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/ethtool.h>
57 #include <linux/socket.h>
58 #include <linux/file.h>
59 #include <linux/net.h>
60 #include <linux/interrupt.h>
61 #include <linux/thread_info.h>
62 #include <linux/rcupdate.h>
63 #include <linux/netdevice.h>
64 #include <linux/proc_fs.h>
65 #include <linux/seq_file.h>
66 #include <linux/mutex.h>
67 #include <linux/if_bridge.h>
68 #include <linux/if_vlan.h>
69 #include <linux/ptp_classify.h>
70 #include <linux/init.h>
71 #include <linux/poll.h>
72 #include <linux/cache.h>
73 #include <linux/module.h>
74 #include <linux/highmem.h>
75 #include <linux/mount.h>
76 #include <linux/pseudo_fs.h>
77 #include <linux/security.h>
78 #include <linux/syscalls.h>
79 #include <linux/compat.h>
80 #include <linux/kmod.h>
81 #include <linux/audit.h>
82 #include <linux/wireless.h>
83 #include <linux/nsproxy.h>
84 #include <linux/magic.h>
85 #include <linux/slab.h>
86 #include <linux/xattr.h>
87 #include <linux/nospec.h>
88 #include <linux/indirect_call_wrapper.h>
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
95 #include <net/cls_cgroup.h>
98 #include <linux/netfilter.h>
100 #include <linux/if_tun.h>
101 #include <linux/ipv6_route.h>
102 #include <linux/route.h>
103 #include <linux/termios.h>
104 #include <linux/sockios.h>
105 #include <net/busy_poll.h>
106 #include <linux/errqueue.h>
107 #include <linux/ptp_clock_kernel.h>
109 #ifdef CONFIG_NET_RX_BUSY_POLL
110 unsigned int sysctl_net_busy_read __read_mostly
;
111 unsigned int sysctl_net_busy_poll __read_mostly
;
114 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
);
115 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
);
116 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
118 static int sock_close(struct inode
*inode
, struct file
*file
);
119 static __poll_t
sock_poll(struct file
*file
,
120 struct poll_table_struct
*wait
);
121 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
123 static long compat_sock_ioctl(struct file
*file
,
124 unsigned int cmd
, unsigned long arg
);
126 static int sock_fasync(int fd
, struct file
*filp
, int on
);
127 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
128 int offset
, size_t size
, loff_t
*ppos
, int more
);
129 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
130 struct pipe_inode_info
*pipe
, size_t len
,
133 #ifdef CONFIG_PROC_FS
134 static void sock_show_fdinfo(struct seq_file
*m
, struct file
*f
)
136 struct socket
*sock
= f
->private_data
;
138 if (sock
->ops
->show_fdinfo
)
139 sock
->ops
->show_fdinfo(m
, sock
);
142 #define sock_show_fdinfo NULL
146 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
147 * in the operation structures but are done directly via the socketcall() multiplexor.
150 static const struct file_operations socket_file_ops
= {
151 .owner
= THIS_MODULE
,
153 .read_iter
= sock_read_iter
,
154 .write_iter
= sock_write_iter
,
156 .unlocked_ioctl
= sock_ioctl
,
158 .compat_ioctl
= compat_sock_ioctl
,
161 .release
= sock_close
,
162 .fasync
= sock_fasync
,
163 .sendpage
= sock_sendpage
,
164 .splice_write
= generic_splice_sendpage
,
165 .splice_read
= sock_splice_read
,
166 .show_fdinfo
= sock_show_fdinfo
,
169 static const char * const pf_family_names
[] = {
170 [PF_UNSPEC
] = "PF_UNSPEC",
171 [PF_UNIX
] = "PF_UNIX/PF_LOCAL",
172 [PF_INET
] = "PF_INET",
173 [PF_AX25
] = "PF_AX25",
175 [PF_APPLETALK
] = "PF_APPLETALK",
176 [PF_NETROM
] = "PF_NETROM",
177 [PF_BRIDGE
] = "PF_BRIDGE",
178 [PF_ATMPVC
] = "PF_ATMPVC",
180 [PF_INET6
] = "PF_INET6",
181 [PF_ROSE
] = "PF_ROSE",
182 [PF_DECnet
] = "PF_DECnet",
183 [PF_NETBEUI
] = "PF_NETBEUI",
184 [PF_SECURITY
] = "PF_SECURITY",
186 [PF_NETLINK
] = "PF_NETLINK/PF_ROUTE",
187 [PF_PACKET
] = "PF_PACKET",
189 [PF_ECONET
] = "PF_ECONET",
190 [PF_ATMSVC
] = "PF_ATMSVC",
193 [PF_IRDA
] = "PF_IRDA",
194 [PF_PPPOX
] = "PF_PPPOX",
195 [PF_WANPIPE
] = "PF_WANPIPE",
198 [PF_MPLS
] = "PF_MPLS",
200 [PF_TIPC
] = "PF_TIPC",
201 [PF_BLUETOOTH
] = "PF_BLUETOOTH",
202 [PF_IUCV
] = "PF_IUCV",
203 [PF_RXRPC
] = "PF_RXRPC",
204 [PF_ISDN
] = "PF_ISDN",
205 [PF_PHONET
] = "PF_PHONET",
206 [PF_IEEE802154
] = "PF_IEEE802154",
207 [PF_CAIF
] = "PF_CAIF",
210 [PF_VSOCK
] = "PF_VSOCK",
212 [PF_QIPCRTR
] = "PF_QIPCRTR",
218 * The protocol list. Each protocol is registered in here.
221 static DEFINE_SPINLOCK(net_family_lock
);
222 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
226 * Move socket addresses back and forth across the kernel/user
227 * divide and look after the messy bits.
231 * move_addr_to_kernel - copy a socket address into kernel space
232 * @uaddr: Address in user space
233 * @kaddr: Address in kernel space
234 * @ulen: Length in user space
236 * The address is copied into kernel space. If the provided address is
237 * too long an error code of -EINVAL is returned. If the copy gives
238 * invalid addresses -EFAULT is returned. On a success 0 is returned.
241 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
243 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
247 if (copy_from_user(kaddr
, uaddr
, ulen
))
249 return audit_sockaddr(ulen
, kaddr
);
253 * move_addr_to_user - copy an address to user space
254 * @kaddr: kernel space address
255 * @klen: length of address in kernel
256 * @uaddr: user space address
257 * @ulen: pointer to user length field
259 * The value pointed to by ulen on entry is the buffer length available.
260 * This is overwritten with the buffer space used. -EINVAL is returned
261 * if an overlong buffer is specified or a negative buffer size. -EFAULT
262 * is returned if either the buffer or the length field are not
264 * After copying the data up to the limit the user specifies, the true
265 * length of the data is written over the length limit the user
266 * specified. Zero is returned for a success.
269 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
270 void __user
*uaddr
, int __user
*ulen
)
275 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
276 err
= get_user(len
, ulen
);
284 if (audit_sockaddr(klen
, kaddr
))
286 if (copy_to_user(uaddr
, kaddr
, len
))
290 * "fromlen shall refer to the value before truncation.."
293 return __put_user(klen
, ulen
);
296 static struct kmem_cache
*sock_inode_cachep __ro_after_init
;
298 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
300 struct socket_alloc
*ei
;
302 ei
= kmem_cache_alloc(sock_inode_cachep
, GFP_KERNEL
);
305 init_waitqueue_head(&ei
->socket
.wq
.wait
);
306 ei
->socket
.wq
.fasync_list
= NULL
;
307 ei
->socket
.wq
.flags
= 0;
309 ei
->socket
.state
= SS_UNCONNECTED
;
310 ei
->socket
.flags
= 0;
311 ei
->socket
.ops
= NULL
;
312 ei
->socket
.sk
= NULL
;
313 ei
->socket
.file
= NULL
;
315 return &ei
->vfs_inode
;
318 static void sock_free_inode(struct inode
*inode
)
320 struct socket_alloc
*ei
;
322 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
323 kmem_cache_free(sock_inode_cachep
, ei
);
326 static void init_once(void *foo
)
328 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
330 inode_init_once(&ei
->vfs_inode
);
333 static void init_inodecache(void)
335 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
336 sizeof(struct socket_alloc
),
338 (SLAB_HWCACHE_ALIGN
|
339 SLAB_RECLAIM_ACCOUNT
|
340 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
342 BUG_ON(sock_inode_cachep
== NULL
);
345 static const struct super_operations sockfs_ops
= {
346 .alloc_inode
= sock_alloc_inode
,
347 .free_inode
= sock_free_inode
,
348 .statfs
= simple_statfs
,
352 * sockfs_dname() is called from d_path().
354 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
356 return dynamic_dname(dentry
, buffer
, buflen
, "socket:[%lu]",
357 d_inode(dentry
)->i_ino
);
360 static const struct dentry_operations sockfs_dentry_operations
= {
361 .d_dname
= sockfs_dname
,
364 static int sockfs_xattr_get(const struct xattr_handler
*handler
,
365 struct dentry
*dentry
, struct inode
*inode
,
366 const char *suffix
, void *value
, size_t size
)
369 if (dentry
->d_name
.len
+ 1 > size
)
371 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
373 return dentry
->d_name
.len
+ 1;
376 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
377 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
378 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
380 static const struct xattr_handler sockfs_xattr_handler
= {
381 .name
= XATTR_NAME_SOCKPROTONAME
,
382 .get
= sockfs_xattr_get
,
385 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
386 struct user_namespace
*mnt_userns
,
387 struct dentry
*dentry
, struct inode
*inode
,
388 const char *suffix
, const void *value
,
389 size_t size
, int flags
)
391 /* Handled by LSM. */
395 static const struct xattr_handler sockfs_security_xattr_handler
= {
396 .prefix
= XATTR_SECURITY_PREFIX
,
397 .set
= sockfs_security_xattr_set
,
400 static const struct xattr_handler
*sockfs_xattr_handlers
[] = {
401 &sockfs_xattr_handler
,
402 &sockfs_security_xattr_handler
,
406 static int sockfs_init_fs_context(struct fs_context
*fc
)
408 struct pseudo_fs_context
*ctx
= init_pseudo(fc
, SOCKFS_MAGIC
);
411 ctx
->ops
= &sockfs_ops
;
412 ctx
->dops
= &sockfs_dentry_operations
;
413 ctx
->xattr
= sockfs_xattr_handlers
;
417 static struct vfsmount
*sock_mnt __read_mostly
;
419 static struct file_system_type sock_fs_type
= {
421 .init_fs_context
= sockfs_init_fs_context
,
422 .kill_sb
= kill_anon_super
,
426 * Obtains the first available file descriptor and sets it up for use.
