1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * NET An implementation of the SOCKET network access protocol.
5 * Version: @(#)socket.c 1.1.93 18/02/95
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
49 * This module is effectively the top level interface to the BSD socket
52 * Based upon Swansea University Computer Society NET3.039
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
96 #include <net/cls_cgroup.h>
99 #include <linux/netfilter.h>
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
109 #include <trace/events/sock.h>
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly
;
113 unsigned int sysctl_net_busy_poll __read_mostly
;
116 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
);
117 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
);
118 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
);
120 static int sock_close(struct inode
*inode
, struct file
*file
);
121 static __poll_t
sock_poll(struct file
*file
,
122 struct poll_table_struct
*wait
);
123 static long sock_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
);
125 static long compat_sock_ioctl(struct file
*file
,
126 unsigned int cmd
, unsigned long arg
);
128 static int sock_fasync(int fd
, struct file
*filp
, int on
);
129 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
130 int offset
, size_t size
, loff_t
*ppos
, int more
);
131 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
132 struct pipe_inode_info
*pipe
, size_t len
,
135 #ifdef CONFIG_PROC_FS
136 static void sock_show_fdinfo(struct seq_file
*m
, struct file
*f
)
138 struct socket
*sock
= f
->private_data
;
140 if (sock
->ops
->show_fdinfo
)
141 sock
->ops
->show_fdinfo(m
, sock
);
144 #define sock_show_fdinfo NULL
148 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
149 * in the operation structures but are done directly via the socketcall() multiplexor.
152 static const struct file_operations socket_file_ops
= {
153 .owner
= THIS_MODULE
,
155 .read_iter
= sock_read_iter
,
156 .write_iter
= sock_write_iter
,
158 .unlocked_ioctl
= sock_ioctl
,
160 .compat_ioctl
= compat_sock_ioctl
,
163 .release
= sock_close
,
164 .fasync
= sock_fasync
,
165 .sendpage
= sock_sendpage
,
166 .splice_write
= generic_splice_sendpage
,
167 .splice_read
= sock_splice_read
,
168 .show_fdinfo
= sock_show_fdinfo
,
171 static const char * const pf_family_names
[] = {
172 [PF_UNSPEC
] = "PF_UNSPEC",
173 [PF_UNIX
] = "PF_UNIX/PF_LOCAL",
174 [PF_INET
] = "PF_INET",
175 [PF_AX25
] = "PF_AX25",
177 [PF_APPLETALK
] = "PF_APPLETALK",
178 [PF_NETROM
] = "PF_NETROM",
179 [PF_BRIDGE
] = "PF_BRIDGE",
180 [PF_ATMPVC
] = "PF_ATMPVC",
182 [PF_INET6
] = "PF_INET6",
183 [PF_ROSE
] = "PF_ROSE",
184 [PF_DECnet
] = "PF_DECnet",
185 [PF_NETBEUI
] = "PF_NETBEUI",
186 [PF_SECURITY
] = "PF_SECURITY",
188 [PF_NETLINK
] = "PF_NETLINK/PF_ROUTE",
189 [PF_PACKET
] = "PF_PACKET",
191 [PF_ECONET
] = "PF_ECONET",
192 [PF_ATMSVC
] = "PF_ATMSVC",
195 [PF_IRDA
] = "PF_IRDA",
196 [PF_PPPOX
] = "PF_PPPOX",
197 [PF_WANPIPE
] = "PF_WANPIPE",
200 [PF_MPLS
] = "PF_MPLS",
202 [PF_TIPC
] = "PF_TIPC",
203 [PF_BLUETOOTH
] = "PF_BLUETOOTH",
204 [PF_IUCV
] = "PF_IUCV",
205 [PF_RXRPC
] = "PF_RXRPC",
206 [PF_ISDN
] = "PF_ISDN",
207 [PF_PHONET
] = "PF_PHONET",
208 [PF_IEEE802154
] = "PF_IEEE802154",
209 [PF_CAIF
] = "PF_CAIF",
212 [PF_VSOCK
] = "PF_VSOCK",
214 [PF_QIPCRTR
] = "PF_QIPCRTR",
217 [PF_MCTP
] = "PF_MCTP",
221 * The protocol list. Each protocol is registered in here.
224 static DEFINE_SPINLOCK(net_family_lock
);
225 static const struct net_proto_family __rcu
*net_families
[NPROTO
] __read_mostly
;
229 * Move socket addresses back and forth across the kernel/user
230 * divide and look after the messy bits.
234 * move_addr_to_kernel - copy a socket address into kernel space
235 * @uaddr: Address in user space
236 * @kaddr: Address in kernel space
237 * @ulen: Length in user space
239 * The address is copied into kernel space. If the provided address is
240 * too long an error code of -EINVAL is returned. If the copy gives
241 * invalid addresses -EFAULT is returned. On a success 0 is returned.
244 int move_addr_to_kernel(void __user
*uaddr
, int ulen
, struct sockaddr_storage
*kaddr
)
246 if (ulen
< 0 || ulen
> sizeof(struct sockaddr_storage
))
250 if (copy_from_user(kaddr
, uaddr
, ulen
))
252 return audit_sockaddr(ulen
, kaddr
);
256 * move_addr_to_user - copy an address to user space
257 * @kaddr: kernel space address
258 * @klen: length of address in kernel
259 * @uaddr: user space address
260 * @ulen: pointer to user length field
262 * The value pointed to by ulen on entry is the buffer length available.
263 * This is overwritten with the buffer space used. -EINVAL is returned
264 * if an overlong buffer is specified or a negative buffer size. -EFAULT
265 * is returned if either the buffer or the length field are not
267 * After copying the data up to the limit the user specifies, the true
268 * length of the data is written over the length limit the user
269 * specified. Zero is returned for a success.
272 static int move_addr_to_user(struct sockaddr_storage
*kaddr
, int klen
,
273 void __user
*uaddr
, int __user
*ulen
)
278 BUG_ON(klen
> sizeof(struct sockaddr_storage
));
279 err
= get_user(len
, ulen
);
287 if (audit_sockaddr(klen
, kaddr
))
289 if (copy_to_user(uaddr
, kaddr
, len
))
293 * "fromlen shall refer to the value before truncation.."
296 return __put_user(klen
, ulen
);
299 static struct kmem_cache
*sock_inode_cachep __ro_after_init
;
301 static struct inode
*sock_alloc_inode(struct super_block
*sb
)
303 struct socket_alloc
*ei
;
305 ei
= alloc_inode_sb(sb
, sock_inode_cachep
, GFP_KERNEL
);
308 init_waitqueue_head(&ei
->socket
.wq
.wait
);
309 ei
->socket
.wq
.fasync_list
= NULL
;
310 ei
->socket
.wq
.flags
= 0;
312 ei
->socket
.state
= SS_UNCONNECTED
;
313 ei
->socket
.flags
= 0;
314 ei
->socket
.ops
= NULL
;
315 ei
->socket
.sk
= NULL
;
316 ei
->socket
.file
= NULL
;
318 return &ei
->vfs_inode
;
321 static void sock_free_inode(struct inode
*inode
)
323 struct socket_alloc
*ei
;
325 ei
= container_of(inode
, struct socket_alloc
, vfs_inode
);
326 kmem_cache_free(sock_inode_cachep
, ei
);
329 static void init_once(void *foo
)
331 struct socket_alloc
*ei
= (struct socket_alloc
*)foo
;
333 inode_init_once(&ei
->vfs_inode
);
336 static void init_inodecache(void)
338 sock_inode_cachep
= kmem_cache_create("sock_inode_cache",
339 sizeof(struct socket_alloc
),
341 (SLAB_HWCACHE_ALIGN
|
342 SLAB_RECLAIM_ACCOUNT
|
343 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
345 BUG_ON(sock_inode_cachep
== NULL
);
348 static const struct super_operations sockfs_ops
= {
349 .alloc_inode
= sock_alloc_inode
,
350 .free_inode
= sock_free_inode
,
351 .statfs
= simple_statfs
,
355 * sockfs_dname() is called from d_path().
357 static char *sockfs_dname(struct dentry
*dentry
, char *buffer
, int buflen
)
359 return dynamic_dname(buffer
, buflen
, "socket:[%lu]",
360 d_inode(dentry
)->i_ino
);
363 static const struct dentry_operations sockfs_dentry_operations
= {
364 .d_dname
= sockfs_dname
,
367 static int sockfs_xattr_get(const struct xattr_handler
*handler
,
368 struct dentry
*dentry
, struct inode
*inode
,
369 const char *suffix
, void *value
, size_t size
)
372 if (dentry
->d_name
.len
+ 1 > size
)
374 memcpy(value
, dentry
->d_name
.name
, dentry
->d_name
.len
+ 1);
376 return dentry
->d_name
.len
+ 1;
379 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
380 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
381 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
383 static const struct xattr_handler sockfs_xattr_handler
= {
384 .name
= XATTR_NAME_SOCKPROTONAME
,
385 .get
= sockfs_xattr_get
,
388 static int sockfs_security_xattr_set(const struct xattr_handler
*handler
,
389 struct mnt_idmap
*idmap
,
390 struct dentry
*dentry
, struct inode
*inode
,
391 const char *suffix
, const void *value
,
392 size_t size
, int flags
)
394 /* Handled by LSM. */
398 static const struct xattr_handler sockfs_security_xattr_handler
= {
399 .prefix
= XATTR_SECURITY_PREFIX
,
400 .set
= sockfs_security_xattr_set
,
403 static const struct xattr_handler
*sockfs_xattr_handlers
[] = {
404 &sockfs_xattr_handler
,
405 &sockfs_security_xattr_handler
,
409 static int sockfs_init_fs_context(struct fs_context
*fc
)
411 struct pseudo_fs_context
*ctx
= init_pseudo(fc
, SOCKFS_MAGIC
);
414 ctx
->ops
= &sockfs_ops
;
415 ctx
->dops
= &sockfs_dentry_operations
;
416 ctx
->xattr
= sockfs_xattr_handlers
;
420 static struct vfsmount
*sock_mnt __read_mostly
;
422 static struct file_system_type sock_fs_type
= {
424 .init_fs_context
= sockfs_init_fs_context
,
425 .kill_sb
= kill_anon_super
,
429 * Obtains the first available file descriptor and sets it up for use.
431 * These functions create file structures and maps them to fd space
432 * of the current process. On success it returns file descriptor
433 * and file struct implicitly stored in sock->file.
434 * Note that another thread may close file descriptor before we return
435 * from this function. We use the fact that now we do not refer
436 * to socket after mapping. If one day we will need it, this
437 * function will increment ref. count on file by 1.
439 * In any case returned fd MAY BE not valid!
440 * This race condition is unavoidable
441 * with shared fd spaces, we cannot solve it inside kernel,
442 * but we take care of internal coherence yet.
446 * sock_alloc_file - Bind a &socket to a &file
448 * @flags: file status flags
449 * @dname: protocol name
451 * Returns the &file bound with @sock, implicitly storing it
452 * in sock->file. If dname is %NULL, sets to "".
454 * On failure @sock is released, and an ERR pointer is returned.
456 * This function uses GFP_KERNEL internally.
459 struct file
*sock_alloc_file(struct socket
*sock
, int flags
, const char *dname
)
464 dname
= sock
->sk
? sock
->sk
->sk_prot_creator
->name
: "";
466 file
= alloc_file_pseudo(SOCK_INODE(sock
), sock_mnt
, dname
,
467 O_RDWR
| (flags
& O_NONBLOCK
),
475 file
->private_data
= sock
;
476 stream_open(SOCK_INODE(sock
), file
);
479 EXPORT_SYMBOL(sock_alloc_file
);
481 static int sock_map_fd(struct socket
*sock
, int flags
)
483 struct file
*newfile
;
484 int fd
= get_unused_fd_flags(flags
);
485 if (unlikely(fd
< 0)) {
490 newfile
= sock_alloc_file(sock
, flags
, NULL
);
491 if (!IS_ERR(newfile
)) {
492 fd_install(fd
, newfile
);
497 return PTR_ERR(newfile
);
501 * sock_from_file - Return the &socket bounded to @file.
