);
3101 if (len
> sizeof(a
))
3104 /* copy_from_user should be SMP safe. */
3105 if (copy_from_user(a
, args
, len
))
3108 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
3117 err
= __sys_socket(a0
, a1
, a
[2]);
3120 err
= __sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3123 err
= __sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3126 err
= __sys_listen(a0
, a1
);
3129 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3130 (int __user
*)a
[2], 0);
3132 case SYS_GETSOCKNAME
:
3134 __sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
3135 (int __user
*)a
[2]);
3137 case SYS_GETPEERNAME
:
3139 __sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
3140 (int __user
*)a
[2]);
3142 case SYS_SOCKETPAIR
:
3143 err
= __sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
3146 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3150 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3151 (struct sockaddr __user
*)a
[4], a
[5]);
3154 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3158 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3159 (struct sockaddr __user
*)a
[4],
3160 (int __user
*)a
[5]);
3163 err
= __sys_shutdown(a0
, a1
);
3165 case SYS_SETSOCKOPT
:
3166 err
= __sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3169 case SYS_GETSOCKOPT
:
3171 __sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3172 (int __user
*)a
[4]);
3175 err
= __sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
,
3179 err
= __sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2],
3183 err
= __sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
,
3187 if (IS_ENABLED(CONFIG_64BIT
))
3188 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3190 (struct __kernel_timespec __user
*)a
[4],
3193 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3195 (struct old_timespec32 __user
*)a
[4]);
3198 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3199 (int __user
*)a
[2], a
[3]);
3208 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3211 * sock_register - add a socket protocol handler
3212 * @ops: description of protocol
3214 * This function is called by a protocol handler that wants to
3215 * advertise its address family, and have it linked into the
3216 * socket interface. The value ops->family corresponds to the
3217 * socket system call protocol family.
3219 int sock_register(const struct net_proto_family
*ops
)
3223 if (ops
->family
>= NPROTO
) {
3224 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
3228 spin_lock(&net_family_lock
);
3229 if (rcu_dereference_protected(net_families
[ops
->family
],
3230 lockdep_is_held(&net_family_lock
)))
3233 rcu_assign_pointer(net_families
[ops
->family
], ops
);
3236 spin_unlock(&net_family_lock
);
3238 pr_info("NET: Registered %s protocol family\n", pf_family_names
[ops
->family
]);
3241 EXPORT_SYMBOL(sock_register
);
3244 * sock_unregister - remove a protocol handler
3245 * @family: protocol family to remove
3247 * This function is called by a protocol handler that wants to
3248 * remove its address family, and have it unlinked from the
3249 * new socket creation.
3251 * If protocol handler is a module, then it can use module reference
3252 * counts to protect against new references. If protocol handler is not
3253 * a module then it needs to provide its own protection in
3254 * the ops->create routine.
3256 void sock_unregister(int family
)
3258 BUG_ON(family
< 0 || family
>= NPROTO
);
3260 spin_lock(&net_family_lock
);
3261 RCU_INIT_POINTER(net_families
[family
], NULL
);
3262 spin_unlock(&net_family_lock
);
3266 pr_info("NET: Unregistered %s protocol family\n", pf_family_names
[family
]);
3268 EXPORT_SYMBOL(sock_unregister
);
3270 bool sock_is_registered(int family
)
3272 return family
< NPROTO
&& rcu_access_pointer(net_families
[family
]);
3275 static int __init
sock_init(void)
3279 * Initialize the network sysctl infrastructure.
3281 err
= net_sysctl_init();
3286 * Initialize skbuff SLAB cache
3291 * Initialize the protocols module.
3296 err
= register_filesystem(&sock_fs_type
);
3299 sock_mnt
= kern_mount(&sock_fs_type
);
3300 if (IS_ERR(sock_mnt
)) {
3301 err
= PTR_ERR(sock_mnt
);
3305 /* The real protocol initialization is performed in later initcalls.
3308 #ifdef CONFIG_NETFILTER
3309 err
= netfilter_init();
3314 ptp_classifier_init();
3320 unregister_filesystem(&sock_fs_type
);
3324 core_initcall(sock_init
); /* early initcall */
3326 #ifdef CONFIG_PROC_FS
3327 void socket_seq_show(struct seq_file
*seq
)
3329 seq_printf(seq
, "sockets: used %d\n",
3330 sock_inuse_get(seq
->private));
3332 #endif /* CONFIG_PROC_FS */
3334 /* Handle the fact that while struct ifreq has the same *layout* on
3335 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3336 * which are handled elsewhere, it still has different *size* due to
3337 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3338 * resulting in struct ifreq being 32 and 40 bytes respectively).