428 * These functions create file structures and maps them to fd space
429 * of the current process. On success it returns file descriptor
430 * and file struct implicitly stored in sock->file.
431 * Note that another thread may close file descriptor before we return
432 * from this function. We use the fact that now we do not refer
433 * to socket after mapping. If one day we will need it, this
434 * function will increment ref. count on file by 1.
436 * In any case returned fd MAY BE not valid!
437 * This race condition is unavoidable
438 * with shared fd spaces, we cannot solve it inside kernel,
439 * but we take care of internal coherence yet.
443 * sock_alloc_file - Bind a &socket to a &file
445 * @flags: file status flags
446 * @dname: protocol name
448 * Returns the &file bound with @sock, implicitly storing it
449 * in sock->file. If dname is %NULL, sets to "".
450 * On failure the return is a ERR pointer (see linux/err.h).
451 * This function uses GFP_KERNEL internally.
454 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
459 dname
= sock
->sk
? sock
->sk
->sk_prot_creator
->name
: "";
461 file
= alloc_file_pseudo(SOCK_INODE(sock
), sock_mnt
, dname
,
462 O_RDWR
| (flags
& O_NONBLOCK
),
470 file
->private_data
= sock
;
471 stream_open(SOCK_INODE(sock
), file
);
474 EXPORT_SYMBOL(sock_alloc_file
);
476 static int sock_map_fd(struct socket
*sock
, int flags
)
478 struct file
*newfile
;
479 int fd
= get_unused_fd_flags(flags
);
480 if (unlikely(fd
< 0)) {
485 newfile
= sock_alloc_file(sock
, flags
, NULL
);
486 if (!IS_ERR(newfile
)) {
487 fd_install(fd
, newfile
);
492 return PTR_ERR(newfile
);
496 * sock_from_file - Return the &socket bounded to @file.
499 * On failure returns %NULL.
502 struct socket
*sock_from_file(struct file
*file
)
504 if (file
->f_op
== &socket_file_ops
)
505 return file
->private_data
; /* set in sock_map_fd */
509 EXPORT_SYMBOL(sock_from_file
);
512 * sockfd_lookup - Go from a file number to its socket slot
514 * @err: pointer to an error code return
516 * The file handle passed in is locked and the socket it is bound
517 * to is returned. If an error occurs the err pointer is overwritten
518 * with a negative errno code and NULL is returned. The function checks
519 * for both invalid handles and passing a handle which is not a socket.
521 * On a success the socket object pointer is returned.
524 struct socket
*sockfd_lookup(int fd
, int *err
)
535 sock
= sock_from_file(file
);
542 EXPORT_SYMBOL(sockfd_lookup
);
544 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
546 struct fd f
= fdget(fd
);
551 sock
= sock_from_file(f
.file
);
553 *fput_needed
= f
.flags
& FDPUT_FPUT
;
562 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
568 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
578 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
583 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
590 static int sockfs_setattr(struct user_namespace
*mnt_userns
,
591 struct dentry
*dentry
, struct iattr
*iattr
)
593 int err
= simple_setattr(&init_user_ns
, dentry
, iattr
);
595 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
596 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
599 sock
->sk
->sk_uid
= iattr
->ia_uid
;
607 static const struct inode_operations sockfs_inode_ops
= {
608 .listxattr
= sockfs_listxattr
,
609 .setattr
= sockfs_setattr
,
613 * sock_alloc - allocate a socket
615 * Allocate a new inode and socket object. The two are bound together
616 * and initialised. The socket is then returned. If we are out of inodes
617 * NULL is returned. This functions uses GFP_KERNEL internally.
620 struct socket
*sock_alloc(void)
625 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
629 sock
= SOCKET_I(inode
);
631 inode
->i_ino
= get_next_ino();
632 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
633 inode
->i_uid
= current_fsuid();
634 inode
->i_gid
= current_fsgid();
635 inode
->i_op
= &sockfs_inode_ops
;
639 EXPORT_SYMBOL(sock_alloc
);
641 static void __sock_release(struct socket
*sock
, struct inode
*inode
)
644 struct module
*owner
= sock
->ops
->owner
;
648 sock
->ops
->release(sock
);
656 if (sock
->wq
.fasync_list
)
657 pr_err("%s: fasync list not empty!\n", __func__
);
660 iput(SOCK_INODE(sock
));
667 * sock_release - close a socket
668 * @sock: socket to close
670 * The socket is released from the protocol stack if it has a release
671 * callback, and the inode is then released if the socket is bound to
672 * an inode not a file.
674 void sock_release(struct socket
*sock
)
676 __sock_release(sock
, NULL
);
678 EXPORT_SYMBOL(sock_release
);
680 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
682 u8 flags
= *tx_flags
;
684 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
)
685 flags
|= SKBTX_HW_TSTAMP
;
687 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
688 flags
|= SKBTX_SW_TSTAMP
;
690 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
691 flags
|= SKBTX_SCHED_TSTAMP
;
695 EXPORT_SYMBOL(__sock_tx_timestamp
);
697 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket
*, struct msghdr
*,
699 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket
*, struct msghdr
*,
701 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
703 int ret
= INDIRECT_CALL_INET(sock
->ops
->sendmsg
, inet6_sendmsg
,
704 inet_sendmsg
, sock
, msg
,
706 BUG_ON(ret
== -EIOCBQUEUED
);
711 * sock_sendmsg - send a message through @sock
713 * @msg: message to send
715 * Sends @msg through @sock, passing through LSM.
716 * Returns the number of bytes sent, or an error code.
718 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
720 int err
= security_socket_sendmsg(sock
, msg
,
723 return err
?: sock_sendmsg_nosec(sock
, msg
);
725 EXPORT_SYMBOL(sock_sendmsg
);
728 * kernel_sendmsg - send a message through @sock (kernel-space)
730 * @msg: message header
732 * @num: vec array length
733 * @size: total message data size
735 * Builds the message data with @vec and sends it through @sock.
736 * Returns the number of bytes sent, or an error code.
739 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
740 struct kvec
*vec
, size_t num
, size_t size
)
742 iov_iter_kvec(&msg
->msg_iter
, WRITE
, vec
, num
, size
);
743 return sock_sendmsg(sock
, msg
);
745 EXPORT_SYMBOL(kernel_sendmsg
);
748 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
750 * @msg: message header
751 * @vec: output s/g array
752 * @num: output s/g array length
753 * @size: total message data size
755 * Builds the message data with @vec and sends it through @sock.
756 * Returns the number of bytes sent, or an error code.
757 * Caller must hold @sk.
760 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
761 struct kvec
*vec
, size_t num
, size_t size
)
763 struct socket
*sock
= sk
->sk_socket
;
765 if (!sock
->ops
->sendmsg_locked
)
766 return sock_no_sendmsg_locked(sk
, msg
, size
);
768 iov_iter_kvec(&msg
->msg_iter
, WRITE
, vec
, num
, size
);
770 return sock
->ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
772 EXPORT_SYMBOL(kernel_sendmsg_locked
);
774 static bool skb_is_err_queue(const struct sk_buff
*skb
)
776 /* pkt_type of skbs enqueued on the error queue are set to
777 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
778 * in recvmsg, since skbs received on a local socket will never
779 * have a pkt_type of PACKET_OUTGOING.
781 return skb
->pkt_type
== PACKET_OUTGOING
;
784 /* On transmit, software and hardware timestamps are returned independently.
785 * As the two skb clones share the hardware timestamp, which may be updated
786 * before the software timestamp is received, a hardware TX timestamp may be
787 * returned only if there is no software TX timestamp. Ignore false software
788 * timestamps, which may be made in the __sock_recv_timestamp() call when the
789 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
790 * hardware timestamp.
792 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
794 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
797 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
)
799 struct scm_ts_pktinfo ts_pktinfo
;
800 struct net_device
*orig_dev
;
802 if (!skb_mac_header_was_set(skb
))
805 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
808 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
810 ts_pktinfo
.if_index
= orig_dev
->ifindex
;
813 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
814 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
815 sizeof(ts_pktinfo
), &ts_pktinfo
);
819 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
821 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
824 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
825 int new_tstamp
= sock_flag(sk
, SOCK_TSTAMP_NEW
);
826 struct scm_timestamping_internal tss
;
828 int empty
= 1, false_tstamp
= 0;
829 struct skb_shared_hwtstamps
*shhwtstamps
=
832 /* Race occurred between timestamp enabling and packet
833 receiving. Fill in the current time for now. */
834 if (need_software_tstamp
&& skb
->tstamp
== 0) {
835 __net_timestamp(skb
);
839 if (need_software_tstamp
) {
840 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
842 struct __kernel_sock_timeval tv
;
844 skb_get_new_timestamp(skb
, &tv
);
845 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_NEW
,
848 struct __kernel_old_timeval tv
;
850 skb_get_timestamp(skb
, &tv
);
851 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_OLD
,
856 struct __kernel_timespec ts
;
858 skb_get_new_timestampns(skb
, &ts
);
859 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_NEW
,
862 struct __kernel_old_timespec ts
;
864 skb_get_timestampns(skb
, &ts
);
865 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_OLD
,
871 memset(&tss
, 0, sizeof(tss
));
872 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
873 ktime_to_timespec64_cond(skb
->tstamp
, tss
.ts
+ 0))
876 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
877 !skb_is_swtx_tstamp(skb
, false_tstamp
)) {
878 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
879 ptp_convert_timestamp(shhwtstamps
, sk
->sk_bind_phc
);
881 if (ktime_to_timespec64_cond(shhwtstamps
->hwtstamp
,
885 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
886 !skb_is_err_queue(skb
))
887 put_ts_pktinfo(msg
, skb
);
891 if (sock_flag(sk
, SOCK_TSTAMP_NEW
))
892 put_cmsg_scm_timestamping64(msg
, &tss
);
894 put_cmsg_scm_timestamping(msg
, &tss
);
896 if (skb_is_err_queue(skb
) && skb
->len
&&
897 SKB_EXT_ERR(skb
)->opt_stats
)
898 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
899 skb
->len
, skb
->data
);
902 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
904 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
909 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
911 if (!skb
->wifi_acked_valid
)
914 ack
= skb
->wifi_acked
;
916 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
918 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
920 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
923 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
924 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
925 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
928 void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
931 sock_recv_timestamp(msg
, sk
, skb
);
932 sock_recv_drops(msg
, sk
, skb
);
934 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops
);
936 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket
*, struct msghdr
*,
938 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket
*, struct msghdr
*,
940 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
943 return INDIRECT_CALL_INET(sock
->ops
->recvmsg
, inet6_recvmsg
,
944 inet_recvmsg
, sock
, msg
, msg_data_left(msg
),
949 * sock_recvmsg - receive a message from @sock
951 * @msg: message to receive
952 * @flags: message flags
954 * Receives @msg from @sock, passing through LSM. Returns the total number
955 * of bytes received, or an error.