504 * On failure returns %NULL.
507 struct socket
*sock_from_file(struct file
*file
)
509 if (file
->f_op
== &socket_file_ops
)
510 return file
->private_data
; /* set in sock_alloc_file */
514 EXPORT_SYMBOL(sock_from_file
);
517 * sockfd_lookup - Go from a file number to its socket slot
519 * @err: pointer to an error code return
521 * The file handle passed in is locked and the socket it is bound
522 * to is returned. If an error occurs the err pointer is overwritten
523 * with a negative errno code and NULL is returned. The function checks
524 * for both invalid handles and passing a handle which is not a socket.
526 * On a success the socket object pointer is returned.
529 struct socket
*sockfd_lookup(int fd
, int *err
)
540 sock
= sock_from_file(file
);
547 EXPORT_SYMBOL(sockfd_lookup
);
549 static struct socket
*sockfd_lookup_light(int fd
, int *err
, int *fput_needed
)
551 struct fd f
= fdget(fd
);
556 sock
= sock_from_file(f
.file
);
558 *fput_needed
= f
.flags
& FDPUT_FPUT
;
567 static ssize_t
sockfs_listxattr(struct dentry
*dentry
, char *buffer
,
573 len
= security_inode_listsecurity(d_inode(dentry
), buffer
, size
);
583 len
= (XATTR_NAME_SOCKPROTONAME_LEN
+ 1);
588 memcpy(buffer
, XATTR_NAME_SOCKPROTONAME
, len
);
595 static int sockfs_setattr(struct mnt_idmap
*idmap
,
596 struct dentry
*dentry
, struct iattr
*iattr
)
598 int err
= simple_setattr(&nop_mnt_idmap
, dentry
, iattr
);
600 if (!err
&& (iattr
->ia_valid
& ATTR_UID
)) {
601 struct socket
*sock
= SOCKET_I(d_inode(dentry
));
604 sock
->sk
->sk_uid
= iattr
->ia_uid
;
612 static const struct inode_operations sockfs_inode_ops
= {
613 .listxattr
= sockfs_listxattr
,
614 .setattr
= sockfs_setattr
,
618 * sock_alloc - allocate a socket
620 * Allocate a new inode and socket object. The two are bound together
621 * and initialised. The socket is then returned. If we are out of inodes
622 * NULL is returned. This functions uses GFP_KERNEL internally.
625 struct socket
*sock_alloc(void)
630 inode
= new_inode_pseudo(sock_mnt
->mnt_sb
);
634 sock
= SOCKET_I(inode
);
636 inode
->i_ino
= get_next_ino();
637 inode
->i_mode
= S_IFSOCK
| S_IRWXUGO
;
638 inode
->i_uid
= current_fsuid();
639 inode
->i_gid
= current_fsgid();
640 inode
->i_op
= &sockfs_inode_ops
;
644 EXPORT_SYMBOL(sock_alloc
);
646 static void __sock_release(struct socket
*sock
, struct inode
*inode
)
649 struct module
*owner
= sock
->ops
->owner
;
653 sock
->ops
->release(sock
);
661 if (sock
->wq
.fasync_list
)
662 pr_err("%s: fasync list not empty!\n", __func__
);
665 iput(SOCK_INODE(sock
));
672 * sock_release - close a socket
673 * @sock: socket to close
675 * The socket is released from the protocol stack if it has a release
676 * callback, and the inode is then released if the socket is bound to
677 * an inode not a file.
679 void sock_release(struct socket
*sock
)
681 __sock_release(sock
, NULL
);
683 EXPORT_SYMBOL(sock_release
);
685 void __sock_tx_timestamp(__u16 tsflags
, __u8
*tx_flags
)
687 u8 flags
= *tx_flags
;
689 if (tsflags
& SOF_TIMESTAMPING_TX_HARDWARE
) {
690 flags
|= SKBTX_HW_TSTAMP
;
692 /* PTP hardware clocks can provide a free running cycle counter
693 * as a time base for virtual clocks. Tell driver to use the
694 * free running cycle counter for timestamp if socket is bound
697 if (tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
698 flags
|= SKBTX_HW_TSTAMP_USE_CYCLES
;
701 if (tsflags
& SOF_TIMESTAMPING_TX_SOFTWARE
)
702 flags
|= SKBTX_SW_TSTAMP
;
704 if (tsflags
& SOF_TIMESTAMPING_TX_SCHED
)
705 flags
|= SKBTX_SCHED_TSTAMP
;
709 EXPORT_SYMBOL(__sock_tx_timestamp
);
711 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket
*, struct msghdr
*,
713 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket
*, struct msghdr
*,
716 static noinline
void call_trace_sock_send_length(struct sock
*sk
, int ret
,
719 trace_sock_send_length(sk
, ret
, 0);
722 static inline int sock_sendmsg_nosec(struct socket
*sock
, struct msghdr
*msg
)
724 int ret
= INDIRECT_CALL_INET(sock
->ops
->sendmsg
, inet6_sendmsg
,
725 inet_sendmsg
, sock
, msg
,
727 BUG_ON(ret
== -EIOCBQUEUED
);
729 if (trace_sock_send_length_enabled())
730 call_trace_sock_send_length(sock
->sk
, ret
, 0);
735 * sock_sendmsg - send a message through @sock
737 * @msg: message to send
739 * Sends @msg through @sock, passing through LSM.
740 * Returns the number of bytes sent, or an error code.
742 int sock_sendmsg(struct socket
*sock
, struct msghdr
*msg
)
744 int err
= security_socket_sendmsg(sock
, msg
,
747 return err
?: sock_sendmsg_nosec(sock
, msg
);
749 EXPORT_SYMBOL(sock_sendmsg
);
752 * kernel_sendmsg - send a message through @sock (kernel-space)
754 * @msg: message header
756 * @num: vec array length
757 * @size: total message data size
759 * Builds the message data with @vec and sends it through @sock.
760 * Returns the number of bytes sent, or an error code.
763 int kernel_sendmsg(struct socket
*sock
, struct msghdr
*msg
,
764 struct kvec
*vec
, size_t num
, size_t size
)
766 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
767 return sock_sendmsg(sock
, msg
);
769 EXPORT_SYMBOL(kernel_sendmsg
);
772 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
774 * @msg: message header
775 * @vec: output s/g array
776 * @num: output s/g array length
777 * @size: total message data size
779 * Builds the message data with @vec and sends it through @sock.
780 * Returns the number of bytes sent, or an error code.
781 * Caller must hold @sk.
784 int kernel_sendmsg_locked(struct sock
*sk
, struct msghdr
*msg
,
785 struct kvec
*vec
, size_t num
, size_t size
)
787 struct socket
*sock
= sk
->sk_socket
;
789 if (!sock
->ops
->sendmsg_locked
)
790 return sock_no_sendmsg_locked(sk
, msg
, size
);
792 iov_iter_kvec(&msg
->msg_iter
, ITER_SOURCE
, vec
, num
, size
);
794 return sock
->ops
->sendmsg_locked(sk
, msg
, msg_data_left(msg
));
796 EXPORT_SYMBOL(kernel_sendmsg_locked
);
798 static bool skb_is_err_queue(const struct sk_buff
*skb
)
800 /* pkt_type of skbs enqueued on the error queue are set to
801 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
802 * in recvmsg, since skbs received on a local socket will never
803 * have a pkt_type of PACKET_OUTGOING.
805 return skb
->pkt_type
== PACKET_OUTGOING
;
808 /* On transmit, software and hardware timestamps are returned independently.
809 * As the two skb clones share the hardware timestamp, which may be updated
810 * before the software timestamp is received, a hardware TX timestamp may be
811 * returned only if there is no software TX timestamp. Ignore false software
812 * timestamps, which may be made in the __sock_recv_timestamp() call when the
813 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
814 * hardware timestamp.
816 static bool skb_is_swtx_tstamp(const struct sk_buff
*skb
, int false_tstamp
)
818 return skb
->tstamp
&& !false_tstamp
&& skb_is_err_queue(skb
);
821 static ktime_t
get_timestamp(struct sock
*sk
, struct sk_buff
*skb
, int *if_index
)
823 bool cycles
= sk
->sk_tsflags
& SOF_TIMESTAMPING_BIND_PHC
;
824 struct skb_shared_hwtstamps
*shhwtstamps
= skb_hwtstamps(skb
);
825 struct net_device
*orig_dev
;
829 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
831 *if_index
= orig_dev
->ifindex
;
832 hwtstamp
= netdev_get_tstamp(orig_dev
, shhwtstamps
, cycles
);
834 hwtstamp
= shhwtstamps
->hwtstamp
;
841 static void put_ts_pktinfo(struct msghdr
*msg
, struct sk_buff
*skb
,
844 struct scm_ts_pktinfo ts_pktinfo
;
845 struct net_device
*orig_dev
;
847 if (!skb_mac_header_was_set(skb
))
850 memset(&ts_pktinfo
, 0, sizeof(ts_pktinfo
));
854 orig_dev
= dev_get_by_napi_id(skb_napi_id(skb
));
856 if_index
= orig_dev
->ifindex
;
859 ts_pktinfo
.if_index
= if_index
;
861 ts_pktinfo
.pkt_length
= skb
->len
- skb_mac_offset(skb
);
862 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_PKTINFO
,
863 sizeof(ts_pktinfo
), &ts_pktinfo
);
867 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
869 void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
872 int need_software_tstamp
= sock_flag(sk
, SOCK_RCVTSTAMP
);
873 int new_tstamp
= sock_flag(sk
, SOCK_TSTAMP_NEW
);
874 struct scm_timestamping_internal tss
;
876 int empty
= 1, false_tstamp
= 0;
877 struct skb_shared_hwtstamps
*shhwtstamps
=
882 /* Race occurred between timestamp enabling and packet
883 receiving. Fill in the current time for now. */
884 if (need_software_tstamp
&& skb
->tstamp
== 0) {
885 __net_timestamp(skb
);
889 if (need_software_tstamp
) {
890 if (!sock_flag(sk
, SOCK_RCVTSTAMPNS
)) {
892 struct __kernel_sock_timeval tv
;
894 skb_get_new_timestamp(skb
, &tv
);
895 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_NEW
,
898 struct __kernel_old_timeval tv
;
900 skb_get_timestamp(skb
, &tv
);
901 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMP_OLD
,
906 struct __kernel_timespec ts
;
908 skb_get_new_timestampns(skb
, &ts
);
909 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_NEW
,
912 struct __kernel_old_timespec ts
;
914 skb_get_timestampns(skb
, &ts
);
915 put_cmsg(msg
, SOL_SOCKET
, SO_TIMESTAMPNS_OLD
,
921 memset(&tss
, 0, sizeof(tss
));
922 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_SOFTWARE
) &&
923 ktime_to_timespec64_cond(skb
->tstamp
, tss
.ts
+ 0))
926 (sk
->sk_tsflags
& SOF_TIMESTAMPING_RAW_HARDWARE
) &&
927 !skb_is_swtx_tstamp(skb
, false_tstamp
)) {
929 if (skb_shinfo(skb
)->tx_flags
& SKBTX_HW_TSTAMP_NETDEV
)
930 hwtstamp
= get_timestamp(sk
, skb
, &if_index
);
932 hwtstamp
= shhwtstamps
->hwtstamp
;
934 if (sk
->sk_tsflags
& SOF_TIMESTAMPING_BIND_PHC
)
935 hwtstamp
= ptp_convert_timestamp(&hwtstamp
,
938 if (ktime_to_timespec64_cond(hwtstamp
, tss
.ts
+ 2)) {
941 if ((sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_PKTINFO
) &&
942 !skb_is_err_queue(skb
))
943 put_ts_pktinfo(msg
, skb
, if_index
);
947 if (sock_flag(sk
, SOCK_TSTAMP_NEW
))
948 put_cmsg_scm_timestamping64(msg
, &tss
);
950 put_cmsg_scm_timestamping(msg
, &tss
);
952 if (skb_is_err_queue(skb
) && skb
->len
&&
953 SKB_EXT_ERR(skb
)->opt_stats
)
954 put_cmsg(msg
, SOL_SOCKET
, SCM_TIMESTAMPING_OPT_STATS
,
955 skb
->len
, skb
->data
);
958 EXPORT_SYMBOL_GPL(__sock_recv_timestamp
);
960 #ifdef CONFIG_WIRELESS
961 void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
966 if (!sock_flag(sk
, SOCK_WIFI_STATUS
))
968 if (!skb
->wifi_acked_valid
)
971 ack
= skb
->wifi_acked
;
973 put_cmsg(msg
, SOL_SOCKET
, SCM_WIFI_STATUS
, sizeof(ack
), &ack
);
975 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status
);
978 static inline void sock_recv_drops(struct msghdr
*msg
, struct sock
*sk
,
981 if (sock_flag(sk
, SOCK_RXQ_OVFL
) && skb
&& SOCK_SKB_CB(skb
)->dropcount
)
982 put_cmsg(msg
, SOL_SOCKET
, SO_RXQ_OVFL
,
983 sizeof(__u32
), &SOCK_SKB_CB(skb
)->dropcount
);
986 static void sock_recv_mark(struct msghdr
*msg
, struct sock
*sk
,
989 if (sock_flag(sk
, SOCK_RCVMARK
) && skb
) {
990 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
991 __u32 mark
= skb
->mark
;
993 put_cmsg(msg
, SOL_SOCKET
, SO_MARK
, sizeof(__u32
), &mark
);
997 void __sock_recv_cmsgs(struct msghdr
*msg
, struct sock
*sk
,
1000 sock_recv_timestamp(msg
, sk
, skb
);
1001 sock_recv_drops(msg
, sk
, skb
);
1002 sock_recv_mark(msg
, sk
, skb
);
1004 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs
);
1006 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket
*, struct msghdr
*,
1008 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket
*, struct msghdr
*,
1011 static noinline
void call_trace_sock_recv_length(struct sock
*sk
, int ret
, int flags
)
1013 trace_sock_recv_length(sk
, ret
, flags
);
1016 static inline int sock_recvmsg_nosec(struct socket
*sock
, struct msghdr
*msg
,
1019 int ret
= INDIRECT_CALL_INET(sock
->ops
->recvmsg
, inet6_recvmsg
,
1020 inet_recvmsg
, sock
, msg
,
1021 msg_data_left(msg
), flags
);
1022 if (trace_sock_recv_length_enabled())
1023 call_trace_sock_recv_length(sock
->sk
, ret
, flags
);
1028 * sock_recvmsg - receive a message from @sock
1030 * @msg: message to receive
1031 * @flags: message flags
1033 * Receives @msg from @sock, passing through LSM. Returns the total number
1034 * of bytes received, or an error.