3339 * As a result, if the struct happens to be at the end of a page and
3340 * the next page isn't readable/writable, we get a fault. To prevent
3341 * that, copy back and forth to the full size.
3343 int get_user_ifreq(struct ifreq
*ifr
, void __user
**ifrdata
, void __user
*arg
)
3345 if (in_compat_syscall()) {
3346 struct compat_ifreq
*ifr32
= (struct compat_ifreq
*)ifr
;
3348 memset(ifr
, 0, sizeof(*ifr
));
3349 if (copy_from_user(ifr32
, arg
, sizeof(*ifr32
)))
3353 *ifrdata
= compat_ptr(ifr32
->ifr_data
);
3358 if (copy_from_user(ifr
, arg
, sizeof(*ifr
)))
3362 *ifrdata
= ifr
->ifr_data
;
3366 EXPORT_SYMBOL(get_user_ifreq
);
3368 int put_user_ifreq(struct ifreq
*ifr
, void __user
*arg
)
3370 size_t size
= sizeof(*ifr
);
3372 if (in_compat_syscall())
3373 size
= sizeof(struct compat_ifreq
);
3375 if (copy_to_user(arg
, ifr
, size
))
3380 EXPORT_SYMBOL(put_user_ifreq
);
3382 #ifdef CONFIG_COMPAT
3383 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3385 compat_uptr_t uptr32
;
3390 if (get_user_ifreq(&ifr
, NULL
, uifr32
))
3393 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
3396 saved
= ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
;
3397 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= compat_ptr(uptr32
);
3399 err
= dev_ioctl(net
, SIOCWANDEV
, &ifr
, NULL
, NULL
);
3401 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= saved
;
3402 if (put_user_ifreq(&ifr
, uifr32
))
3408 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3409 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3410 struct compat_ifreq __user
*u_ifreq32
)
3415 if (!is_socket_ioctl_cmd(cmd
))
3417 if (get_user_ifreq(&ifreq
, &data
, u_ifreq32
))
3419 ifreq
.ifr_data
= data
;
3421 return dev_ioctl(net
, cmd
, &ifreq
, data
, NULL
);
3424 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3425 unsigned int cmd
, unsigned long arg
)
3427 void __user
*argp
= compat_ptr(arg
);
3428 struct sock
*sk
= sock
->sk
;
3429 struct net
*net
= sock_net(sk
);
3430 const struct proto_ops
*ops
;
3432 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3433 return sock_ioctl(file
, cmd
, (unsigned long)argp
);
3437 return compat_siocwandev(net
, argp
);
3438 case SIOCGSTAMP_OLD
:
3439 case SIOCGSTAMPNS_OLD
:
3440 ops
= READ_ONCE(sock
->ops
);
3441 if (!ops
->gettstamp
)
3442 return -ENOIOCTLCMD
;
3443 return ops
->gettstamp(sock
, argp
, cmd
== SIOCGSTAMP_OLD
,
3444 !COMPAT_USE_64BIT_TIME
);
3447 case SIOCBONDSLAVEINFOQUERY
:
3448 case SIOCBONDINFOQUERY
:
3451 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3462 case SIOCGSTAMP_NEW
:
3463 case SIOCGSTAMPNS_NEW
:
3467 return sock_ioctl(file
, cmd
, arg
);
3486 case SIOCSIFHWBROADCAST
:
3488 case SIOCGIFBRDADDR
:
3489 case SIOCSIFBRDADDR
:
3490 case SIOCGIFDSTADDR
:
3491 case SIOCSIFDSTADDR
:
3492 case SIOCGIFNETMASK
:
3493 case SIOCSIFNETMASK
:
3505 case SIOCBONDENSLAVE
:
3506 case SIOCBONDRELEASE
:
3507 case SIOCBONDSETHWADDR
:
3508 case SIOCBONDCHANGEACTIVE
:
3515 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3518 return -ENOIOCTLCMD
;
3521 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3524 struct socket
*sock
= file
->private_data
;
3525 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
3526 int ret
= -ENOIOCTLCMD
;
3533 if (ops
->compat_ioctl
)
3534 ret
= ops
->compat_ioctl(sock
, cmd
, arg
);
3536 if (ret
== -ENOIOCTLCMD
&&
3537 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3538 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3540 if (ret
== -ENOIOCTLCMD
)
3541 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3548 * kernel_bind - bind an address to a socket (kernel space)
3551 * @addrlen: length of address
3553 * Returns 0 or an error.