957 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
959 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
961 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
963 EXPORT_SYMBOL(sock_recvmsg
);
966 * kernel_recvmsg - Receive a message from a socket (kernel space)
967 * @sock: The socket to receive the message from
968 * @msg: Received message
969 * @vec: Input s/g array for message data
970 * @num: Size of input s/g array
971 * @size: Number of bytes to read
972 * @flags: Message flags (MSG_DONTWAIT, etc...)
974 * On return the msg structure contains the scatter/gather array passed in the
975 * vec argument. The array is modified so that it consists of the unfilled
976 * portion of the original array.
978 * The returned value is the total number of bytes received, or an error.
981 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
982 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
984 msg
->msg_control_is_user
= false;
985 iov_iter_kvec(&msg
->msg_iter
, READ
, vec
, num
, size
);
986 return sock_recvmsg(sock
, msg
, flags
);
988 EXPORT_SYMBOL(kernel_recvmsg
);
990 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
991 int offset
, size_t size
, loff_t
*ppos
, int more
)
996 sock
= file
->private_data
;
998 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
999 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1002 return kernel_sendpage(sock
, page
, offset
, size
, flags
);
1005 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
1006 struct pipe_inode_info
*pipe
, size_t len
,
1009 struct socket
*sock
= file
->private_data
;
1011 if (unlikely(!sock
->ops
->splice_read
))
1012 return generic_file_splice_read(file
, ppos
, pipe
, len
, flags
);
1014 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
1017 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1019 struct file
*file
= iocb
->ki_filp
;
1020 struct socket
*sock
= file
->private_data
;
1021 struct msghdr msg
= {.msg_iter
= *to
,
1025 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1026 msg
.msg_flags
= MSG_DONTWAIT
;
1028 if (iocb
->ki_pos
!= 0)
1031 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
1034 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
1039 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1041 struct file
*file
= iocb
->ki_filp
;
1042 struct socket
*sock
= file
->private_data
;
1043 struct msghdr msg
= {.msg_iter
= *from
,
1047 if (iocb
->ki_pos
!= 0)
1050 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1051 msg
.msg_flags
= MSG_DONTWAIT
;
1053 if (sock
->type
== SOCK_SEQPACKET
)
1054 msg
.msg_flags
|= MSG_EOR
;
1056 res
= sock_sendmsg(sock
, &msg
);
1057 *from
= msg
.msg_iter
;
1062 * Atomic setting of ioctl hooks to avoid race
1063 * with module unload.
1066 static DEFINE_MUTEX(br_ioctl_mutex
);
1067 static int (*br_ioctl_hook
) (struct net
*, unsigned int cmd
, void __user
*arg
);
1069 void brioctl_set(int (*hook
) (struct net
*, unsigned int, void __user
*))
1071 mutex_lock(&br_ioctl_mutex
);
1072 br_ioctl_hook
= hook
;
1073 mutex_unlock(&br_ioctl_mutex
);
1075 EXPORT_SYMBOL(brioctl_set
);
1077 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1078 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1080 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1082 mutex_lock(&vlan_ioctl_mutex
);
1083 vlan_ioctl_hook
= hook
;
1084 mutex_unlock(&vlan_ioctl_mutex
);
1086 EXPORT_SYMBOL(vlan_ioctl_set
);
1088 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1089 unsigned int cmd
, unsigned long arg
)
1092 void __user
*argp
= (void __user
*)arg
;
1094 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1097 * If this ioctl is unknown try to hand it down
1098 * to the NIC driver.
1100 if (err
!= -ENOIOCTLCMD
)
1103 if (cmd
== SIOCGIFCONF
) {
1105 if (copy_from_user(&ifc
, argp
, sizeof(struct ifconf
)))
1108 err
= dev_ifconf(net
, &ifc
, sizeof(struct ifreq
));
1110 if (!err
&& copy_to_user(argp
, &ifc
, sizeof(struct ifconf
)))
1115 if (copy_from_user(&ifr
, argp
, sizeof(struct ifreq
)))
1117 err
= dev_ioctl(net
, cmd
, &ifr
, &need_copyout
);
1118 if (!err
&& need_copyout
)
1119 if (copy_to_user(argp
, &ifr
, sizeof(struct ifreq
)))
1126 * With an ioctl, arg may well be a user mode pointer, but we don't know
1127 * what to do with it - that's up to the protocol still.
1130 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1132 struct socket
*sock
;
1134 void __user
*argp
= (void __user
*)arg
;
1138 sock
= file
->private_data
;
1141 if (unlikely(cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))) {
1144 if (copy_from_user(&ifr
, argp
, sizeof(struct ifreq
)))
1146 err
= dev_ioctl(net
, cmd
, &ifr
, &need_copyout
);
1147 if (!err
&& need_copyout
)
1148 if (copy_to_user(argp
, &ifr
, sizeof(struct ifreq
)))
1151 #ifdef CONFIG_WEXT_CORE
1152 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1153 err
= wext_handle_ioctl(net
, cmd
, argp
);
1160 if (get_user(pid
, (int __user
*)argp
))
1162 err
= f_setown(sock
->file
, pid
, 1);
1166 err
= put_user(f_getown(sock
->file
),
1167 (int __user
*)argp
);
1175 request_module("bridge");
1177 mutex_lock(&br_ioctl_mutex
);
1179 err
= br_ioctl_hook(net
, cmd
, argp
);
1180 mutex_unlock(&br_ioctl_mutex
);
1185 if (!vlan_ioctl_hook
)
1186 request_module("8021q");
1188 mutex_lock(&vlan_ioctl_mutex
);
1189 if (vlan_ioctl_hook
)
1190 err
= vlan_ioctl_hook(net
, argp
);
1191 mutex_unlock(&vlan_ioctl_mutex
);
1195 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1198 err
= open_related_ns(&net
->ns
, get_net_ns
);
1200 case SIOCGSTAMP_OLD
:
1201 case SIOCGSTAMPNS_OLD
:
1202 if (!sock
->ops
->gettstamp
) {
1206 err
= sock
->ops
->gettstamp(sock
, argp
,
1207 cmd
== SIOCGSTAMP_OLD
,
1208 !IS_ENABLED(CONFIG_64BIT
));
1210 case SIOCGSTAMP_NEW
:
1211 case SIOCGSTAMPNS_NEW
:
1212 if (!sock
->ops
->gettstamp
) {
1216 err
= sock
->ops
->gettstamp(sock
, argp
,
1217 cmd
== SIOCGSTAMP_NEW
,
1221 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1228 * sock_create_lite - creates a socket
1229 * @family: protocol family (AF_INET, ...)
1230 * @type: communication type (SOCK_STREAM, ...)
1231 * @protocol: protocol (0, ...)
1234 * Creates a new socket and assigns it to @res, passing through LSM.
1235 * The new socket initialization is not complete, see kernel_accept().
1236 * Returns 0 or an error. On failure @res is set to %NULL.
1237 * This function internally uses GFP_KERNEL.
1240 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1243 struct socket
*sock
= NULL
;
1245 err
= security_socket_create(family
, type
, protocol
, 1);
1249 sock
= sock_alloc();
1256 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1268 EXPORT_SYMBOL(sock_create_lite
);
1270 /* No kernel lock held - perfect */
1271 static __poll_t
sock_poll(struct file
*file
, poll_table
*wait
)
1273 struct socket
*sock
= file
->private_data
;
1274 __poll_t events
= poll_requested_events(wait
), flag
= 0;
1276 if (!sock
->ops
->poll
)
1279 if (sk_can_busy_loop(sock
->sk
)) {
1280 /* poll once if requested by the syscall */
1281 if (events
& POLL_BUSY_LOOP
)
1282 sk_busy_loop(sock
->sk
, 1);
1284 /* if this socket can poll_ll, tell the system call */
1285 flag
= POLL_BUSY_LOOP
;
1288 return sock
->ops
->poll(file
, sock
, wait
) | flag
;
1291 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1293 struct socket
*sock
= file
->private_data
;
1295 return sock
->ops
->mmap(file
, sock
, vma
);
1298 static int sock_close(struct inode
*inode
, struct file
*filp
)
1300 __sock_release(SOCKET_I(inode
), inode
);
1305 * Update the socket async list
1307 * Fasync_list locking strategy.
1309 * 1. fasync_list is modified only under process context socket lock
1310 * i.e. under semaphore.
1311 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1312 * or under socket lock
1315 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1317 struct socket
*sock
= filp
->private_data
;
1318 struct sock
*sk
= sock
->sk
;
1319 struct socket_wq
*wq
= &sock
->wq
;
1325 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1327 if (!wq
->fasync_list
)
1328 sock_reset_flag(sk
, SOCK_FASYNC
);
1330 sock_set_flag(sk
, SOCK_FASYNC
);
1336 /* This function may be called only under rcu_lock */
1338 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1340 if (!wq
|| !wq
->fasync_list
)
1344 case SOCK_WAKE_WAITD
:
1345 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1348 case SOCK_WAKE_SPACE
:
1349 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1354 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1357 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1362 EXPORT_SYMBOL(sock_wake_async
);
1365 * __sock_create - creates a socket
1366 * @net: net namespace
1367 * @family: protocol family (AF_INET, ...)
1368 * @type: communication type (SOCK_STREAM, ...)
1369 * @protocol: protocol (0, ...)
1371 * @kern: boolean for kernel space sockets
1373 * Creates a new socket and assigns it to @res, passing through LSM.
1374 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1375 * be set to true if the socket resides in kernel space.
1376 * This function internally uses GFP_KERNEL.
1379 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1380 struct socket
**res
, int kern
)
1383 struct socket
*sock
;
1384 const struct net_proto_family
*pf
;
1387 * Check protocol is in range
1389 if (family
< 0 || family
>= NPROTO
)
1390 return -EAFNOSUPPORT
;
1391 if (type
< 0 || type
>= SOCK_MAX
)
1396 This uglymoron is moved from INET layer to here to avoid
1397 deadlock in module load.