1036 int sock_recvmsg(struct socket
*sock
, struct msghdr
*msg
, int flags
)
1038 int err
= security_socket_recvmsg(sock
, msg
, msg_data_left(msg
), flags
);
1040 return err
?: sock_recvmsg_nosec(sock
, msg
, flags
);
1042 EXPORT_SYMBOL(sock_recvmsg
);
1045 * kernel_recvmsg - Receive a message from a socket (kernel space)
1046 * @sock: The socket to receive the message from
1047 * @msg: Received message
1048 * @vec: Input s/g array for message data
1049 * @num: Size of input s/g array
1050 * @size: Number of bytes to read
1051 * @flags: Message flags (MSG_DONTWAIT, etc...)
1053 * On return the msg structure contains the scatter/gather array passed in the
1054 * vec argument. The array is modified so that it consists of the unfilled
1055 * portion of the original array.
1057 * The returned value is the total number of bytes received, or an error.
1060 int kernel_recvmsg(struct socket
*sock
, struct msghdr
*msg
,
1061 struct kvec
*vec
, size_t num
, size_t size
, int flags
)
1063 msg
->msg_control_is_user
= false;
1064 iov_iter_kvec(&msg
->msg_iter
, ITER_DEST
, vec
, num
, size
);
1065 return sock_recvmsg(sock
, msg
, flags
);
1067 EXPORT_SYMBOL(kernel_recvmsg
);
1069 static ssize_t
sock_sendpage(struct file
*file
, struct page
*page
,
1070 int offset
, size_t size
, loff_t
*ppos
, int more
)
1072 struct socket
*sock
;
1076 sock
= file
->private_data
;
1078 flags
= (file
->f_flags
& O_NONBLOCK
) ? MSG_DONTWAIT
: 0;
1079 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1082 ret
= kernel_sendpage(sock
, page
, offset
, size
, flags
);
1084 if (trace_sock_send_length_enabled())
1085 call_trace_sock_send_length(sock
->sk
, ret
, 0);
1089 static ssize_t
sock_splice_read(struct file
*file
, loff_t
*ppos
,
1090 struct pipe_inode_info
*pipe
, size_t len
,
1093 struct socket
*sock
= file
->private_data
;
1095 if (unlikely(!sock
->ops
->splice_read
))
1096 return generic_file_splice_read(file
, ppos
, pipe
, len
, flags
);
1098 return sock
->ops
->splice_read(sock
, ppos
, pipe
, len
, flags
);
1101 static ssize_t
sock_read_iter(struct kiocb
*iocb
, struct iov_iter
*to
)
1103 struct file
*file
= iocb
->ki_filp
;
1104 struct socket
*sock
= file
->private_data
;
1105 struct msghdr msg
= {.msg_iter
= *to
,
1109 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1110 msg
.msg_flags
= MSG_DONTWAIT
;
1112 if (iocb
->ki_pos
!= 0)
1115 if (!iov_iter_count(to
)) /* Match SYS5 behaviour */
1118 res
= sock_recvmsg(sock
, &msg
, msg
.msg_flags
);
1123 static ssize_t
sock_write_iter(struct kiocb
*iocb
, struct iov_iter
*from
)
1125 struct file
*file
= iocb
->ki_filp
;
1126 struct socket
*sock
= file
->private_data
;
1127 struct msghdr msg
= {.msg_iter
= *from
,
1131 if (iocb
->ki_pos
!= 0)
1134 if (file
->f_flags
& O_NONBLOCK
|| (iocb
->ki_flags
& IOCB_NOWAIT
))
1135 msg
.msg_flags
= MSG_DONTWAIT
;
1137 if (sock
->type
== SOCK_SEQPACKET
)
1138 msg
.msg_flags
|= MSG_EOR
;
1140 res
= sock_sendmsg(sock
, &msg
);
1141 *from
= msg
.msg_iter
;
1146 * Atomic setting of ioctl hooks to avoid race
1147 * with module unload.
1150 static DEFINE_MUTEX(br_ioctl_mutex
);
1151 static int (*br_ioctl_hook
)(struct net
*net
, struct net_bridge
*br
,
1152 unsigned int cmd
, struct ifreq
*ifr
,
1155 void brioctl_set(int (*hook
)(struct net
*net
, struct net_bridge
*br
,
1156 unsigned int cmd
, struct ifreq
*ifr
,
1159 mutex_lock(&br_ioctl_mutex
);
1160 br_ioctl_hook
= hook
;
1161 mutex_unlock(&br_ioctl_mutex
);
1163 EXPORT_SYMBOL(brioctl_set
);
1165 int br_ioctl_call(struct net
*net
, struct net_bridge
*br
, unsigned int cmd
,
1166 struct ifreq
*ifr
, void __user
*uarg
)
1171 request_module("bridge");
1173 mutex_lock(&br_ioctl_mutex
);
1175 err
= br_ioctl_hook(net
, br
, cmd
, ifr
, uarg
);
1176 mutex_unlock(&br_ioctl_mutex
);
1181 static DEFINE_MUTEX(vlan_ioctl_mutex
);
1182 static int (*vlan_ioctl_hook
) (struct net
*, void __user
*arg
);
1184 void vlan_ioctl_set(int (*hook
) (struct net
*, void __user
*))
1186 mutex_lock(&vlan_ioctl_mutex
);
1187 vlan_ioctl_hook
= hook
;
1188 mutex_unlock(&vlan_ioctl_mutex
);
1190 EXPORT_SYMBOL(vlan_ioctl_set
);
1192 static long sock_do_ioctl(struct net
*net
, struct socket
*sock
,
1193 unsigned int cmd
, unsigned long arg
)
1198 void __user
*argp
= (void __user
*)arg
;
1201 err
= sock
->ops
->ioctl(sock
, cmd
, arg
);
1204 * If this ioctl is unknown try to hand it down
1205 * to the NIC driver.
1207 if (err
!= -ENOIOCTLCMD
)
1210 if (!is_socket_ioctl_cmd(cmd
))
1213 if (get_user_ifreq(&ifr
, &data
, argp
))
1215 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1216 if (!err
&& need_copyout
)
1217 if (put_user_ifreq(&ifr
, argp
))
1224 * With an ioctl, arg may well be a user mode pointer, but we don't know
1225 * what to do with it - that's up to the protocol still.
1228 static long sock_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
1230 struct socket
*sock
;
1232 void __user
*argp
= (void __user
*)arg
;
1236 sock
= file
->private_data
;
1239 if (unlikely(cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))) {
1243 if (get_user_ifreq(&ifr
, &data
, argp
))
1245 err
= dev_ioctl(net
, cmd
, &ifr
, data
, &need_copyout
);
1246 if (!err
&& need_copyout
)
1247 if (put_user_ifreq(&ifr
, argp
))
1250 #ifdef CONFIG_WEXT_CORE
1251 if (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
) {
1252 err
= wext_handle_ioctl(net
, cmd
, argp
);
1259 if (get_user(pid
, (int __user
*)argp
))
1261 err
= f_setown(sock
->file
, pid
, 1);
1265 err
= put_user(f_getown(sock
->file
),
1266 (int __user
*)argp
);
1272 err
= br_ioctl_call(net
, NULL
, cmd
, NULL
, argp
);
1277 if (!vlan_ioctl_hook
)
1278 request_module("8021q");
1280 mutex_lock(&vlan_ioctl_mutex
);
1281 if (vlan_ioctl_hook
)
1282 err
= vlan_ioctl_hook(net
, argp
);
1283 mutex_unlock(&vlan_ioctl_mutex
);
1287 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
1290 err
= open_related_ns(&net
->ns
, get_net_ns
);
1292 case SIOCGSTAMP_OLD
:
1293 case SIOCGSTAMPNS_OLD
:
1294 if (!sock
->ops
->gettstamp
) {
1298 err
= sock
->ops
->gettstamp(sock
, argp
,
1299 cmd
== SIOCGSTAMP_OLD
,
1300 !IS_ENABLED(CONFIG_64BIT
));
1302 case SIOCGSTAMP_NEW
:
1303 case SIOCGSTAMPNS_NEW
:
1304 if (!sock
->ops
->gettstamp
) {
1308 err
= sock
->ops
->gettstamp(sock
, argp
,
1309 cmd
== SIOCGSTAMP_NEW
,
1314 err
= dev_ifconf(net
, argp
);
1318 err
= sock_do_ioctl(net
, sock
, cmd
, arg
);
1325 * sock_create_lite - creates a socket
1326 * @family: protocol family (AF_INET, ...)
1327 * @type: communication type (SOCK_STREAM, ...)
1328 * @protocol: protocol (0, ...)
1331 * Creates a new socket and assigns it to @res, passing through LSM.
1332 * The new socket initialization is not complete, see kernel_accept().
1333 * Returns 0 or an error. On failure @res is set to %NULL.
1334 * This function internally uses GFP_KERNEL.