3556 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3558 struct sockaddr_storage address
;
3560 memcpy(&address
, addr
, addrlen
);
3562 return READ_ONCE(sock
->ops
)->bind(sock
, (struct sockaddr
*)&address
,
3565 EXPORT_SYMBOL(kernel_bind
);
3568 * kernel_listen - move socket to listening state (kernel space)
3570 * @backlog: pending connections queue size
3572 * Returns 0 or an error.
3575 int kernel_listen(struct socket
*sock
, int backlog
)
3577 return READ_ONCE(sock
->ops
)->listen(sock
, backlog
);
3579 EXPORT_SYMBOL(kernel_listen
);
3582 * kernel_accept - accept a connection (kernel space)
3583 * @sock: listening socket
3584 * @newsock: new connected socket
3587 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3588 * If it fails, @newsock is guaranteed to be %NULL.
3589 * Returns 0 or an error.
3592 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3594 struct sock
*sk
= sock
->sk
;
3595 const struct proto_ops
*ops
= READ_ONCE(sock
->ops
);
3596 struct proto_accept_arg arg
= {
3602 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3607 err
= ops
->accept(sock
, *newsock
, &arg
);
3609 sock_release(*newsock
);
3614 (*newsock
)->ops
= ops
;
3615 __module_get(ops
->owner
);
3620 EXPORT_SYMBOL(kernel_accept
);
3623 * kernel_connect - connect a socket (kernel space)
3626 * @addrlen: address length
3627 * @flags: flags (O_NONBLOCK, ...)
3629 * For datagram sockets, @addr is the address to which datagrams are sent
3630 * by default, and the only address from which datagrams are received.
3631 * For stream sockets, attempts to connect to @addr.
3632 * Returns 0 or an error code.
3635 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3638 struct sockaddr_storage address
;
3640 memcpy(&address
, addr
, addrlen
);
3642 return READ_ONCE(sock
->ops
)->connect(sock
, (struct sockaddr
*)&address
,
3645 EXPORT_SYMBOL(kernel_connect
);
3648 * kernel_getsockname - get the address which the socket is bound (kernel space)
3650 * @addr: address holder
3652 * Fills the @addr pointer with the address which the socket is bound.
3653 * Returns the length of the address in bytes or an error code.
3656 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
)
3658 return READ_ONCE(sock
->ops
)->getname(sock
, addr
, 0);
3660 EXPORT_SYMBOL(kernel_getsockname
);
3663 * kernel_getpeername - get the address which the socket is connected (kernel space)
3665 * @addr: address holder
3667 * Fills the @addr pointer with the address which the socket is connected.
3668 * Returns the length of the address in bytes or an error code.
3671 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
)
3673 return READ_ONCE(sock
->ops
)->getname(sock
, addr
, 1);
3675 EXPORT_SYMBOL(kernel_getpeername
);
3678 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3680 * @how: connection part
3682 * Returns 0 or an error.
3685 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3687 return READ_ONCE(sock
->ops
)->shutdown(sock
, how
);
3689 EXPORT_SYMBOL(kernel_sock_shutdown
);
3692 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3695 * This routine returns the IP overhead imposed by a socket i.e.
3696 * the length of the underlying IP header, depending on whether
3697 * this is an IPv4 or IPv6 socket and the length from IP options turned
3698 * on at the socket. Assumes that the caller has a lock on the socket.
3701 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3703 struct inet_sock
*inet
;
3704 struct ip_options_rcu
*opt
;
3706 #if IS_ENABLED(CONFIG_IPV6)
3707 struct ipv6_pinfo
*np
;
3708 struct ipv6_txoptions
*optv6
= NULL
;
3709 #endif /* IS_ENABLED(CONFIG_IPV6) */
3714 switch (sk
->sk_family
) {
3717 overhead
+= sizeof(struct iphdr
);
3718 opt
= rcu_dereference_protected(inet
->inet_opt
,
3719 sock_owned_by_user(sk
));
3721 overhead
+= opt
->opt
.optlen
;
3723 #if IS_ENABLED(CONFIG_IPV6)
3726 overhead
+= sizeof(struct ipv6hdr
);
3728 optv6
= rcu_dereference_protected(np
->opt
,
3729 sock_owned_by_user(sk
));
3731 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3733 #endif /* IS_ENABLED(CONFIG_IPV6) */
3734 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3738 EXPORT_SYMBOL(kernel_sock_ip_overhead
);