1399 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1400 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1405 err
= security_socket_create(family
, type
, protocol
, kern
);
1410 * Allocate the socket and allow the family to set things up. if
1411 * the protocol is 0, the family is instructed to select an appropriate
1414 sock
= sock_alloc();
1416 net_warn_ratelimited("socket: no more sockets\n");
1417 return -ENFILE
; /* Not exactly a match, but its the
1418 closest posix thing */
1423 #ifdef CONFIG_MODULES
1424 /* Attempt to load a protocol module if the find failed.
1426 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1427 * requested real, full-featured networking support upon configuration.
1428 * Otherwise module support will break!
1430 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1431 request_module("net-pf-%d", family
);
1435 pf
= rcu_dereference(net_families
[family
]);
1436 err
= -EAFNOSUPPORT
;
1441 * We will call the ->create function, that possibly is in a loadable
1442 * module, so we have to bump that loadable module refcnt first.
1444 if (!try_module_get(pf
->owner
))
1447 /* Now protected by module ref count */
1450 err
= pf
->create(net
, sock
, protocol
, kern
);
1452 goto out_module_put
;
1455 * Now to bump the refcnt of the [loadable] module that owns this
1456 * socket at sock_release time we decrement its refcnt.
1458 if (!try_module_get(sock
->ops
->owner
))
1459 goto out_module_busy
;
1462 * Now that we're done with the ->create function, the [loadable]
1463 * module can have its refcnt decremented
1465 module_put(pf
->owner
);
1466 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1468 goto out_sock_release
;
1474 err
= -EAFNOSUPPORT
;
1477 module_put(pf
->owner
);
1484 goto out_sock_release
;
1486 EXPORT_SYMBOL(__sock_create
);
1489 * sock_create - creates a socket
1490 * @family: protocol family (AF_INET, ...)
1491 * @type: communication type (SOCK_STREAM, ...)
1492 * @protocol: protocol (0, ...)
1495 * A wrapper around __sock_create().
1496 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1499 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1501 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1503 EXPORT_SYMBOL(sock_create
);
1506 * sock_create_kern - creates a socket (kernel space)
1507 * @net: net namespace
1508 * @family: protocol family (AF_INET, ...)
1509 * @type: communication type (SOCK_STREAM, ...)
1510 * @protocol: protocol (0, ...)
1513 * A wrapper around __sock_create().
1514 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1517 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1519 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1521 EXPORT_SYMBOL(sock_create_kern
);
1523 int __sys_socket(int family
, int type
, int protocol
)
1526 struct socket
*sock
;
1529 /* Check the SOCK_* constants for consistency. */
1530 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1531 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1532 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1533 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1535 flags
= type
& ~SOCK_TYPE_MASK
;
1536 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1538 type
&= SOCK_TYPE_MASK
;
1540 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1541 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1543 retval
= sock_create(family
, type
, protocol
, &sock
);
1547 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1550 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1552 return __sys_socket(family
, type
, protocol
);
1556 * Create a pair of connected sockets.
1559 int __sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1561 struct socket
*sock1
, *sock2
;
1563 struct file
*newfile1
, *newfile2
;
1566 flags
= type
& ~SOCK_TYPE_MASK
;
1567 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1569 type
&= SOCK_TYPE_MASK
;
1571 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1572 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1575 * reserve descriptors and make sure we won't fail
1576 * to return them to userland.
1578 fd1
= get_unused_fd_flags(flags
);
1579 if (unlikely(fd1
< 0))
1582 fd2
= get_unused_fd_flags(flags
);
1583 if (unlikely(fd2
< 0)) {
1588 err
= put_user(fd1
, &usockvec
[0]);
1592 err
= put_user(fd2
, &usockvec
[1]);
1597 * Obtain the first socket and check if the underlying protocol
1598 * supports the socketpair call.
1601 err
= sock_create(family
, type
, protocol
, &sock1
);
1602 if (unlikely(err
< 0))
1605 err
= sock_create(family
, type
, protocol
, &sock2
);
1606 if (unlikely(err
< 0)) {
1607 sock_release(sock1
);
1611 err
= security_socket_socketpair(sock1
, sock2
);
1612 if (unlikely(err
)) {
1613 sock_release(sock2
);
1614 sock_release(sock1
);
1618 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1619 if (unlikely(err
< 0)) {
1620 sock_release(sock2
);
1621 sock_release(sock1
);
1625 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1626 if (IS_ERR(newfile1
)) {
1627 err
= PTR_ERR(newfile1
);
1628 sock_release(sock2
);
1632 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1633 if (IS_ERR(newfile2
)) {
1634 err
= PTR_ERR(newfile2
);
1639 audit_fd_pair(fd1
, fd2
);
1641 fd_install(fd1
, newfile1
);
1642 fd_install(fd2
, newfile2
);
1651 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1652 int __user
*, usockvec
)
1654 return __sys_socketpair(family
, type
, protocol
, usockvec
);
1658 * Bind a name to a socket. Nothing much to do here since it's
1659 * the protocol's responsibility to handle the local address.
1661 * We move the socket address to kernel space before we call
1662 * the protocol layer (having also checked the address is ok).
1665 int __sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1667 struct socket
*sock
;
1668 struct sockaddr_storage address
;
1669 int err
, fput_needed
;
1671 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1673 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1675 err
= security_socket_bind(sock
,
1676 (struct sockaddr
*)&address
,
1679 err
= sock
->ops
->bind(sock
,
1683 fput_light(sock
->file
, fput_needed
);
1688 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1690 return __sys_bind(fd
, umyaddr
, addrlen
);
1694 * Perform a listen. Basically, we allow the protocol to do anything
1695 * necessary for a listen, and if that works, we mark the socket as
1696 * ready for listening.
1699 int __sys_listen(int fd
, int backlog
)
1701 struct socket
*sock
;
1702 int err
, fput_needed
;
1705 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1707 somaxconn
= sock_net(sock
->sk
)->core
.sysctl_somaxconn
;
1708 if ((unsigned int)backlog
> somaxconn
)
1709 backlog
= somaxconn
;
1711 err
= security_socket_listen(sock
, backlog
);
1713 err
= sock
->ops
->listen(sock
, backlog
);
1715 fput_light(sock
->file
, fput_needed
);
1720 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1722 return __sys_listen(fd
, backlog
);
1725 int __sys_accept4_file(struct file
*file
, unsigned file_flags
,
1726 struct sockaddr __user
*upeer_sockaddr
,
1727 int __user
*upeer_addrlen
, int flags
,
1728 unsigned long nofile
)
1730 struct socket
*sock
, *newsock
;
1731 struct file
*newfile
;
1732 int err
, len
, newfd
;
1733 struct sockaddr_storage address
;
1735 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1738 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1739 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1741 sock
= sock_from_file(file
);
1748 newsock
= sock_alloc();
1752 newsock
->type
= sock
->type
;
1753 newsock
->ops
= sock
->ops
;
1756 * We don't need try_module_get here, as the listening socket (sock)
1757 * has the protocol module (sock->ops->owner) held.
1759 __module_get(newsock
->ops
->owner
);
1761 newfd
= __get_unused_fd_flags(flags
, nofile
);
1762 if (unlikely(newfd
< 0)) {
1764 sock_release(newsock
);
1767 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1768 if (IS_ERR(newfile
)) {
1769 err
= PTR_ERR(newfile
);
1770 put_unused_fd(newfd
);
1774 err
= security_socket_accept(sock
, newsock
);
1778 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
| file_flags
,
1783 if (upeer_sockaddr
) {
1784 len
= newsock
->ops
->getname(newsock
,
1785 (struct sockaddr
*)&address
, 2);
1787 err
= -ECONNABORTED
;
1790 err
= move_addr_to_user(&address
,
1791 len
, upeer_sockaddr
, upeer_addrlen
);
1796 /* File flags are not inherited via accept() unlike another OSes. */
1798 fd_install(newfd
, newfile
);
1804 put_unused_fd(newfd
);
1810 * For accept, we attempt to create a new socket, set up the link
1811 * with the client, wake up the client, then return the new
1812 * connected fd. We collect the address of the connector in kernel
1813 * space and move it to user at the very end. This is unclean because
1814 * we open the socket then return an error.
1816 * 1003.1g adds the ability to recvmsg() to query connection pending
1817 * status to recvmsg. We need to add that support in a way thats
1818 * clean when we restructure accept also.
1821 int __sys_accept4(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1822 int __user
*upeer_addrlen
, int flags
)
1829 ret
= __sys_accept4_file(f
.file
, 0, upeer_sockaddr
,
1830 upeer_addrlen
, flags
,
1831 rlimit(RLIMIT_NOFILE
));
1838 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1839 int __user
*, upeer_addrlen
, int, flags
)
1841 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, flags
);
1844 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1845 int __user
*, upeer_addrlen
)
1847 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1851 * Attempt to connect to a socket with the server address. The address
1852 * is in user space so we verify it is OK and move it to kernel space.
1854 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1857 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1858 * other SEQPACKET protocols that take time to connect() as it doesn't
1859 * include the -EINPROGRESS status for such sockets.
1862 int __sys_connect_file(struct file
*file
, struct sockaddr_storage
*address
,
1863 int addrlen
, int file_flags
)
1865 struct socket
*sock
;
1868 sock
= sock_from_file(file
);
1875 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
1879 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
1880 sock
->file
->f_flags
| file_flags
);
1885 int __sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
1892 struct sockaddr_storage address
;
1894 ret
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
1896 ret
= __sys_connect_file(f
.file
, &address
, addrlen
, 0);
1903 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
1906 return __sys_connect(fd
, uservaddr
, addrlen
);
1910 * Get the local address ('name') of a socket object. Move the obtained
1911 * name to user space.
1914 int __sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
1915 int __user
*usockaddr_len
)
1917 struct socket
*sock
;
1918 struct sockaddr_storage address
;
1919 int err
, fput_needed
;
1921 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1925 err
= security_socket_getsockname(sock
);
1929 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 0);
1932 /* "err" is actually length in this case */
1933 err
= move_addr_to_user(&address
, err
, usockaddr
, usockaddr_len
);
1936 fput_light(sock
->file
, fput_needed
);
1941 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
1942 int __user
*, usockaddr_len
)
1944 return __sys_getsockname(fd
, usockaddr
, usockaddr_len
);
1948 * Get the remote address ('name') of a socket object. Move the obtained
1949 * name to user space.