1337 int sock_create_lite(int family
, int type
, int protocol
, struct socket
**res
)
1340 struct socket
*sock
= NULL
;
1342 err
= security_socket_create(family
, type
, protocol
, 1);
1346 sock
= sock_alloc();
1353 err
= security_socket_post_create(sock
, family
, type
, protocol
, 1);
1365 EXPORT_SYMBOL(sock_create_lite
);
1367 /* No kernel lock held - perfect */
1368 static __poll_t
sock_poll(struct file
*file
, poll_table
*wait
)
1370 struct socket
*sock
= file
->private_data
;
1371 __poll_t events
= poll_requested_events(wait
), flag
= 0;
1373 if (!sock
->ops
->poll
)
1376 if (sk_can_busy_loop(sock
->sk
)) {
1377 /* poll once if requested by the syscall */
1378 if (events
& POLL_BUSY_LOOP
)
1379 sk_busy_loop(sock
->sk
, 1);
1381 /* if this socket can poll_ll, tell the system call */
1382 flag
= POLL_BUSY_LOOP
;
1385 return sock
->ops
->poll(file
, sock
, wait
) | flag
;
1388 static int sock_mmap(struct file
*file
, struct vm_area_struct
*vma
)
1390 struct socket
*sock
= file
->private_data
;
1392 return sock
->ops
->mmap(file
, sock
, vma
);
1395 static int sock_close(struct inode
*inode
, struct file
*filp
)
1397 __sock_release(SOCKET_I(inode
), inode
);
1402 * Update the socket async list
1404 * Fasync_list locking strategy.
1406 * 1. fasync_list is modified only under process context socket lock
1407 * i.e. under semaphore.
1408 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1409 * or under socket lock
1412 static int sock_fasync(int fd
, struct file
*filp
, int on
)
1414 struct socket
*sock
= filp
->private_data
;
1415 struct sock
*sk
= sock
->sk
;
1416 struct socket_wq
*wq
= &sock
->wq
;
1422 fasync_helper(fd
, filp
, on
, &wq
->fasync_list
);
1424 if (!wq
->fasync_list
)
1425 sock_reset_flag(sk
, SOCK_FASYNC
);
1427 sock_set_flag(sk
, SOCK_FASYNC
);
1433 /* This function may be called only under rcu_lock */
1435 int sock_wake_async(struct socket_wq
*wq
, int how
, int band
)
1437 if (!wq
|| !wq
->fasync_list
)
1441 case SOCK_WAKE_WAITD
:
1442 if (test_bit(SOCKWQ_ASYNC_WAITDATA
, &wq
->flags
))
1445 case SOCK_WAKE_SPACE
:
1446 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE
, &wq
->flags
))
1451 kill_fasync(&wq
->fasync_list
, SIGIO
, band
);
1454 kill_fasync(&wq
->fasync_list
, SIGURG
, band
);
1459 EXPORT_SYMBOL(sock_wake_async
);
1462 * __sock_create - creates a socket
1463 * @net: net namespace
1464 * @family: protocol family (AF_INET, ...)
1465 * @type: communication type (SOCK_STREAM, ...)
1466 * @protocol: protocol (0, ...)
1468 * @kern: boolean for kernel space sockets
1470 * Creates a new socket and assigns it to @res, passing through LSM.
1471 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1472 * be set to true if the socket resides in kernel space.
1473 * This function internally uses GFP_KERNEL.
1476 int __sock_create(struct net
*net
, int family
, int type
, int protocol
,
1477 struct socket
**res
, int kern
)
1480 struct socket
*sock
;
1481 const struct net_proto_family
*pf
;
1484 * Check protocol is in range
1486 if (family
< 0 || family
>= NPROTO
)
1487 return -EAFNOSUPPORT
;
1488 if (type
< 0 || type
>= SOCK_MAX
)
1493 This uglymoron is moved from INET layer to here to avoid
1494 deadlock in module load.
1496 if (family
== PF_INET
&& type
== SOCK_PACKET
) {
1497 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1502 err
= security_socket_create(family
, type
, protocol
, kern
);
1507 * Allocate the socket and allow the family to set things up. if
1508 * the protocol is 0, the family is instructed to select an appropriate
1511 sock
= sock_alloc();
1513 net_warn_ratelimited("socket: no more sockets\n");
1514 return -ENFILE
; /* Not exactly a match, but its the
1515 closest posix thing */
1520 #ifdef CONFIG_MODULES
1521 /* Attempt to load a protocol module if the find failed.
1523 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1524 * requested real, full-featured networking support upon configuration.
1525 * Otherwise module support will break!
1527 if (rcu_access_pointer(net_families
[family
]) == NULL
)
1528 request_module("net-pf-%d", family
);
1532 pf
= rcu_dereference(net_families
[family
]);
1533 err
= -EAFNOSUPPORT
;
1538 * We will call the ->create function, that possibly is in a loadable
1539 * module, so we have to bump that loadable module refcnt first.
1541 if (!try_module_get(pf
->owner
))
1544 /* Now protected by module ref count */
1547 err
= pf
->create(net
, sock
, protocol
, kern
);
1549 goto out_module_put
;
1552 * Now to bump the refcnt of the [loadable] module that owns this
1553 * socket at sock_release time we decrement its refcnt.
1555 if (!try_module_get(sock
->ops
->owner
))
1556 goto out_module_busy
;
1559 * Now that we're done with the ->create function, the [loadable]
1560 * module can have its refcnt decremented
1562 module_put(pf
->owner
);
1563 err
= security_socket_post_create(sock
, family
, type
, protocol
, kern
);
1565 goto out_sock_release
;
1571 err
= -EAFNOSUPPORT
;
1574 module_put(pf
->owner
);
1581 goto out_sock_release
;
1583 EXPORT_SYMBOL(__sock_create
);
1586 * sock_create - creates a socket
1587 * @family: protocol family (AF_INET, ...)
1588 * @type: communication type (SOCK_STREAM, ...)
1589 * @protocol: protocol (0, ...)
1592 * A wrapper around __sock_create().
1593 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1596 int sock_create(int family
, int type
, int protocol
, struct socket
**res
)
1598 return __sock_create(current
->nsproxy
->net_ns
, family
, type
, protocol
, res
, 0);
1600 EXPORT_SYMBOL(sock_create
);
1603 * sock_create_kern - creates a socket (kernel space)
1604 * @net: net namespace
1605 * @family: protocol family (AF_INET, ...)
1606 * @type: communication type (SOCK_STREAM, ...)
1607 * @protocol: protocol (0, ...)
1610 * A wrapper around __sock_create().
1611 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1614 int sock_create_kern(struct net
*net
, int family
, int type
, int protocol
, struct socket
**res
)
1616 return __sock_create(net
, family
, type
, protocol
, res
, 1);
1618 EXPORT_SYMBOL(sock_create_kern
);
1620 static struct socket
*__sys_socket_create(int family
, int type
, int protocol
)
1622 struct socket
*sock
;
1625 /* Check the SOCK_* constants for consistency. */
1626 BUILD_BUG_ON(SOCK_CLOEXEC
!= O_CLOEXEC
);
1627 BUILD_BUG_ON((SOCK_MAX
| SOCK_TYPE_MASK
) != SOCK_TYPE_MASK
);
1628 BUILD_BUG_ON(SOCK_CLOEXEC
& SOCK_TYPE_MASK
);
1629 BUILD_BUG_ON(SOCK_NONBLOCK
& SOCK_TYPE_MASK
);
1631 if ((type
& ~SOCK_TYPE_MASK
) & ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1632 return ERR_PTR(-EINVAL
);
1633 type
&= SOCK_TYPE_MASK
;
1635 retval
= sock_create(family
, type
, protocol
, &sock
);
1637 return ERR_PTR(retval
);
1642 struct file
*__sys_socket_file(int family
, int type
, int protocol
)
1644 struct socket
*sock
;
1647 sock
= __sys_socket_create(family
, type
, protocol
);
1649 return ERR_CAST(sock
);
1651 flags
= type
& ~SOCK_TYPE_MASK
;
1652 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1653 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1655 return sock_alloc_file(sock
, flags
, NULL
);
1658 int __sys_socket(int family
, int type
, int protocol
)
1660 struct socket
*sock
;
1663 sock
= __sys_socket_create(family
, type
, protocol
);
1665 return PTR_ERR(sock
);
1667 flags
= type
& ~SOCK_TYPE_MASK
;
1668 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1669 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1671 return sock_map_fd(sock
, flags
& (O_CLOEXEC
| O_NONBLOCK
));
1674 SYSCALL_DEFINE3(socket
, int, family
, int, type
, int, protocol
)
1676 return __sys_socket(family
, type
, protocol
);
1680 * Create a pair of connected sockets.
1683 int __sys_socketpair(int family
, int type
, int protocol
, int __user
*usockvec
)
1685 struct socket
*sock1
, *sock2
;
1687 struct file
*newfile1
, *newfile2
;
1690 flags
= type
& ~SOCK_TYPE_MASK
;
1691 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1693 type
&= SOCK_TYPE_MASK
;
1695 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1696 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1699 * reserve descriptors and make sure we won't fail
1700 * to return them to userland.
1702 fd1
= get_unused_fd_flags(flags
);
1703 if (unlikely(fd1
< 0))
1706 fd2
= get_unused_fd_flags(flags
);
1707 if (unlikely(fd2
< 0)) {
1712 err
= put_user(fd1
, &usockvec
[0]);
1716 err
= put_user(fd2
, &usockvec
[1]);
1721 * Obtain the first socket and check if the underlying protocol
1722 * supports the socketpair call.
1725 err
= sock_create(family
, type
, protocol
, &sock1
);
1726 if (unlikely(err
< 0))
1729 err
= sock_create(family
, type
, protocol
, &sock2
);
1730 if (unlikely(err
< 0)) {
1731 sock_release(sock1
);
1735 err
= security_socket_socketpair(sock1
, sock2
);
1736 if (unlikely(err
)) {
1737 sock_release(sock2
);
1738 sock_release(sock1
);
1742 err
= sock1
->ops
->socketpair(sock1
, sock2
);
1743 if (unlikely(err
< 0)) {
1744 sock_release(sock2
);
1745 sock_release(sock1
);
1749 newfile1
= sock_alloc_file(sock1
, flags
, NULL
);
1750 if (IS_ERR(newfile1
)) {
1751 err
= PTR_ERR(newfile1
);
1752 sock_release(sock2
);
1756 newfile2
= sock_alloc_file(sock2
, flags
, NULL
);
1757 if (IS_ERR(newfile2
)) {
1758 err
= PTR_ERR(newfile2
);
1763 audit_fd_pair(fd1
, fd2
);
1765 fd_install(fd1
, newfile1
);
1766 fd_install(fd2
, newfile2
);
1775 SYSCALL_DEFINE4(socketpair
, int, family
, int, type
, int, protocol
,
1776 int __user
*, usockvec
)
1778 return __sys_socketpair(family
, type
, protocol
, usockvec
);
1782 * Bind a name to a socket. Nothing much to do here since it's
1783 * the protocol's responsibility to handle the local address.
1785 * We move the socket address to kernel space before we call
1786 * the protocol layer (having also checked the address is ok).
1789 int __sys_bind(int fd
, struct sockaddr __user
*umyaddr
, int addrlen
)
1791 struct socket
*sock
;
1792 struct sockaddr_storage address
;
1793 int err
, fput_needed
;
1795 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1797 err
= move_addr_to_kernel(umyaddr
, addrlen
, &address
);
1799 err
= security_socket_bind(sock
,
1800 (struct sockaddr
*)&address
,
1803 err
= sock
->ops
->bind(sock
,
1807 fput_light(sock
->file
, fput_needed
);
1812 SYSCALL_DEFINE3(bind
, int, fd
, struct sockaddr __user
*, umyaddr
, int, addrlen
)
1814 return __sys_bind(fd
, umyaddr
, addrlen
);
1818 * Perform a listen. Basically, we allow the protocol to do anything
1819 * necessary for a listen, and if that works, we mark the socket as
1820 * ready for listening.