1952 int __sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
1953 int __user
*usockaddr_len
)
1955 struct socket
*sock
;
1956 struct sockaddr_storage address
;
1957 int err
, fput_needed
;
1959 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1961 err
= security_socket_getpeername(sock
);
1963 fput_light(sock
->file
, fput_needed
);
1967 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 1);
1969 /* "err" is actually length in this case */
1970 err
= move_addr_to_user(&address
, err
, usockaddr
,
1972 fput_light(sock
->file
, fput_needed
);
1977 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
1978 int __user
*, usockaddr_len
)
1980 return __sys_getpeername(fd
, usockaddr
, usockaddr_len
);
1984 * Send a datagram to a given address. We move the address into kernel
1985 * space and check the user space data area is readable before invoking
1988 int __sys_sendto(int fd
, void __user
*buff
, size_t len
, unsigned int flags
,
1989 struct sockaddr __user
*addr
, int addr_len
)
1991 struct socket
*sock
;
1992 struct sockaddr_storage address
;
1998 err
= import_single_range(WRITE
, buff
, len
, &iov
, &msg
.msg_iter
);
2001 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2005 msg
.msg_name
= NULL
;
2006 msg
.msg_control
= NULL
;
2007 msg
.msg_controllen
= 0;
2008 msg
.msg_namelen
= 0;
2010 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
2013 msg
.msg_name
= (struct sockaddr
*)&address
;
2014 msg
.msg_namelen
= addr_len
;
2016 if (sock
->file
->f_flags
& O_NONBLOCK
)
2017 flags
|= MSG_DONTWAIT
;
2018 msg
.msg_flags
= flags
;
2019 err
= sock_sendmsg(sock
, &msg
);
2022 fput_light(sock
->file
, fput_needed
);
2027 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
2028 unsigned int, flags
, struct sockaddr __user
*, addr
,
2031 return __sys_sendto(fd
, buff
, len
, flags
, addr
, addr_len
);
2035 * Send a datagram down a socket.
2038 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
2039 unsigned int, flags
)
2041 return __sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
2045 * Receive a frame from the socket and optionally record the address of the
2046 * sender. We verify the buffers are writable and if needed move the
2047 * sender address from kernel to user space.
2049 int __sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
, unsigned int flags
,
2050 struct sockaddr __user
*addr
, int __user
*addr_len
)
2052 struct socket
*sock
;
2055 struct sockaddr_storage address
;
2059 err
= import_single_range(READ
, ubuf
, size
, &iov
, &msg
.msg_iter
);
2062 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2066 msg
.msg_control
= NULL
;
2067 msg
.msg_controllen
= 0;
2068 /* Save some cycles and don't copy the address if not needed */
2069 msg
.msg_name
= addr
? (struct sockaddr
*)&address
: NULL
;
2070 /* We assume all kernel code knows the size of sockaddr_storage */
2071 msg
.msg_namelen
= 0;
2072 msg
.msg_iocb
= NULL
;
2074 if (sock
->file
->f_flags
& O_NONBLOCK
)
2075 flags
|= MSG_DONTWAIT
;
2076 err
= sock_recvmsg(sock
, &msg
, flags
);
2078 if (err
>= 0 && addr
!= NULL
) {
2079 err2
= move_addr_to_user(&address
,
2080 msg
.msg_namelen
, addr
, addr_len
);
2085 fput_light(sock
->file
, fput_needed
);
2090 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
2091 unsigned int, flags
, struct sockaddr __user
*, addr
,
2092 int __user
*, addr_len
)
2094 return __sys_recvfrom(fd
, ubuf
, size
, flags
, addr
, addr_len
);
2098 * Receive a datagram from a socket.
2101 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
2102 unsigned int, flags
)
2104 return __sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
2107 static bool sock_use_custom_sol_socket(const struct socket
*sock
)
2109 const struct sock
*sk
= sock
->sk
;
2111 /* Use sock->ops->setsockopt() for MPTCP */
2112 return IS_ENABLED(CONFIG_MPTCP
) &&
2113 sk
->sk_protocol
== IPPROTO_MPTCP
&&
2114 sk
->sk_type
== SOCK_STREAM
&&
2115 (sk
->sk_family
== AF_INET
|| sk
->sk_family
== AF_INET6
);
2119 * Set a socket option. Because we don't know the option lengths we have
2120 * to pass the user mode parameter for the protocols to sort out.
2122 int __sys_setsockopt(int fd
, int level
, int optname
, char __user
*user_optval
,
2125 sockptr_t optval
= USER_SOCKPTR(user_optval
);
2126 char *kernel_optval
= NULL
;
2127 int err
, fput_needed
;
2128 struct socket
*sock
;
2133 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2137 err
= security_socket_setsockopt(sock
, level
, optname
);
2141 if (!in_compat_syscall())
2142 err
= BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock
->sk
, &level
, &optname
,
2143 user_optval
, &optlen
,
2153 optval
= KERNEL_SOCKPTR(kernel_optval
);
2154 if (level
== SOL_SOCKET
&& !sock_use_custom_sol_socket(sock
))
2155 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2156 else if (unlikely(!sock
->ops
->setsockopt
))
2159 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2161 kfree(kernel_optval
);
2163 fput_light(sock
->file
, fput_needed
);
2167 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
2168 char __user
*, optval
, int, optlen
)
2170 return __sys_setsockopt(fd
, level
, optname
, optval
, optlen
);
2173 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level
,
2177 * Get a socket option. Because we don't know the option lengths we have
2178 * to pass a user mode parameter for the protocols to sort out.
2180 int __sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
,
2183 int err
, fput_needed
;
2184 struct socket
*sock
;
2187 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2191 err
= security_socket_getsockopt(sock
, level
, optname
);
2195 if (!in_compat_syscall())
2196 max_optlen
= BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen
);
2198 if (level
== SOL_SOCKET
)
2199 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2200 else if (unlikely(!sock
->ops
->getsockopt
))
2203 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2206 if (!in_compat_syscall())
2207 err
= BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock
->sk
, level
, optname
,
2208 optval
, optlen
, max_optlen
,
2211 fput_light(sock
->file
, fput_needed
);
2215 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
2216 char __user
*, optval
, int __user
*, optlen
)
2218 return __sys_getsockopt(fd
, level
, optname
, optval
, optlen
);
2222 * Shutdown a socket.
2225 int __sys_shutdown_sock(struct socket
*sock
, int how
)
2229 err
= security_socket_shutdown(sock
, how
);
2231 err
= sock
->ops
->shutdown(sock
, how
);
2236 int __sys_shutdown(int fd
, int how
)
2238 int err
, fput_needed
;
2239 struct socket
*sock
;
2241 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2243 err
= __sys_shutdown_sock(sock
, how
);
2244 fput_light(sock
->file
, fput_needed
);
2249 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
2251 return __sys_shutdown(fd
, how
);
2254 /* A couple of helpful macros for getting the address of the 32/64 bit
2255 * fields which are the same type (int / unsigned) on our platforms.
2257 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2258 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2259 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2261 struct used_address
{
2262 struct sockaddr_storage name
;
2263 unsigned int name_len
;
2266 int __copy_msghdr_from_user(struct msghdr
*kmsg
,
2267 struct user_msghdr __user
*umsg
,
2268 struct sockaddr __user
**save_addr
,
2269 struct iovec __user
**uiov
, size_t *nsegs
)
2271 struct user_msghdr msg
;
2274 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
2277 kmsg
->msg_control_is_user
= true;
2278 kmsg
->msg_control_user
= msg
.msg_control
;
2279 kmsg
->msg_controllen
= msg
.msg_controllen
;
2280 kmsg
->msg_flags
= msg
.msg_flags
;
2282 kmsg
->msg_namelen
= msg
.msg_namelen
;
2284 kmsg
->msg_namelen
= 0;
2286 if (kmsg
->msg_namelen
< 0)
2289 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2290 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2293 *save_addr
= msg
.msg_name
;
2295 if (msg
.msg_name
&& kmsg
->msg_namelen
) {
2297 err
= move_addr_to_kernel(msg
.msg_name
,
2304 kmsg
->msg_name
= NULL
;
2305 kmsg
->msg_namelen
= 0;
2308 if (msg
.msg_iovlen
> UIO_MAXIOV
)
2311 kmsg
->msg_iocb
= NULL
;
2312 *uiov
= msg
.msg_iov
;
2313 *nsegs
= msg
.msg_iovlen
;
2317 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
2318 struct user_msghdr __user
*umsg
,
2319 struct sockaddr __user
**save_addr
,
2322 struct user_msghdr msg
;
2325 err
= __copy_msghdr_from_user(kmsg
, umsg
, save_addr
, &msg
.msg_iov
,
2330 err
= import_iovec(save_addr
? READ
: WRITE
,
2331 msg
.msg_iov
, msg
.msg_iovlen
,
2332 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
2333 return err
< 0 ? err
: 0;
2336 static int ____sys_sendmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2337 unsigned int flags
, struct used_address
*used_address
,
2338 unsigned int allowed_msghdr_flags
)
2340 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2341 __aligned(sizeof(__kernel_size_t
));
2342 /* 20 is size of ipv6_pktinfo */
2343 unsigned char *ctl_buf
= ctl
;
2349 if (msg_sys
->msg_controllen
> INT_MAX
)
2351 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
2352 ctl_len
= msg_sys
->msg_controllen
;
2353 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2355 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2359 ctl_buf
= msg_sys
->msg_control
;
2360 ctl_len
= msg_sys
->msg_controllen
;
2361 } else if (ctl_len
) {
2362 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2363 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2364 if (ctl_len
> sizeof(ctl
)) {
2365 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2366 if (ctl_buf
== NULL
)
2370 if (copy_from_user(ctl_buf
, msg_sys
->msg_control_user
, ctl_len
))
2372 msg_sys
->msg_control
= ctl_buf
;
2373 msg_sys
->msg_control_is_user
= false;
2375 msg_sys
->msg_flags
= flags
;
2377 if (sock
->file
->f_flags
& O_NONBLOCK
)
2378 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2380 * If this is sendmmsg() and current destination address is same as
2381 * previously succeeded address, omit asking LSM's decision.
2382 * used_address->name_len is initialized to UINT_MAX so that the first
2383 * destination address never matches.