1823 int __sys_listen(int fd
, int backlog
)
1825 struct socket
*sock
;
1826 int err
, fput_needed
;
1829 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
1831 somaxconn
= READ_ONCE(sock_net(sock
->sk
)->core
.sysctl_somaxconn
);
1832 if ((unsigned int)backlog
> somaxconn
)
1833 backlog
= somaxconn
;
1835 err
= security_socket_listen(sock
, backlog
);
1837 err
= sock
->ops
->listen(sock
, backlog
);
1839 fput_light(sock
->file
, fput_needed
);
1844 SYSCALL_DEFINE2(listen
, int, fd
, int, backlog
)
1846 return __sys_listen(fd
, backlog
);
1849 struct file
*do_accept(struct file
*file
, unsigned file_flags
,
1850 struct sockaddr __user
*upeer_sockaddr
,
1851 int __user
*upeer_addrlen
, int flags
)
1853 struct socket
*sock
, *newsock
;
1854 struct file
*newfile
;
1856 struct sockaddr_storage address
;
1858 sock
= sock_from_file(file
);
1860 return ERR_PTR(-ENOTSOCK
);
1862 newsock
= sock_alloc();
1864 return ERR_PTR(-ENFILE
);
1866 newsock
->type
= sock
->type
;
1867 newsock
->ops
= sock
->ops
;
1870 * We don't need try_module_get here, as the listening socket (sock)
1871 * has the protocol module (sock->ops->owner) held.
1873 __module_get(newsock
->ops
->owner
);
1875 newfile
= sock_alloc_file(newsock
, flags
, sock
->sk
->sk_prot_creator
->name
);
1876 if (IS_ERR(newfile
))
1879 err
= security_socket_accept(sock
, newsock
);
1883 err
= sock
->ops
->accept(sock
, newsock
, sock
->file
->f_flags
| file_flags
,
1888 if (upeer_sockaddr
) {
1889 len
= newsock
->ops
->getname(newsock
,
1890 (struct sockaddr
*)&address
, 2);
1892 err
= -ECONNABORTED
;
1895 err
= move_addr_to_user(&address
,
1896 len
, upeer_sockaddr
, upeer_addrlen
);
1901 /* File flags are not inherited via accept() unlike another OSes. */
1905 return ERR_PTR(err
);
1908 static int __sys_accept4_file(struct file
*file
, struct sockaddr __user
*upeer_sockaddr
,
1909 int __user
*upeer_addrlen
, int flags
)
1911 struct file
*newfile
;
1914 if (flags
& ~(SOCK_CLOEXEC
| SOCK_NONBLOCK
))
1917 if (SOCK_NONBLOCK
!= O_NONBLOCK
&& (flags
& SOCK_NONBLOCK
))
1918 flags
= (flags
& ~SOCK_NONBLOCK
) | O_NONBLOCK
;
1920 newfd
= get_unused_fd_flags(flags
);
1921 if (unlikely(newfd
< 0))
1924 newfile
= do_accept(file
, 0, upeer_sockaddr
, upeer_addrlen
,
1926 if (IS_ERR(newfile
)) {
1927 put_unused_fd(newfd
);
1928 return PTR_ERR(newfile
);
1930 fd_install(newfd
, newfile
);
1935 * For accept, we attempt to create a new socket, set up the link
1936 * with the client, wake up the client, then return the new
1937 * connected fd. We collect the address of the connector in kernel
1938 * space and move it to user at the very end. This is unclean because
1939 * we open the socket then return an error.
1941 * 1003.1g adds the ability to recvmsg() to query connection pending
1942 * status to recvmsg. We need to add that support in a way thats
1943 * clean when we restructure accept also.
1946 int __sys_accept4(int fd
, struct sockaddr __user
*upeer_sockaddr
,
1947 int __user
*upeer_addrlen
, int flags
)
1954 ret
= __sys_accept4_file(f
.file
, upeer_sockaddr
,
1955 upeer_addrlen
, flags
);
1962 SYSCALL_DEFINE4(accept4
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1963 int __user
*, upeer_addrlen
, int, flags
)
1965 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, flags
);
1968 SYSCALL_DEFINE3(accept
, int, fd
, struct sockaddr __user
*, upeer_sockaddr
,
1969 int __user
*, upeer_addrlen
)
1971 return __sys_accept4(fd
, upeer_sockaddr
, upeer_addrlen
, 0);
1975 * Attempt to connect to a socket with the server address. The address
1976 * is in user space so we verify it is OK and move it to kernel space.
1978 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1981 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1982 * other SEQPACKET protocols that take time to connect() as it doesn't
1983 * include the -EINPROGRESS status for such sockets.
1986 int __sys_connect_file(struct file
*file
, struct sockaddr_storage
*address
,
1987 int addrlen
, int file_flags
)
1989 struct socket
*sock
;
1992 sock
= sock_from_file(file
);
1999 security_socket_connect(sock
, (struct sockaddr
*)address
, addrlen
);
2003 err
= sock
->ops
->connect(sock
, (struct sockaddr
*)address
, addrlen
,
2004 sock
->file
->f_flags
| file_flags
);
2009 int __sys_connect(int fd
, struct sockaddr __user
*uservaddr
, int addrlen
)
2016 struct sockaddr_storage address
;
2018 ret
= move_addr_to_kernel(uservaddr
, addrlen
, &address
);
2020 ret
= __sys_connect_file(f
.file
, &address
, addrlen
, 0);
2027 SYSCALL_DEFINE3(connect
, int, fd
, struct sockaddr __user
*, uservaddr
,
2030 return __sys_connect(fd
, uservaddr
, addrlen
);
2034 * Get the local address ('name') of a socket object. Move the obtained
2035 * name to user space.
2038 int __sys_getsockname(int fd
, struct sockaddr __user
*usockaddr
,
2039 int __user
*usockaddr_len
)
2041 struct socket
*sock
;
2042 struct sockaddr_storage address
;
2043 int err
, fput_needed
;
2045 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2049 err
= security_socket_getsockname(sock
);
2053 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 0);
2056 /* "err" is actually length in this case */
2057 err
= move_addr_to_user(&address
, err
, usockaddr
, usockaddr_len
);
2060 fput_light(sock
->file
, fput_needed
);
2065 SYSCALL_DEFINE3(getsockname
, int, fd
, struct sockaddr __user
*, usockaddr
,
2066 int __user
*, usockaddr_len
)
2068 return __sys_getsockname(fd
, usockaddr
, usockaddr_len
);
2072 * Get the remote address ('name') of a socket object. Move the obtained
2073 * name to user space.
2076 int __sys_getpeername(int fd
, struct sockaddr __user
*usockaddr
,
2077 int __user
*usockaddr_len
)
2079 struct socket
*sock
;
2080 struct sockaddr_storage address
;
2081 int err
, fput_needed
;
2083 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2085 err
= security_socket_getpeername(sock
);
2087 fput_light(sock
->file
, fput_needed
);
2091 err
= sock
->ops
->getname(sock
, (struct sockaddr
*)&address
, 1);
2093 /* "err" is actually length in this case */
2094 err
= move_addr_to_user(&address
, err
, usockaddr
,
2096 fput_light(sock
->file
, fput_needed
);
2101 SYSCALL_DEFINE3(getpeername
, int, fd
, struct sockaddr __user
*, usockaddr
,
2102 int __user
*, usockaddr_len
)
2104 return __sys_getpeername(fd
, usockaddr
, usockaddr_len
);
2108 * Send a datagram to a given address. We move the address into kernel
2109 * space and check the user space data area is readable before invoking
2112 int __sys_sendto(int fd
, void __user
*buff
, size_t len
, unsigned int flags
,
2113 struct sockaddr __user
*addr
, int addr_len
)
2115 struct socket
*sock
;
2116 struct sockaddr_storage address
;
2122 err
= import_single_range(ITER_SOURCE
, buff
, len
, &iov
, &msg
.msg_iter
);
2125 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2129 msg
.msg_name
= NULL
;
2130 msg
.msg_control
= NULL
;
2131 msg
.msg_controllen
= 0;
2132 msg
.msg_namelen
= 0;
2133 msg
.msg_ubuf
= NULL
;
2135 err
= move_addr_to_kernel(addr
, addr_len
, &address
);
2138 msg
.msg_name
= (struct sockaddr
*)&address
;
2139 msg
.msg_namelen
= addr_len
;
2141 if (sock
->file
->f_flags
& O_NONBLOCK
)
2142 flags
|= MSG_DONTWAIT
;
2143 msg
.msg_flags
= flags
;
2144 err
= sock_sendmsg(sock
, &msg
);
2147 fput_light(sock
->file
, fput_needed
);
2152 SYSCALL_DEFINE6(sendto
, int, fd
, void __user
*, buff
, size_t, len
,
2153 unsigned int, flags
, struct sockaddr __user
*, addr
,
2156 return __sys_sendto(fd
, buff
, len
, flags
, addr
, addr_len
);
2160 * Send a datagram down a socket.
2163 SYSCALL_DEFINE4(send
, int, fd
, void __user
*, buff
, size_t, len
,
2164 unsigned int, flags
)
2166 return __sys_sendto(fd
, buff
, len
, flags
, NULL
, 0);
2170 * Receive a frame from the socket and optionally record the address of the
2171 * sender. We verify the buffers are writable and if needed move the
2172 * sender address from kernel to user space.
2174 int __sys_recvfrom(int fd
, void __user
*ubuf
, size_t size
, unsigned int flags
,
2175 struct sockaddr __user
*addr
, int __user
*addr_len
)
2177 struct sockaddr_storage address
;
2178 struct msghdr msg
= {
2179 /* Save some cycles and don't copy the address if not needed */
2180 .msg_name
= addr
? (struct sockaddr
*)&address
: NULL
,
2182 struct socket
*sock
;
2187 err
= import_single_range(ITER_DEST
, ubuf
, size
, &iov
, &msg
.msg_iter
);
2190 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2194 if (sock
->file
->f_flags
& O_NONBLOCK
)
2195 flags
|= MSG_DONTWAIT
;
2196 err
= sock_recvmsg(sock
, &msg
, flags
);
2198 if (err
>= 0 && addr
!= NULL
) {
2199 err2
= move_addr_to_user(&address
,
2200 msg
.msg_namelen
, addr
, addr_len
);
2205 fput_light(sock
->file
, fput_needed
);
2210 SYSCALL_DEFINE6(recvfrom
, int, fd
, void __user
*, ubuf
, size_t, size
,
2211 unsigned int, flags
, struct sockaddr __user
*, addr
,
2212 int __user
*, addr_len
)
2214 return __sys_recvfrom(fd
, ubuf
, size
, flags
, addr
, addr_len
);
2218 * Receive a datagram from a socket.
2221 SYSCALL_DEFINE4(recv
, int, fd
, void __user
*, ubuf
, size_t, size
,
2222 unsigned int, flags
)
2224 return __sys_recvfrom(fd
, ubuf
, size
, flags
, NULL
, NULL
);
2227 static bool sock_use_custom_sol_socket(const struct socket
*sock
)
2229 return test_bit(SOCK_CUSTOM_SOCKOPT
, &sock
->flags
);
2233 * Set a socket option. Because we don't know the option lengths we have
2234 * to pass the user mode parameter for the protocols to sort out.
2236 int __sys_setsockopt(int fd
, int level
, int optname
, char __user
*user_optval
,
2239 sockptr_t optval
= USER_SOCKPTR(user_optval
);
2240 char *kernel_optval
= NULL
;
2241 int err
, fput_needed
;
2242 struct socket
*sock
;
2247 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2251 err
= security_socket_setsockopt(sock
, level
, optname
);
2255 if (!in_compat_syscall())
2256 err
= BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock
->sk
, &level
, &optname
,
2257 user_optval
, &optlen
,
2267 optval
= KERNEL_SOCKPTR(kernel_optval
);
2268 if (level
== SOL_SOCKET
&& !sock_use_custom_sol_socket(sock
))
2269 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
2270 else if (unlikely(!sock
->ops
->setsockopt
))
2273 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
2275 kfree(kernel_optval
);
2277 fput_light(sock
->file
, fput_needed
);
2281 SYSCALL_DEFINE5(setsockopt
, int, fd
, int, level
, int, optname
,
2282 char __user
*, optval
, int, optlen
)
2284 return __sys_setsockopt(fd
, level
, optname
, optval
, optlen
);
2287 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level
,
2291 * Get a socket option. Because we don't know the option lengths we have
2292 * to pass a user mode parameter for the protocols to sort out.