2385 if (used_address
&& msg_sys
->msg_name
&&
2386 used_address
->name_len
== msg_sys
->msg_namelen
&&
2387 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2388 used_address
->name_len
)) {
2389 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2392 err
= sock_sendmsg(sock
, msg_sys
);
2394 * If this is sendmmsg() and sending to current destination address was
2395 * successful, remember it.
2397 if (used_address
&& err
>= 0) {
2398 used_address
->name_len
= msg_sys
->msg_namelen
;
2399 if (msg_sys
->msg_name
)
2400 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2401 used_address
->name_len
);
2406 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2411 int sendmsg_copy_msghdr(struct msghdr
*msg
,
2412 struct user_msghdr __user
*umsg
, unsigned flags
,
2417 if (flags
& MSG_CMSG_COMPAT
) {
2418 struct compat_msghdr __user
*msg_compat
;
2420 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2421 err
= get_compat_msghdr(msg
, msg_compat
, NULL
, iov
);
2423 err
= copy_msghdr_from_user(msg
, umsg
, NULL
, iov
);
2431 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2432 struct msghdr
*msg_sys
, unsigned int flags
,
2433 struct used_address
*used_address
,
2434 unsigned int allowed_msghdr_flags
)
2436 struct sockaddr_storage address
;
2437 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2440 msg_sys
->msg_name
= &address
;
2442 err
= sendmsg_copy_msghdr(msg_sys
, msg
, flags
, &iov
);
2446 err
= ____sys_sendmsg(sock
, msg_sys
, flags
, used_address
,
2447 allowed_msghdr_flags
);
2453 * BSD sendmsg interface
2455 long __sys_sendmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2458 return ____sys_sendmsg(sock
, msg
, flags
, NULL
, 0);
2461 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2462 bool forbid_cmsg_compat
)
2464 int fput_needed
, err
;
2465 struct msghdr msg_sys
;
2466 struct socket
*sock
;
2468 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2471 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2475 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2477 fput_light(sock
->file
, fput_needed
);
2482 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2484 return __sys_sendmsg(fd
, msg
, flags
, true);
2488 * Linux sendmmsg interface
2491 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2492 unsigned int flags
, bool forbid_cmsg_compat
)
2494 int fput_needed
, err
, datagrams
;
2495 struct socket
*sock
;
2496 struct mmsghdr __user
*entry
;
2497 struct compat_mmsghdr __user
*compat_entry
;
2498 struct msghdr msg_sys
;
2499 struct used_address used_address
;
2500 unsigned int oflags
= flags
;
2502 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2505 if (vlen
> UIO_MAXIOV
)
2510 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2514 used_address
.name_len
= UINT_MAX
;
2516 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2520 while (datagrams
< vlen
) {
2521 if (datagrams
== vlen
- 1)
2524 if (MSG_CMSG_COMPAT
& flags
) {
2525 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2526 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2529 err
= __put_user(err
, &compat_entry
->msg_len
);
2532 err
= ___sys_sendmsg(sock
,
2533 (struct user_msghdr __user
*)entry
,
2534 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2537 err
= put_user(err
, &entry
->msg_len
);
2544 if (msg_data_left(&msg_sys
))
2549 fput_light(sock
->file
, fput_needed
);
2551 /* We only return an error if no datagrams were able to be sent */
2558 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2559 unsigned int, vlen
, unsigned int, flags
)
2561 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
, true);
2564 int recvmsg_copy_msghdr(struct msghdr
*msg
,
2565 struct user_msghdr __user
*umsg
, unsigned flags
,
2566 struct sockaddr __user
**uaddr
,
2571 if (MSG_CMSG_COMPAT
& flags
) {
2572 struct compat_msghdr __user
*msg_compat
;
2574 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2575 err
= get_compat_msghdr(msg
, msg_compat
, uaddr
, iov
);
2577 err
= copy_msghdr_from_user(msg
, umsg
, uaddr
, iov
);
2585 static int ____sys_recvmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2586 struct user_msghdr __user
*msg
,
2587 struct sockaddr __user
*uaddr
,
2588 unsigned int flags
, int nosec
)
2590 struct compat_msghdr __user
*msg_compat
=
2591 (struct compat_msghdr __user
*) msg
;
2592 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2593 struct sockaddr_storage addr
;
2594 unsigned long cmsg_ptr
;
2598 msg_sys
->msg_name
= &addr
;
2599 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2600 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2602 /* We assume all kernel code knows the size of sockaddr_storage */
2603 msg_sys
->msg_namelen
= 0;
2605 if (sock
->file
->f_flags
& O_NONBLOCK
)
2606 flags
|= MSG_DONTWAIT
;
2608 if (unlikely(nosec
))
2609 err
= sock_recvmsg_nosec(sock
, msg_sys
, flags
);
2611 err
= sock_recvmsg(sock
, msg_sys
, flags
);
2617 if (uaddr
!= NULL
) {
2618 err
= move_addr_to_user(&addr
,
2619 msg_sys
->msg_namelen
, uaddr
,
2624 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2628 if (MSG_CMSG_COMPAT
& flags
)
2629 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2630 &msg_compat
->msg_controllen
);
2632 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2633 &msg
->msg_controllen
);
2641 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2642 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2644 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2645 /* user mode address pointers */
2646 struct sockaddr __user
*uaddr
;
2649 err
= recvmsg_copy_msghdr(msg_sys
, msg
, flags
, &uaddr
, &iov
);
2653 err
= ____sys_recvmsg(sock
, msg_sys
, msg
, uaddr
, flags
, nosec
);
2659 * BSD recvmsg interface
2662 long __sys_recvmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2663 struct user_msghdr __user
*umsg
,
2664 struct sockaddr __user
*uaddr
, unsigned int flags
)
2666 return ____sys_recvmsg(sock
, msg
, umsg
, uaddr
, flags
, 0);
2669 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2670 bool forbid_cmsg_compat
)
2672 int fput_needed
, err
;
2673 struct msghdr msg_sys
;
2674 struct socket
*sock
;
2676 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2679 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2683 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2685 fput_light(sock
->file
, fput_needed
);
2690 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2691 unsigned int, flags
)
2693 return __sys_recvmsg(fd
, msg
, flags
, true);
2697 * Linux recvmmsg interface
2700 static int do_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2701 unsigned int vlen
, unsigned int flags
,
2702 struct timespec64
*timeout
)
2704 int fput_needed
, err
, datagrams
;
2705 struct socket
*sock
;
2706 struct mmsghdr __user
*entry
;
2707 struct compat_mmsghdr __user
*compat_entry
;
2708 struct msghdr msg_sys
;
2709 struct timespec64 end_time
;
2710 struct timespec64 timeout64
;
2713 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2719 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2723 if (likely(!(flags
& MSG_ERRQUEUE
))) {
2724 err
= sock_error(sock
->sk
);
2732 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2734 while (datagrams
< vlen
) {
2736 * No need to ask LSM for more than the first datagram.
2738 if (MSG_CMSG_COMPAT
& flags
) {
2739 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2740 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2744 err
= __put_user(err
, &compat_entry
->msg_len
);
2747 err
= ___sys_recvmsg(sock
,
2748 (struct user_msghdr __user
*)entry
,
2749 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2753 err
= put_user(err
, &entry
->msg_len
);
2761 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2762 if (flags
& MSG_WAITFORONE
)
2763 flags
|= MSG_DONTWAIT
;
2766 ktime_get_ts64(&timeout64
);
2767 *timeout
= timespec64_sub(end_time
, timeout64
);
2768 if (timeout
->tv_sec
< 0) {
2769 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2773 /* Timeout, return less than vlen datagrams */
2774 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2778 /* Out of band data, return right away */
2779 if (msg_sys
.msg_flags
& MSG_OOB
)
2787 if (datagrams
== 0) {
2793 * We may return less entries than requested (vlen) if the
2794 * sock is non block and there aren't enough datagrams...
2796 if (err
!= -EAGAIN
) {
2798 * ... or if recvmsg returns an error after we
2799 * received some datagrams, where we record the
2800 * error to return on the next call or if the
2801 * app asks about it using getsockopt(SO_ERROR).
2803 sock
->sk
->sk_err
= -err
;
2806 fput_light(sock
->file
, fput_needed
);
2811 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2812 unsigned int vlen
, unsigned int flags
,
2813 struct __kernel_timespec __user
*timeout
,
2814 struct old_timespec32 __user
*timeout32
)
2817 struct timespec64 timeout_sys
;
2819 if (timeout
&& get_timespec64(&timeout_sys
, timeout
))
2822 if (timeout32
&& get_old_timespec32(&timeout_sys
, timeout32
))
2825 if (!timeout
&& !timeout32
)
2826 return do_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2828 datagrams
= do_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2833 if (timeout
&& put_timespec64(&timeout_sys
, timeout
))
2834 datagrams
= -EFAULT
;
2836 if (timeout32
&& put_old_timespec32(&timeout_sys
, timeout32
))
2837 datagrams
= -EFAULT
;
2842 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2843 unsigned int, vlen
, unsigned int, flags
,
2844 struct __kernel_timespec __user
*, timeout
)
2846 if (flags
& MSG_CMSG_COMPAT
)
2849 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, timeout
, NULL
);
2852 #ifdef CONFIG_COMPAT_32BIT_TIME
2853 SYSCALL_DEFINE5(recvmmsg_time32
, int, fd
, struct mmsghdr __user
*, mmsg
,
2854 unsigned int, vlen
, unsigned int, flags
,
2855 struct old_timespec32 __user
*, timeout
)
2857 if (flags
& MSG_CMSG_COMPAT
)
2860 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
, timeout
);
2864 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2865 /* Argument list sizes for sys_socketcall */
2866 #define AL(x) ((x) * sizeof(unsigned long))
2867 static const unsigned char nargs
[21] = {
2868 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2869 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2870 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2877 * System call vectors.
2879 * Argument checking cleaned up. Saved 20% in size.
2880 * This function doesn't need to set the kernel lock because
2881 * it is set by the callees.
2884 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2886 unsigned long a
[AUDITSC_ARGS
];
2887 unsigned long a0
, a1
;
2891 if (call
< 1 || call
> SYS_SENDMMSG
)
2893 call
= array_index_nospec(call
, SYS_SENDMMSG
+ 1);
2896 if (len
> sizeof(a
))
2899 /* copy_from_user should be SMP safe. */
2900 if (copy_from_user(a
, args
, len
))
2903 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
2912 err
= __sys_socket(a0
, a1
, a
[2]);
2915 err
= __sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2918 err
= __sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
2921 err
= __sys_listen(a0
, a1
);
2924 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2925 (int __user
*)a
[2], 0);
2927 case SYS_GETSOCKNAME
:
2929 __sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
2930 (int __user
*)a
[2]);
2932 case SYS_GETPEERNAME
:
2934 __sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
2935 (int __user
*)a
[2]);
2937 case SYS_SOCKETPAIR
:
2938 err
= __sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
2941 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2945 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
2946 (struct sockaddr __user
*)a
[4], a
[5]);
2949 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2953 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
2954 (struct sockaddr __user
*)a
[4],
2955 (int __user
*)a
[5]);
2958 err
= __sys_shutdown(a0
, a1
);
2960 case SYS_SETSOCKOPT
:
2961 err
= __sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2964 case SYS_GETSOCKOPT
:
2966 __sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
2967 (int __user
*)a
[4]);
2970 err
= __sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
,
2974 err
= __sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2],
2978 err
= __sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
,
2982 if (IS_ENABLED(CONFIG_64BIT
))
2983 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
2985 (struct __kernel_timespec __user
*)a
[4],
2988 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
2990 (struct old_timespec32 __user
*)a
[4]);
2993 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
2994 (int __user
*)a
[2], a
[3]);
3003 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3006 * sock_register - add a socket protocol handler
3007 * @ops: description of protocol
3009 * This function is called by a protocol handler that wants to
3010 * advertise its address family, and have it linked into the
3011 * socket interface. The value ops->family corresponds to the
3012 * socket system call protocol family.
3014 int sock_register(const struct net_proto_family
*ops
)
3018 if (ops
->family
>= NPROTO
) {
3019 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
3023 spin_lock(&net_family_lock
);
3024 if (rcu_dereference_protected(net_families
[ops
->family
],
3025 lockdep_is_held(&net_family_lock
)))
3028 rcu_assign_pointer(net_families
[ops
->family
], ops
);
3031 spin_unlock(&net_family_lock
);
3033 pr_info("NET: Registered %s protocol family\n", pf_family_names
[ops
->family
]);
3036 EXPORT_SYMBOL(sock_register
);
3039 * sock_unregister - remove a protocol handler
3040 * @family: protocol family to remove
3042 * This function is called by a protocol handler that wants to
3043 * remove its address family, and have it unlinked from the
3044 * new socket creation.
3046 * If protocol handler is a module, then it can use module reference
3047 * counts to protect against new references. If protocol handler is not
3048 * a module then it needs to provide its own protection in
3049 * the ops->create routine.
3051 void sock_unregister(int family
)
3053 BUG_ON(family
< 0 || family
>= NPROTO
);
3055 spin_lock(&net_family_lock
);
3056 RCU_INIT_POINTER(net_families
[family
], NULL
);
3057 spin_unlock(&net_family_lock
);
3061 pr_info("NET: Unregistered %s protocol family\n", pf_family_names
[family
]);
3063 EXPORT_SYMBOL(sock_unregister
);
3065 bool sock_is_registered(int family
)
3067 return family
< NPROTO
&& rcu_access_pointer(net_families
[family
]);
3070 static int __init
sock_init(void)
3074 * Initialize the network sysctl infrastructure.
3076 err
= net_sysctl_init();
3081 * Initialize skbuff SLAB cache
3086 * Initialize the protocols module.
3091 err
= register_filesystem(&sock_fs_type
);
3094 sock_mnt
= kern_mount(&sock_fs_type
);
3095 if (IS_ERR(sock_mnt
)) {
3096 err
= PTR_ERR(sock_mnt
);
3100 /* The real protocol initialization is performed in later initcalls.
3103 #ifdef CONFIG_NETFILTER
3104 err
= netfilter_init();
3109 ptp_classifier_init();
3115 unregister_filesystem(&sock_fs_type
);
3119 core_initcall(sock_init
); /* early initcall */
3121 #ifdef CONFIG_PROC_FS
3122 void socket_seq_show(struct seq_file
*seq
)
3124 seq_printf(seq
, "sockets: used %d\n",
3125 sock_inuse_get(seq
->private));
3127 #endif /* CONFIG_PROC_FS */
3129 #ifdef CONFIG_COMPAT
3130 static int compat_dev_ifconf(struct net
*net
, struct compat_ifconf __user
*uifc32
)
3132 struct compat_ifconf ifc32
;
3136 if (copy_from_user(&ifc32
, uifc32
, sizeof(struct compat_ifconf
)))
3139 ifc
.ifc_len
= ifc32
.ifc_len
;
3140 ifc
.ifc_req
= compat_ptr(ifc32
.ifcbuf
);
3143 err
= dev_ifconf(net
, &ifc
, sizeof(struct compat_ifreq
));
3148 ifc32
.ifc_len
= ifc
.ifc_len
;
3149 if (copy_to_user(uifc32
, &ifc32
, sizeof(struct compat_ifconf
)))
3155 static int ethtool_ioctl(struct net
*net
, struct compat_ifreq __user
*ifr32
)
3157 struct compat_ethtool_rxnfc __user
*compat_rxnfc
;
3158 bool convert_in
= false, convert_out
= false;
3159 size_t buf_size
= 0;
3160 struct ethtool_rxnfc __user
*rxnfc
= NULL
;
3162 u32 rule_cnt
= 0, actual_rule_cnt
;
3167 if (get_user(data
, &ifr32
->ifr_ifru
.ifru_data
))
3170 compat_rxnfc
= compat_ptr(data
);
3172 if (get_user(ethcmd
, &compat_rxnfc
->cmd
))
3175 /* Most ethtool structures are defined without padding.
3176 * Unfortunately struct ethtool_rxnfc is an exception.
3181 case ETHTOOL_GRXCLSRLALL
:
3182 /* Buffer size is variable */
3183 if (get_user(rule_cnt
, &compat_rxnfc
->rule_cnt
))
3185 if (rule_cnt
> KMALLOC_MAX_SIZE
/ sizeof(u32
))
3187 buf_size
+= rule_cnt
* sizeof(u32
);
3189 case ETHTOOL_GRXRINGS
:
3190 case ETHTOOL_GRXCLSRLCNT
:
3191 case ETHTOOL_GRXCLSRULE
:
3192 case ETHTOOL_SRXCLSRLINS
:
3195 case ETHTOOL_SRXCLSRLDEL
:
3196 buf_size
+= sizeof(struct ethtool_rxnfc
);
3198 rxnfc
= compat_alloc_user_space(buf_size
);
3202 if (copy_from_user(&ifr
.ifr_name
, &ifr32
->ifr_name
, IFNAMSIZ
))
3205 ifr
.ifr_data
= convert_in
? rxnfc
: (void __user
*)compat_rxnfc
;
3208 /* We expect there to be holes between fs.m_ext and
3209 * fs.ring_cookie and at the end of fs, but nowhere else.
3211 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc
, fs
.m_ext
) +
3212 sizeof(compat_rxnfc
->fs
.m_ext
) !=
3213 offsetof(struct ethtool_rxnfc
, fs
.m_ext
) +
3214 sizeof(rxnfc
->fs
.m_ext
));
3216 offsetof(struct compat_ethtool_rxnfc
, fs
.location
) -
3217 offsetof(struct compat_ethtool_rxnfc
, fs
.ring_cookie
) !=
3218 offsetof(struct ethtool_rxnfc
, fs
.location
) -
3219 offsetof(struct ethtool_rxnfc
, fs
.ring_cookie
));
3221 if (copy_in_user(rxnfc
, compat_rxnfc
,
3222 (void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
3223 (void __user
*)rxnfc
) ||
3224 copy_in_user(&rxnfc
->fs
.ring_cookie
,
3225 &compat_rxnfc
->fs
.ring_cookie
,
3226 (void __user
*)(&rxnfc
->fs
.location
+ 1) -
3227 (void __user
*)&rxnfc
->fs
.ring_cookie
))
3229 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
3230 if (put_user(rule_cnt
, &rxnfc
->rule_cnt
))
3232 } else if (copy_in_user(&rxnfc
->rule_cnt
,
3233 &compat_rxnfc
->rule_cnt
,
3234 sizeof(rxnfc
->rule_cnt
)))
3238 ret
= dev_ioctl(net
, SIOCETHTOOL
, &ifr
, NULL
);
3243 if (copy_in_user(compat_rxnfc
, rxnfc
,
3244 (const void __user
*)(&rxnfc
->fs
.m_ext
+ 1) -
3245 (const void __user
*)rxnfc
) ||
3246 copy_in_user(&compat_rxnfc
->fs
.ring_cookie
,
3247 &rxnfc
->fs
.ring_cookie
,
3248 (const void __user
*)(&rxnfc
->fs
.location
+ 1) -
3249 (const void __user
*)&rxnfc
->fs
.ring_cookie
) ||
3250 copy_in_user(&compat_rxnfc
->rule_cnt
, &rxnfc
->rule_cnt
,
3251 sizeof(rxnfc
->rule_cnt
)))
3254 if (ethcmd
== ETHTOOL_GRXCLSRLALL
) {
3255 /* As an optimisation, we only copy the actual
3256 * number of rules that the underlying
3257 * function returned. Since Mallory might
3258 * change the rule count in user memory, we
3259 * check that it is less than the rule count
3260 * originally given (as the user buffer size),
3261 * which has been range-checked.
3263 if (get_user(actual_rule_cnt
, &rxnfc
->rule_cnt
))
3265 if (actual_rule_cnt
< rule_cnt
)
3266 rule_cnt
= actual_rule_cnt
;
3267 if (copy_in_user(&compat_rxnfc
->rule_locs
[0],
3268 &rxnfc
->rule_locs
[0],
3269 rule_cnt
* sizeof(u32
)))
3277 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3279 compat_uptr_t uptr32
;
3284 if (copy_from_user(&ifr
, uifr32
, sizeof(struct compat_ifreq
)))
3287 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
3290 saved
= ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
;
3291 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= compat_ptr(uptr32
);
3293 err
= dev_ioctl(net
, SIOCWANDEV
, &ifr
, NULL
);
3295 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= saved
;
3296 if (copy_to_user(uifr32
, &ifr
, sizeof(struct compat_ifreq
)))
3302 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3303 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3304 struct compat_ifreq __user
*u_ifreq32
)
3309 if (copy_from_user(ifreq
.ifr_name
, u_ifreq32
->ifr_name
, IFNAMSIZ
))
3311 if (get_user(data32
, &u_ifreq32
->ifr_data
))
3313 ifreq
.ifr_data
= compat_ptr(data32
);
3315 return dev_ioctl(net
, cmd
, &ifreq
, NULL
);
3318 static int compat_ifreq_ioctl(struct net
*net
, struct socket
*sock
,
3320 struct compat_ifreq __user
*uifr32
)
3322 struct ifreq __user
*uifr
;
3325 /* Handle the fact that while struct ifreq has the same *layout* on
3326 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3327 * which are handled elsewhere, it still has different *size* due to
3328 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3329 * resulting in struct ifreq being 32 and 40 bytes respectively).