2294 int __sys_getsockopt(int fd
, int level
, int optname
, char __user
*optval
,
2297 int max_optlen __maybe_unused
;
2298 int err
, fput_needed
;
2299 struct socket
*sock
;
2301 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2305 err
= security_socket_getsockopt(sock
, level
, optname
);
2309 if (!in_compat_syscall())
2310 max_optlen
= BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen
);
2312 if (level
== SOL_SOCKET
)
2313 err
= sock_getsockopt(sock
, level
, optname
, optval
, optlen
);
2314 else if (unlikely(!sock
->ops
->getsockopt
))
2317 err
= sock
->ops
->getsockopt(sock
, level
, optname
, optval
,
2320 if (!in_compat_syscall())
2321 err
= BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock
->sk
, level
, optname
,
2322 optval
, optlen
, max_optlen
,
2325 fput_light(sock
->file
, fput_needed
);
2329 SYSCALL_DEFINE5(getsockopt
, int, fd
, int, level
, int, optname
,
2330 char __user
*, optval
, int __user
*, optlen
)
2332 return __sys_getsockopt(fd
, level
, optname
, optval
, optlen
);
2336 * Shutdown a socket.
2339 int __sys_shutdown_sock(struct socket
*sock
, int how
)
2343 err
= security_socket_shutdown(sock
, how
);
2345 err
= sock
->ops
->shutdown(sock
, how
);
2350 int __sys_shutdown(int fd
, int how
)
2352 int err
, fput_needed
;
2353 struct socket
*sock
;
2355 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2357 err
= __sys_shutdown_sock(sock
, how
);
2358 fput_light(sock
->file
, fput_needed
);
2363 SYSCALL_DEFINE2(shutdown
, int, fd
, int, how
)
2365 return __sys_shutdown(fd
, how
);
2368 /* A couple of helpful macros for getting the address of the 32/64 bit
2369 * fields which are the same type (int / unsigned) on our platforms.
2371 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2372 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2373 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2375 struct used_address
{
2376 struct sockaddr_storage name
;
2377 unsigned int name_len
;
2380 int __copy_msghdr(struct msghdr
*kmsg
,
2381 struct user_msghdr
*msg
,
2382 struct sockaddr __user
**save_addr
)
2386 kmsg
->msg_control_is_user
= true;
2387 kmsg
->msg_get_inq
= 0;
2388 kmsg
->msg_control_user
= msg
->msg_control
;
2389 kmsg
->msg_controllen
= msg
->msg_controllen
;
2390 kmsg
->msg_flags
= msg
->msg_flags
;
2392 kmsg
->msg_namelen
= msg
->msg_namelen
;
2394 kmsg
->msg_namelen
= 0;
2396 if (kmsg
->msg_namelen
< 0)
2399 if (kmsg
->msg_namelen
> sizeof(struct sockaddr_storage
))
2400 kmsg
->msg_namelen
= sizeof(struct sockaddr_storage
);
2403 *save_addr
= msg
->msg_name
;
2405 if (msg
->msg_name
&& kmsg
->msg_namelen
) {
2407 err
= move_addr_to_kernel(msg
->msg_name
,
2414 kmsg
->msg_name
= NULL
;
2415 kmsg
->msg_namelen
= 0;
2418 if (msg
->msg_iovlen
> UIO_MAXIOV
)
2421 kmsg
->msg_iocb
= NULL
;
2422 kmsg
->msg_ubuf
= NULL
;
2426 static int copy_msghdr_from_user(struct msghdr
*kmsg
,
2427 struct user_msghdr __user
*umsg
,
2428 struct sockaddr __user
**save_addr
,
2431 struct user_msghdr msg
;
2434 if (copy_from_user(&msg
, umsg
, sizeof(*umsg
)))
2437 err
= __copy_msghdr(kmsg
, &msg
, save_addr
);
2441 err
= import_iovec(save_addr
? ITER_DEST
: ITER_SOURCE
,
2442 msg
.msg_iov
, msg
.msg_iovlen
,
2443 UIO_FASTIOV
, iov
, &kmsg
->msg_iter
);
2444 return err
< 0 ? err
: 0;
2447 static int ____sys_sendmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2448 unsigned int flags
, struct used_address
*used_address
,
2449 unsigned int allowed_msghdr_flags
)
2451 unsigned char ctl
[sizeof(struct cmsghdr
) + 20]
2452 __aligned(sizeof(__kernel_size_t
));
2453 /* 20 is size of ipv6_pktinfo */
2454 unsigned char *ctl_buf
= ctl
;
2460 if (msg_sys
->msg_controllen
> INT_MAX
)
2462 flags
|= (msg_sys
->msg_flags
& allowed_msghdr_flags
);
2463 ctl_len
= msg_sys
->msg_controllen
;
2464 if ((MSG_CMSG_COMPAT
& flags
) && ctl_len
) {
2466 cmsghdr_from_user_compat_to_kern(msg_sys
, sock
->sk
, ctl
,
2470 ctl_buf
= msg_sys
->msg_control
;
2471 ctl_len
= msg_sys
->msg_controllen
;
2472 } else if (ctl_len
) {
2473 BUILD_BUG_ON(sizeof(struct cmsghdr
) !=
2474 CMSG_ALIGN(sizeof(struct cmsghdr
)));
2475 if (ctl_len
> sizeof(ctl
)) {
2476 ctl_buf
= sock_kmalloc(sock
->sk
, ctl_len
, GFP_KERNEL
);
2477 if (ctl_buf
== NULL
)
2481 if (copy_from_user(ctl_buf
, msg_sys
->msg_control_user
, ctl_len
))
2483 msg_sys
->msg_control
= ctl_buf
;
2484 msg_sys
->msg_control_is_user
= false;
2486 msg_sys
->msg_flags
= flags
;
2488 if (sock
->file
->f_flags
& O_NONBLOCK
)
2489 msg_sys
->msg_flags
|= MSG_DONTWAIT
;
2491 * If this is sendmmsg() and current destination address is same as
2492 * previously succeeded address, omit asking LSM's decision.
2493 * used_address->name_len is initialized to UINT_MAX so that the first
2494 * destination address never matches.
2496 if (used_address
&& msg_sys
->msg_name
&&
2497 used_address
->name_len
== msg_sys
->msg_namelen
&&
2498 !memcmp(&used_address
->name
, msg_sys
->msg_name
,
2499 used_address
->name_len
)) {
2500 err
= sock_sendmsg_nosec(sock
, msg_sys
);
2503 err
= sock_sendmsg(sock
, msg_sys
);
2505 * If this is sendmmsg() and sending to current destination address was
2506 * successful, remember it.
2508 if (used_address
&& err
>= 0) {
2509 used_address
->name_len
= msg_sys
->msg_namelen
;
2510 if (msg_sys
->msg_name
)
2511 memcpy(&used_address
->name
, msg_sys
->msg_name
,
2512 used_address
->name_len
);
2517 sock_kfree_s(sock
->sk
, ctl_buf
, ctl_len
);
2522 int sendmsg_copy_msghdr(struct msghdr
*msg
,
2523 struct user_msghdr __user
*umsg
, unsigned flags
,
2528 if (flags
& MSG_CMSG_COMPAT
) {
2529 struct compat_msghdr __user
*msg_compat
;
2531 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2532 err
= get_compat_msghdr(msg
, msg_compat
, NULL
, iov
);
2534 err
= copy_msghdr_from_user(msg
, umsg
, NULL
, iov
);
2542 static int ___sys_sendmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2543 struct msghdr
*msg_sys
, unsigned int flags
,
2544 struct used_address
*used_address
,
2545 unsigned int allowed_msghdr_flags
)
2547 struct sockaddr_storage address
;
2548 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2551 msg_sys
->msg_name
= &address
;
2553 err
= sendmsg_copy_msghdr(msg_sys
, msg
, flags
, &iov
);
2557 err
= ____sys_sendmsg(sock
, msg_sys
, flags
, used_address
,
2558 allowed_msghdr_flags
);
2564 * BSD sendmsg interface
2566 long __sys_sendmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2569 return ____sys_sendmsg(sock
, msg
, flags
, NULL
, 0);
2572 long __sys_sendmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2573 bool forbid_cmsg_compat
)
2575 int fput_needed
, err
;
2576 struct msghdr msg_sys
;
2577 struct socket
*sock
;
2579 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2582 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2586 err
= ___sys_sendmsg(sock
, msg
, &msg_sys
, flags
, NULL
, 0);
2588 fput_light(sock
->file
, fput_needed
);
2593 SYSCALL_DEFINE3(sendmsg
, int, fd
, struct user_msghdr __user
*, msg
, unsigned int, flags
)
2595 return __sys_sendmsg(fd
, msg
, flags
, true);
2599 * Linux sendmmsg interface
2602 int __sys_sendmmsg(int fd
, struct mmsghdr __user
*mmsg
, unsigned int vlen
,
2603 unsigned int flags
, bool forbid_cmsg_compat
)
2605 int fput_needed
, err
, datagrams
;
2606 struct socket
*sock
;
2607 struct mmsghdr __user
*entry
;
2608 struct compat_mmsghdr __user
*compat_entry
;
2609 struct msghdr msg_sys
;
2610 struct used_address used_address
;
2611 unsigned int oflags
= flags
;
2613 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2616 if (vlen
> UIO_MAXIOV
)
2621 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2625 used_address
.name_len
= UINT_MAX
;
2627 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2631 while (datagrams
< vlen
) {
2632 if (datagrams
== vlen
- 1)
2635 if (MSG_CMSG_COMPAT
& flags
) {
2636 err
= ___sys_sendmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2637 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2640 err
= __put_user(err
, &compat_entry
->msg_len
);
2643 err
= ___sys_sendmsg(sock
,
2644 (struct user_msghdr __user
*)entry
,
2645 &msg_sys
, flags
, &used_address
, MSG_EOR
);
2648 err
= put_user(err
, &entry
->msg_len
);
2655 if (msg_data_left(&msg_sys
))
2660 fput_light(sock
->file
, fput_needed
);
2662 /* We only return an error if no datagrams were able to be sent */
2669 SYSCALL_DEFINE4(sendmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2670 unsigned int, vlen
, unsigned int, flags
)
2672 return __sys_sendmmsg(fd
, mmsg
, vlen
, flags
, true);
2675 int recvmsg_copy_msghdr(struct msghdr
*msg
,
2676 struct user_msghdr __user
*umsg
, unsigned flags
,
2677 struct sockaddr __user
**uaddr
,
2682 if (MSG_CMSG_COMPAT
& flags
) {
2683 struct compat_msghdr __user
*msg_compat
;
2685 msg_compat
= (struct compat_msghdr __user
*) umsg
;
2686 err
= get_compat_msghdr(msg
, msg_compat
, uaddr
, iov
);
2688 err
= copy_msghdr_from_user(msg
, umsg
, uaddr
, iov
);
2696 static int ____sys_recvmsg(struct socket
*sock
, struct msghdr
*msg_sys
,
2697 struct user_msghdr __user
*msg
,
2698 struct sockaddr __user
*uaddr
,
2699 unsigned int flags
, int nosec
)
2701 struct compat_msghdr __user
*msg_compat
=
2702 (struct compat_msghdr __user
*) msg
;
2703 int __user
*uaddr_len
= COMPAT_NAMELEN(msg
);
2704 struct sockaddr_storage addr
;
2705 unsigned long cmsg_ptr
;
2709 msg_sys
->msg_name
= &addr
;
2710 cmsg_ptr
= (unsigned long)msg_sys
->msg_control
;
2711 msg_sys
->msg_flags
= flags
& (MSG_CMSG_CLOEXEC
|MSG_CMSG_COMPAT
);
2713 /* We assume all kernel code knows the size of sockaddr_storage */
2714 msg_sys
->msg_namelen
= 0;
2716 if (sock
->file
->f_flags
& O_NONBLOCK
)
2717 flags
|= MSG_DONTWAIT
;
2719 if (unlikely(nosec
))
2720 err
= sock_recvmsg_nosec(sock
, msg_sys
, flags
);
2722 err
= sock_recvmsg(sock
, msg_sys
, flags
);
2728 if (uaddr
!= NULL
) {
2729 err
= move_addr_to_user(&addr
,
2730 msg_sys
->msg_namelen
, uaddr
,
2735 err
= __put_user((msg_sys
->msg_flags
& ~MSG_CMSG_COMPAT
),
2739 if (MSG_CMSG_COMPAT
& flags
)
2740 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2741 &msg_compat
->msg_controllen
);
2743 err
= __put_user((unsigned long)msg_sys
->msg_control
- cmsg_ptr
,
2744 &msg
->msg_controllen
);
2752 static int ___sys_recvmsg(struct socket
*sock
, struct user_msghdr __user
*msg
,
2753 struct msghdr
*msg_sys
, unsigned int flags
, int nosec
)
2755 struct iovec iovstack
[UIO_FASTIOV
], *iov
= iovstack
;
2756 /* user mode address pointers */
2757 struct sockaddr __user
*uaddr
;
2760 err
= recvmsg_copy_msghdr(msg_sys
, msg
, flags
, &uaddr
, &iov
);
2764 err
= ____sys_recvmsg(sock
, msg_sys
, msg
, uaddr
, flags
, nosec
);
2770 * BSD recvmsg interface
2773 long __sys_recvmsg_sock(struct socket
*sock
, struct msghdr
*msg
,
2774 struct user_msghdr __user
*umsg
,
2775 struct sockaddr __user
*uaddr
, unsigned int flags
)
2777 return ____sys_recvmsg(sock
, msg
, umsg
, uaddr
, flags
, 0);
2780 long __sys_recvmsg(int fd
, struct user_msghdr __user
*msg
, unsigned int flags
,
2781 bool forbid_cmsg_compat
)
2783 int fput_needed
, err
;
2784 struct msghdr msg_sys
;
2785 struct socket
*sock
;
2787 if (forbid_cmsg_compat
&& (flags
& MSG_CMSG_COMPAT
))
2790 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2794 err
= ___sys_recvmsg(sock
, msg
, &msg_sys
, flags
, 0);
2796 fput_light(sock
->file
, fput_needed
);
2801 SYSCALL_DEFINE3(recvmsg
, int, fd
, struct user_msghdr __user
*, msg
,
2802 unsigned int, flags
)
2804 return __sys_recvmsg(fd
, msg
, flags
, true);
2808 * Linux recvmmsg interface
2811 static int do_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2812 unsigned int vlen
, unsigned int flags
,
2813 struct timespec64
*timeout
)
2815 int fput_needed
, err
, datagrams
;
2816 struct socket
*sock
;
2817 struct mmsghdr __user
*entry
;
2818 struct compat_mmsghdr __user
*compat_entry
;
2819 struct msghdr msg_sys
;
2820 struct timespec64 end_time
;
2821 struct timespec64 timeout64
;
2824 poll_select_set_timeout(&end_time
, timeout
->tv_sec
,
2830 sock
= sockfd_lookup_light(fd
, &err
, &fput_needed
);
2834 if (likely(!(flags
& MSG_ERRQUEUE
))) {
2835 err
= sock_error(sock
->sk
);
2843 compat_entry
= (struct compat_mmsghdr __user
*)mmsg
;
2845 while (datagrams
< vlen
) {
2847 * No need to ask LSM for more than the first datagram.
2849 if (MSG_CMSG_COMPAT
& flags
) {
2850 err
= ___sys_recvmsg(sock
, (struct user_msghdr __user
*)compat_entry
,
2851 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2855 err
= __put_user(err
, &compat_entry
->msg_len
);
2858 err
= ___sys_recvmsg(sock
,
2859 (struct user_msghdr __user
*)entry
,
2860 &msg_sys
, flags
& ~MSG_WAITFORONE
,
2864 err
= put_user(err
, &entry
->msg_len
);
2872 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2873 if (flags
& MSG_WAITFORONE
)
2874 flags
|= MSG_DONTWAIT
;
2877 ktime_get_ts64(&timeout64
);
2878 *timeout
= timespec64_sub(end_time
, timeout64
);
2879 if (timeout
->tv_sec
< 0) {
2880 timeout
->tv_sec
= timeout
->tv_nsec
= 0;
2884 /* Timeout, return less than vlen datagrams */
2885 if (timeout
->tv_nsec
== 0 && timeout
->tv_sec
== 0)
2889 /* Out of band data, return right away */
2890 if (msg_sys
.msg_flags
& MSG_OOB
)
2898 if (datagrams
== 0) {
2904 * We may return less entries than requested (vlen) if the
2905 * sock is non block and there aren't enough datagrams...
2907 if (err
!= -EAGAIN
) {
2909 * ... or if recvmsg returns an error after we
2910 * received some datagrams, where we record the
2911 * error to return on the next call or if the
2912 * app asks about it using getsockopt(SO_ERROR).
2914 sock
->sk
->sk_err
= -err
;
2917 fput_light(sock
->file
, fput_needed
);
2922 int __sys_recvmmsg(int fd
, struct mmsghdr __user
*mmsg
,
2923 unsigned int vlen
, unsigned int flags
,
2924 struct __kernel_timespec __user
*timeout
,
2925 struct old_timespec32 __user
*timeout32
)
2928 struct timespec64 timeout_sys
;
2930 if (timeout
&& get_timespec64(&timeout_sys
, timeout
))
2933 if (timeout32
&& get_old_timespec32(&timeout_sys
, timeout32
))
2936 if (!timeout
&& !timeout32
)
2937 return do_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
);
2939 datagrams
= do_recvmmsg(fd
, mmsg
, vlen
, flags
, &timeout_sys
);
2944 if (timeout
&& put_timespec64(&timeout_sys
, timeout
))
2945 datagrams
= -EFAULT
;
2947 if (timeout32
&& put_old_timespec32(&timeout_sys
, timeout32
))
2948 datagrams
= -EFAULT
;
2953 SYSCALL_DEFINE5(recvmmsg
, int, fd
, struct mmsghdr __user
*, mmsg
,
2954 unsigned int, vlen
, unsigned int, flags
,
2955 struct __kernel_timespec __user
*, timeout
)
2957 if (flags
& MSG_CMSG_COMPAT
)
2960 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, timeout
, NULL
);
2963 #ifdef CONFIG_COMPAT_32BIT_TIME
2964 SYSCALL_DEFINE5(recvmmsg_time32
, int, fd
, struct mmsghdr __user
*, mmsg
,
2965 unsigned int, vlen
, unsigned int, flags
,
2966 struct old_timespec32 __user
*, timeout
)
2968 if (flags
& MSG_CMSG_COMPAT
)
2971 return __sys_recvmmsg(fd
, mmsg
, vlen
, flags
, NULL
, timeout
);
2975 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2976 /* Argument list sizes for sys_socketcall */
2977 #define AL(x) ((x) * sizeof(unsigned long))
2978 static const unsigned char nargs
[21] = {
2979 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2980 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2981 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2988 * System call vectors.
2990 * Argument checking cleaned up. Saved 20% in size.
2991 * This function doesn't need to set the kernel lock because
2992 * it is set by the callees.
2995 SYSCALL_DEFINE2(socketcall
, int, call
, unsigned long __user
*, args
)
2997 unsigned long a
[AUDITSC_ARGS
];
2998 unsigned long a0
, a1
;
3002 if (call
< 1 || call
> SYS_SENDMMSG
)
3004 call
= array_index_nospec(call
, SYS_SENDMMSG
+ 1);
3007 if (len
> sizeof(a
))
3010 /* copy_from_user should be SMP safe. */
3011 if (copy_from_user(a
, args
, len
))
3014 err
= audit_socketcall(nargs
[call
] / sizeof(unsigned long), a
);
3023 err
= __sys_socket(a0
, a1
, a
[2]);
3026 err
= __sys_bind(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3029 err
= __sys_connect(a0
, (struct sockaddr __user
*)a1
, a
[2]);
3032 err
= __sys_listen(a0
, a1
);
3035 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3036 (int __user
*)a
[2], 0);
3038 case SYS_GETSOCKNAME
:
3040 __sys_getsockname(a0
, (struct sockaddr __user
*)a1
,
3041 (int __user
*)a
[2]);
3043 case SYS_GETPEERNAME
:
3045 __sys_getpeername(a0
, (struct sockaddr __user
*)a1
,
3046 (int __user
*)a
[2]);
3048 case SYS_SOCKETPAIR
:
3049 err
= __sys_socketpair(a0
, a1
, a
[2], (int __user
*)a
[3]);
3052 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3056 err
= __sys_sendto(a0
, (void __user
*)a1
, a
[2], a
[3],
3057 (struct sockaddr __user
*)a
[4], a
[5]);
3060 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3064 err
= __sys_recvfrom(a0
, (void __user
*)a1
, a
[2], a
[3],
3065 (struct sockaddr __user
*)a
[4],
3066 (int __user
*)a
[5]);
3069 err
= __sys_shutdown(a0
, a1
);
3071 case SYS_SETSOCKOPT
:
3072 err
= __sys_setsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3075 case SYS_GETSOCKOPT
:
3077 __sys_getsockopt(a0
, a1
, a
[2], (char __user
*)a
[3],
3078 (int __user
*)a
[4]);
3081 err
= __sys_sendmsg(a0
, (struct user_msghdr __user
*)a1
,
3085 err
= __sys_sendmmsg(a0
, (struct mmsghdr __user
*)a1
, a
[2],
3089 err
= __sys_recvmsg(a0
, (struct user_msghdr __user
*)a1
,
3093 if (IS_ENABLED(CONFIG_64BIT
))
3094 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3096 (struct __kernel_timespec __user
*)a
[4],
3099 err
= __sys_recvmmsg(a0
, (struct mmsghdr __user
*)a1
,
3101 (struct old_timespec32 __user
*)a
[4]);
3104 err
= __sys_accept4(a0
, (struct sockaddr __user
*)a1
,
3105 (int __user
*)a
[2], a
[3]);
3114 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3117 * sock_register - add a socket protocol handler
3118 * @ops: description of protocol
3120 * This function is called by a protocol handler that wants to
3121 * advertise its address family, and have it linked into the
3122 * socket interface. The value ops->family corresponds to the
3123 * socket system call protocol family.
3125 int sock_register(const struct net_proto_family
*ops
)
3129 if (ops
->family
>= NPROTO
) {
3130 pr_crit("protocol %d >= NPROTO(%d)\n", ops
->family
, NPROTO
);
3134 spin_lock(&net_family_lock
);
3135 if (rcu_dereference_protected(net_families
[ops
->family
],
3136 lockdep_is_held(&net_family_lock
)))
3139 rcu_assign_pointer(net_families
[ops
->family
], ops
);
3142 spin_unlock(&net_family_lock
);
3144 pr_info("NET: Registered %s protocol family\n", pf_family_names
[ops
->family
]);
3147 EXPORT_SYMBOL(sock_register
);
3150 * sock_unregister - remove a protocol handler
3151 * @family: protocol family to remove
3153 * This function is called by a protocol handler that wants to
3154 * remove its address family, and have it unlinked from the
3155 * new socket creation.
3157 * If protocol handler is a module, then it can use module reference
3158 * counts to protect against new references. If protocol handler is not
3159 * a module then it needs to provide its own protection in
3160 * the ops->create routine.
3162 void sock_unregister(int family
)
3164 BUG_ON(family
< 0 || family
>= NPROTO
);
3166 spin_lock(&net_family_lock
);
3167 RCU_INIT_POINTER(net_families
[family
], NULL
);
3168 spin_unlock(&net_family_lock
);
3172 pr_info("NET: Unregistered %s protocol family\n", pf_family_names
[family
]);
3174 EXPORT_SYMBOL(sock_unregister
);
3176 bool sock_is_registered(int family
)
3178 return family
< NPROTO
&& rcu_access_pointer(net_families
[family
]);
3181 static int __init
sock_init(void)
3185 * Initialize the network sysctl infrastructure.
3187 err
= net_sysctl_init();
3192 * Initialize skbuff SLAB cache
3197 * Initialize the protocols module.
3202 err
= register_filesystem(&sock_fs_type
);
3205 sock_mnt
= kern_mount(&sock_fs_type
);
3206 if (IS_ERR(sock_mnt
)) {
3207 err
= PTR_ERR(sock_mnt
);
3211 /* The real protocol initialization is performed in later initcalls.