3330 * As a result, if the struct happens to be at the end of a page and
3331 * the next page isn't readable/writable, we get a fault. To prevent
3332 * that, copy back and forth to the full size.
3335 uifr
= compat_alloc_user_space(sizeof(*uifr
));
3336 if (copy_in_user(uifr
, uifr32
, sizeof(*uifr32
)))
3339 err
= sock_do_ioctl(net
, sock
, cmd
, (unsigned long)uifr
);
3350 case SIOCGIFBRDADDR
:
3351 case SIOCGIFDSTADDR
:
3352 case SIOCGIFNETMASK
:
3358 if (copy_in_user(uifr32
, uifr
, sizeof(*uifr32
)))
3366 static int compat_sioc_ifmap(struct net
*net
, unsigned int cmd
,
3367 struct compat_ifreq __user
*uifr32
)
3370 struct compat_ifmap __user
*uifmap32
;
3373 uifmap32
= &uifr32
->ifr_ifru
.ifru_map
;
3374 err
= copy_from_user(&ifr
, uifr32
, sizeof(ifr
.ifr_name
));
3375 err
|= get_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3376 err
|= get_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3377 err
|= get_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3378 err
|= get_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3379 err
|= get_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3380 err
|= get_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3384 err
= dev_ioctl(net
, cmd
, &ifr
, NULL
);
3386 if (cmd
== SIOCGIFMAP
&& !err
) {
3387 err
= copy_to_user(uifr32
, &ifr
, sizeof(ifr
.ifr_name
));
3388 err
|= put_user(ifr
.ifr_map
.mem_start
, &uifmap32
->mem_start
);
3389 err
|= put_user(ifr
.ifr_map
.mem_end
, &uifmap32
->mem_end
);
3390 err
|= put_user(ifr
.ifr_map
.base_addr
, &uifmap32
->base_addr
);
3391 err
|= put_user(ifr
.ifr_map
.irq
, &uifmap32
->irq
);
3392 err
|= put_user(ifr
.ifr_map
.dma
, &uifmap32
->dma
);
3393 err
|= put_user(ifr
.ifr_map
.port
, &uifmap32
->port
);
3400 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3401 * for some operations; this forces use of the newer bridge-utils that
3402 * use compatible ioctls
3404 static int old_bridge_ioctl(compat_ulong_t __user
*argp
)
3408 if (get_user(tmp
, argp
))
3410 if (tmp
== BRCTL_GET_VERSION
)
3411 return BRCTL_VERSION
+ 1;
3415 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3416 unsigned int cmd
, unsigned long arg
)
3418 void __user
*argp
= compat_ptr(arg
);
3419 struct sock
*sk
= sock
->sk
;
3420 struct net
*net
= sock_net(sk
);
3422 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3423 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3428 return old_bridge_ioctl(argp
);
3430 return compat_dev_ifconf(net
, argp
);
3432 return ethtool_ioctl(net
, argp
);
3434 return compat_siocwandev(net
, argp
);
3437 return compat_sioc_ifmap(net
, cmd
, argp
);
3438 case SIOCGSTAMP_OLD
:
3439 case SIOCGSTAMPNS_OLD
:
3440 if (!sock
->ops
->gettstamp
)
3441 return -ENOIOCTLCMD
;
3442 return sock
->ops
->gettstamp(sock
, argp
, cmd
== SIOCGSTAMP_OLD
,
3443 !COMPAT_USE_64BIT_TIME
);
3445 case SIOCBONDSLAVEINFOQUERY
:
3446 case SIOCBONDINFOQUERY
:
3449 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3460 case SIOCGSTAMP_NEW
:
3461 case SIOCGSTAMPNS_NEW
:
3462 return sock_ioctl(file
, cmd
, arg
);
3479 case SIOCSIFHWBROADCAST
:
3481 case SIOCGIFBRDADDR
:
3482 case SIOCSIFBRDADDR
:
3483 case SIOCGIFDSTADDR
:
3484 case SIOCSIFDSTADDR
:
3485 case SIOCGIFNETMASK
:
3486 case SIOCSIFNETMASK
:
3498 case SIOCBONDENSLAVE
:
3499 case SIOCBONDRELEASE
:
3500 case SIOCBONDSETHWADDR
:
3501 case SIOCBONDCHANGEACTIVE
:
3502 return compat_ifreq_ioctl(net
, sock
, cmd
, argp
);
3510 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3513 return -ENOIOCTLCMD
;
3516 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3519 struct socket
*sock
= file
->private_data
;
3520 int ret
= -ENOIOCTLCMD
;
3527 if (sock
->ops
->compat_ioctl
)
3528 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3530 if (ret
== -ENOIOCTLCMD
&&
3531 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3532 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3534 if (ret
== -ENOIOCTLCMD
)
3535 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3542 * kernel_bind - bind an address to a socket (kernel space)
3545 * @addrlen: length of address
3547 * Returns 0 or an error.
3550 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3552 return sock
->ops
->bind(sock
, addr
, addrlen
);
3554 EXPORT_SYMBOL(kernel_bind
);
3557 * kernel_listen - move socket to listening state (kernel space)
3559 * @backlog: pending connections queue size
3561 * Returns 0 or an error.
3564 int kernel_listen(struct socket
*sock
, int backlog
)
3566 return sock
->ops
->listen(sock
, backlog
);
3568 EXPORT_SYMBOL(kernel_listen
);
3571 * kernel_accept - accept a connection (kernel space)
3572 * @sock: listening socket
3573 * @newsock: new connected socket
3576 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3577 * If it fails, @newsock is guaranteed to be %NULL.
3578 * Returns 0 or an error.
3581 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3583 struct sock
*sk
= sock
->sk
;
3586 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3591 err
= sock
->ops
->accept(sock
, *newsock
, flags
, true);
3593 sock_release(*newsock
);
3598 (*newsock
)->ops
= sock
->ops
;
3599 __module_get((*newsock
)->ops
->owner
);
3604 EXPORT_SYMBOL(kernel_accept
);
3607 * kernel_connect - connect a socket (kernel space)
3610 * @addrlen: address length
3611 * @flags: flags (O_NONBLOCK, ...)
3613 * For datagram sockets, @addr is the address to which datagrams are sent
3614 * by default, and the only address from which datagrams are received.
3615 * For stream sockets, attempts to connect to @addr.
3616 * Returns 0 or an error code.
3619 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3622 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3624 EXPORT_SYMBOL(kernel_connect
);
3627 * kernel_getsockname - get the address which the socket is bound (kernel space)
3629 * @addr: address holder
3631 * Fills the @addr pointer with the address which the socket is bound.
3632 * Returns 0 or an error code.
3635 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
)
3637 return sock
->ops
->getname(sock
, addr
, 0);
3639 EXPORT_SYMBOL(kernel_getsockname
);
3642 * kernel_getpeername - get the address which the socket is connected (kernel space)
3644 * @addr: address holder
3646 * Fills the @addr pointer with the address which the socket is connected.
3647 * Returns 0 or an error code.
3650 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
)
3652 return sock
->ops
->getname(sock
, addr
, 1);
3654 EXPORT_SYMBOL(kernel_getpeername
);
3657 * kernel_sendpage - send a &page through a socket (kernel space)
3660 * @offset: page offset
3661 * @size: total size in bytes
3662 * @flags: flags (MSG_DONTWAIT, ...)
3664 * Returns the total amount sent in bytes or an error.
3667 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3668 size_t size
, int flags
)
3670 if (sock
->ops
->sendpage
) {
3671 /* Warn in case the improper page to zero-copy send */
3672 WARN_ONCE(!sendpage_ok(page
), "improper page for zero-copy send");
3673 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3675 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3677 EXPORT_SYMBOL(kernel_sendpage
);
3680 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3683 * @offset: page offset
3684 * @size: total size in bytes
3685 * @flags: flags (MSG_DONTWAIT, ...)
3687 * Returns the total amount sent in bytes or an error.
3688 * Caller must hold @sk.
3691 int kernel_sendpage_locked(struct sock
*sk
, struct page
*page
, int offset
,
3692 size_t size
, int flags
)
3694 struct socket
*sock
= sk
->sk_socket
;
3696 if (sock
->ops
->sendpage_locked
)
3697 return sock
->ops
->sendpage_locked(sk
, page
, offset
, size
,
3700 return sock_no_sendpage_locked(sk
, page
, offset
, size
, flags
);
3702 EXPORT_SYMBOL(kernel_sendpage_locked
);
3705 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3707 * @how: connection part
3709 * Returns 0 or an error.
3712 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3714 return sock
->ops
->shutdown(sock
, how
);
3716 EXPORT_SYMBOL(kernel_sock_shutdown
);
3719 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3722 * This routine returns the IP overhead imposed by a socket i.e.
3723 * the length of the underlying IP header, depending on whether
3724 * this is an IPv4 or IPv6 socket and the length from IP options turned
3725 * on at the socket. Assumes that the caller has a lock on the socket.
3728 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3730 struct inet_sock
*inet
;
3731 struct ip_options_rcu
*opt
;
3733 #if IS_ENABLED(CONFIG_IPV6)
3734 struct ipv6_pinfo
*np
;
3735 struct ipv6_txoptions
*optv6
= NULL
;
3736 #endif /* IS_ENABLED(CONFIG_IPV6) */
3741 switch (sk
->sk_family
) {
3744 overhead
+= sizeof(struct iphdr
);
3745 opt
= rcu_dereference_protected(inet
->inet_opt
,
3746 sock_owned_by_user(sk
));
3748 overhead
+= opt
->opt
.optlen
;
3750 #if IS_ENABLED(CONFIG_IPV6)
3753 overhead
+= sizeof(struct ipv6hdr
);
3755 optv6
= rcu_dereference_protected(np
->opt
,
3756 sock_owned_by_user(sk
));
3758 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3760 #endif /* IS_ENABLED(CONFIG_IPV6) */
3761 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3765 EXPORT_SYMBOL(kernel_sock_ip_overhead
);