3214 #ifdef CONFIG_NETFILTER
3215 err
= netfilter_init();
3220 ptp_classifier_init();
3226 unregister_filesystem(&sock_fs_type
);
3230 core_initcall(sock_init
); /* early initcall */
3232 #ifdef CONFIG_PROC_FS
3233 void socket_seq_show(struct seq_file
*seq
)
3235 seq_printf(seq
, "sockets: used %d\n",
3236 sock_inuse_get(seq
->private));
3238 #endif /* CONFIG_PROC_FS */
3240 /* Handle the fact that while struct ifreq has the same *layout* on
3241 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3242 * which are handled elsewhere, it still has different *size* due to
3243 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3244 * resulting in struct ifreq being 32 and 40 bytes respectively).
3245 * As a result, if the struct happens to be at the end of a page and
3246 * the next page isn't readable/writable, we get a fault. To prevent
3247 * that, copy back and forth to the full size.
3249 int get_user_ifreq(struct ifreq
*ifr
, void __user
**ifrdata
, void __user
*arg
)
3251 if (in_compat_syscall()) {
3252 struct compat_ifreq
*ifr32
= (struct compat_ifreq
*)ifr
;
3254 memset(ifr
, 0, sizeof(*ifr
));
3255 if (copy_from_user(ifr32
, arg
, sizeof(*ifr32
)))
3259 *ifrdata
= compat_ptr(ifr32
->ifr_data
);
3264 if (copy_from_user(ifr
, arg
, sizeof(*ifr
)))
3268 *ifrdata
= ifr
->ifr_data
;
3272 EXPORT_SYMBOL(get_user_ifreq
);
3274 int put_user_ifreq(struct ifreq
*ifr
, void __user
*arg
)
3276 size_t size
= sizeof(*ifr
);
3278 if (in_compat_syscall())
3279 size
= sizeof(struct compat_ifreq
);
3281 if (copy_to_user(arg
, ifr
, size
))
3286 EXPORT_SYMBOL(put_user_ifreq
);
3288 #ifdef CONFIG_COMPAT
3289 static int compat_siocwandev(struct net
*net
, struct compat_ifreq __user
*uifr32
)
3291 compat_uptr_t uptr32
;
3296 if (get_user_ifreq(&ifr
, NULL
, uifr32
))
3299 if (get_user(uptr32
, &uifr32
->ifr_settings
.ifs_ifsu
))
3302 saved
= ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
;
3303 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= compat_ptr(uptr32
);
3305 err
= dev_ioctl(net
, SIOCWANDEV
, &ifr
, NULL
, NULL
);
3307 ifr
.ifr_settings
.ifs_ifsu
.raw_hdlc
= saved
;
3308 if (put_user_ifreq(&ifr
, uifr32
))
3314 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3315 static int compat_ifr_data_ioctl(struct net
*net
, unsigned int cmd
,
3316 struct compat_ifreq __user
*u_ifreq32
)
3321 if (!is_socket_ioctl_cmd(cmd
))
3323 if (get_user_ifreq(&ifreq
, &data
, u_ifreq32
))
3325 ifreq
.ifr_data
= data
;
3327 return dev_ioctl(net
, cmd
, &ifreq
, data
, NULL
);
3330 static int compat_sock_ioctl_trans(struct file
*file
, struct socket
*sock
,
3331 unsigned int cmd
, unsigned long arg
)
3333 void __user
*argp
= compat_ptr(arg
);
3334 struct sock
*sk
= sock
->sk
;
3335 struct net
*net
= sock_net(sk
);
3337 if (cmd
>= SIOCDEVPRIVATE
&& cmd
<= (SIOCDEVPRIVATE
+ 15))
3338 return sock_ioctl(file
, cmd
, (unsigned long)argp
);
3342 return compat_siocwandev(net
, argp
);
3343 case SIOCGSTAMP_OLD
:
3344 case SIOCGSTAMPNS_OLD
:
3345 if (!sock
->ops
->gettstamp
)
3346 return -ENOIOCTLCMD
;
3347 return sock
->ops
->gettstamp(sock
, argp
, cmd
== SIOCGSTAMP_OLD
,
3348 !COMPAT_USE_64BIT_TIME
);
3351 case SIOCBONDSLAVEINFOQUERY
:
3352 case SIOCBONDINFOQUERY
:
3355 return compat_ifr_data_ioctl(net
, cmd
, argp
);
3366 case SIOCGSTAMP_NEW
:
3367 case SIOCGSTAMPNS_NEW
:
3371 return sock_ioctl(file
, cmd
, arg
);
3390 case SIOCSIFHWBROADCAST
:
3392 case SIOCGIFBRDADDR
:
3393 case SIOCSIFBRDADDR
:
3394 case SIOCGIFDSTADDR
:
3395 case SIOCSIFDSTADDR
:
3396 case SIOCGIFNETMASK
:
3397 case SIOCSIFNETMASK
:
3409 case SIOCBONDENSLAVE
:
3410 case SIOCBONDRELEASE
:
3411 case SIOCBONDSETHWADDR
:
3412 case SIOCBONDCHANGEACTIVE
:
3419 return sock_do_ioctl(net
, sock
, cmd
, arg
);
3422 return -ENOIOCTLCMD
;
3425 static long compat_sock_ioctl(struct file
*file
, unsigned int cmd
,
3428 struct socket
*sock
= file
->private_data
;
3429 int ret
= -ENOIOCTLCMD
;
3436 if (sock
->ops
->compat_ioctl
)
3437 ret
= sock
->ops
->compat_ioctl(sock
, cmd
, arg
);
3439 if (ret
== -ENOIOCTLCMD
&&
3440 (cmd
>= SIOCIWFIRST
&& cmd
<= SIOCIWLAST
))
3441 ret
= compat_wext_handle_ioctl(net
, cmd
, arg
);
3443 if (ret
== -ENOIOCTLCMD
)
3444 ret
= compat_sock_ioctl_trans(file
, sock
, cmd
, arg
);
3451 * kernel_bind - bind an address to a socket (kernel space)
3454 * @addrlen: length of address
3456 * Returns 0 or an error.
3459 int kernel_bind(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
)
3461 return sock
->ops
->bind(sock
, addr
, addrlen
);
3463 EXPORT_SYMBOL(kernel_bind
);
3466 * kernel_listen - move socket to listening state (kernel space)
3468 * @backlog: pending connections queue size
3470 * Returns 0 or an error.
3473 int kernel_listen(struct socket
*sock
, int backlog
)
3475 return sock
->ops
->listen(sock
, backlog
);
3477 EXPORT_SYMBOL(kernel_listen
);
3480 * kernel_accept - accept a connection (kernel space)
3481 * @sock: listening socket
3482 * @newsock: new connected socket
3485 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3486 * If it fails, @newsock is guaranteed to be %NULL.
3487 * Returns 0 or an error.
3490 int kernel_accept(struct socket
*sock
, struct socket
**newsock
, int flags
)
3492 struct sock
*sk
= sock
->sk
;
3495 err
= sock_create_lite(sk
->sk_family
, sk
->sk_type
, sk
->sk_protocol
,
3500 err
= sock
->ops
->accept(sock
, *newsock
, flags
, true);
3502 sock_release(*newsock
);
3507 (*newsock
)->ops
= sock
->ops
;
3508 __module_get((*newsock
)->ops
->owner
);
3513 EXPORT_SYMBOL(kernel_accept
);
3516 * kernel_connect - connect a socket (kernel space)
3519 * @addrlen: address length
3520 * @flags: flags (O_NONBLOCK, ...)
3522 * For datagram sockets, @addr is the address to which datagrams are sent
3523 * by default, and the only address from which datagrams are received.
3524 * For stream sockets, attempts to connect to @addr.
3525 * Returns 0 or an error code.
3528 int kernel_connect(struct socket
*sock
, struct sockaddr
*addr
, int addrlen
,
3531 return sock
->ops
->connect(sock
, addr
, addrlen
, flags
);
3533 EXPORT_SYMBOL(kernel_connect
);
3536 * kernel_getsockname - get the address which the socket is bound (kernel space)
3538 * @addr: address holder
3540 * Fills the @addr pointer with the address which the socket is bound.
3541 * Returns the length of the address in bytes or an error code.
3544 int kernel_getsockname(struct socket
*sock
, struct sockaddr
*addr
)
3546 return sock
->ops
->getname(sock
, addr
, 0);
3548 EXPORT_SYMBOL(kernel_getsockname
);
3551 * kernel_getpeername - get the address which the socket is connected (kernel space)
3553 * @addr: address holder
3555 * Fills the @addr pointer with the address which the socket is connected.
3556 * Returns the length of the address in bytes or an error code.
3559 int kernel_getpeername(struct socket
*sock
, struct sockaddr
*addr
)
3561 return sock
->ops
->getname(sock
, addr
, 1);
3563 EXPORT_SYMBOL(kernel_getpeername
);
3566 * kernel_sendpage - send a &page through a socket (kernel space)
3569 * @offset: page offset
3570 * @size: total size in bytes
3571 * @flags: flags (MSG_DONTWAIT, ...)
3573 * Returns the total amount sent in bytes or an error.
3576 int kernel_sendpage(struct socket
*sock
, struct page
*page
, int offset
,
3577 size_t size
, int flags
)
3579 if (sock
->ops
->sendpage
) {
3580 /* Warn in case the improper page to zero-copy send */
3581 WARN_ONCE(!sendpage_ok(page
), "improper page for zero-copy send");
3582 return sock
->ops
->sendpage(sock
, page
, offset
, size
, flags
);
3584 return sock_no_sendpage(sock
, page
, offset
, size
, flags
);
3586 EXPORT_SYMBOL(kernel_sendpage
);
3589 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3592 * @offset: page offset
3593 * @size: total size in bytes
3594 * @flags: flags (MSG_DONTWAIT, ...)
3596 * Returns the total amount sent in bytes or an error.
3597 * Caller must hold @sk.
3600 int kernel_sendpage_locked(struct sock
*sk
, struct page
*page
, int offset
,
3601 size_t size
, int flags
)
3603 struct socket
*sock
= sk
->sk_socket
;
3605 if (sock
->ops
->sendpage_locked
)
3606 return sock
->ops
->sendpage_locked(sk
, page
, offset
, size
,
3609 return sock_no_sendpage_locked(sk
, page
, offset
, size
, flags
);
3611 EXPORT_SYMBOL(kernel_sendpage_locked
);
3614 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3616 * @how: connection part
3618 * Returns 0 or an error.
3621 int kernel_sock_shutdown(struct socket
*sock
, enum sock_shutdown_cmd how
)
3623 return sock
->ops
->shutdown(sock
, how
);
3625 EXPORT_SYMBOL(kernel_sock_shutdown
);
3628 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3631 * This routine returns the IP overhead imposed by a socket i.e.
3632 * the length of the underlying IP header, depending on whether
3633 * this is an IPv4 or IPv6 socket and the length from IP options turned
3634 * on at the socket. Assumes that the caller has a lock on the socket.
3637 u32
kernel_sock_ip_overhead(struct sock
*sk
)
3639 struct inet_sock
*inet
;
3640 struct ip_options_rcu
*opt
;
3642 #if IS_ENABLED(CONFIG_IPV6)
3643 struct ipv6_pinfo
*np
;
3644 struct ipv6_txoptions
*optv6
= NULL
;
3645 #endif /* IS_ENABLED(CONFIG_IPV6) */
3650 switch (sk
->sk_family
) {
3653 overhead
+= sizeof(struct iphdr
);
3654 opt
= rcu_dereference_protected(inet
->inet_opt
,
3655 sock_owned_by_user(sk
));
3657 overhead
+= opt
->opt
.optlen
;
3659 #if IS_ENABLED(CONFIG_IPV6)
3662 overhead
+= sizeof(struct ipv6hdr
);
3664 optv6
= rcu_dereference_protected(np
->opt
,
3665 sock_owned_by_user(sk
));
3667 overhead
+= (optv6
->opt_flen
+ optv6
->opt_nflen
);
3669 #endif /* IS_ENABLED(CONFIG_IPV6) */
3670 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3674 EXPORT_SYMBOL(kernel_sock_ip_overhead
);