xref: /openbmc/linux/net/socket.c (revision dd98d289)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * NET		An implementation of the SOCKET network access protocol.
4  *
5  * Version:	@(#)socket.c	1.1.93	18/02/95
6  *
7  * Authors:	Orest Zborowski, <obz@Kodak.COM>
8  *		Ross Biro
9  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10  *
11  * Fixes:
12  *		Anonymous	:	NOTSOCK/BADF cleanup. Error fix in
13  *					shutdown()
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
18  *					top level.
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
23  *					tty drivers).
24  *		Niibe Yutaka	:	Asynchronous I/O for writes (4.4BSD style)
25  *		Jeff Uphoff	:	Made max number of sockets command-line
26  *					configurable.
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
35  *					stuff.
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
41  *					moment.
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
48  *
49  *	This module is effectively the top level interface to the BSD socket
50  *	paradigm.
51  *
52  *	Based upon Swansea University Computer Society NET3.039
53  */
54 
55 #include <linux/ethtool.h>
56 #include <linux/mm.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>
89 
90 #include <linux/uaccess.h>
91 #include <asm/unistd.h>
92 
93 #include <net/compat.h>
94 #include <net/wext.h>
95 #include <net/cls_cgroup.h>
96 
97 #include <net/sock.h>
98 #include <linux/netfilter.h>
99 
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>
108 
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;
112 #endif
113 
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);
117 
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);
122 #ifdef CONFIG_COMPAT
123 static long compat_sock_ioctl(struct file *file,
124 			      unsigned int cmd, unsigned long arg);
125 #endif
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,
131 				unsigned int flags);
132 
133 #ifdef CONFIG_PROC_FS
134 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135 {
136 	struct socket *sock = f->private_data;
137 
138 	if (sock->ops->show_fdinfo)
139 		sock->ops->show_fdinfo(m, sock);
140 }
141 #else
142 #define sock_show_fdinfo NULL
143 #endif
144 
145 /*
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.
148  */
149 
150 static const struct file_operations socket_file_ops = {
151 	.owner =	THIS_MODULE,
152 	.llseek =	no_llseek,
153 	.read_iter =	sock_read_iter,
154 	.write_iter =	sock_write_iter,
155 	.poll =		sock_poll,
156 	.unlocked_ioctl = sock_ioctl,
157 #ifdef CONFIG_COMPAT
158 	.compat_ioctl = compat_sock_ioctl,
159 #endif
160 	.mmap =		sock_mmap,
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,
167 };
168 
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",
174 	[PF_IPX]	= "PF_IPX",
175 	[PF_APPLETALK]	= "PF_APPLETALK",
176 	[PF_NETROM]	= "PF_NETROM",
177 	[PF_BRIDGE]	= "PF_BRIDGE",
178 	[PF_ATMPVC]	= "PF_ATMPVC",
179 	[PF_X25]	= "PF_X25",
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",
185 	[PF_KEY]	= "PF_KEY",
186 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
187 	[PF_PACKET]	= "PF_PACKET",
188 	[PF_ASH]	= "PF_ASH",
189 	[PF_ECONET]	= "PF_ECONET",
190 	[PF_ATMSVC]	= "PF_ATMSVC",
191 	[PF_RDS]	= "PF_RDS",
192 	[PF_SNA]	= "PF_SNA",
193 	[PF_IRDA]	= "PF_IRDA",
194 	[PF_PPPOX]	= "PF_PPPOX",
195 	[PF_WANPIPE]	= "PF_WANPIPE",
196 	[PF_LLC]	= "PF_LLC",
197 	[PF_IB]		= "PF_IB",
198 	[PF_MPLS]	= "PF_MPLS",
199 	[PF_CAN]	= "PF_CAN",
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",
208 	[PF_ALG]	= "PF_ALG",
209 	[PF_NFC]	= "PF_NFC",
210 	[PF_VSOCK]	= "PF_VSOCK",
211 	[PF_KCM]	= "PF_KCM",
212 	[PF_QIPCRTR]	= "PF_QIPCRTR",
213 	[PF_SMC]	= "PF_SMC",
214 	[PF_XDP]	= "PF_XDP",
215 };
216 
217 /*
218  *	The protocol list. Each protocol is registered in here.
219  */
220 
221 static DEFINE_SPINLOCK(net_family_lock);
222 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
223 
224 /*
225  * Support routines.
226  * Move socket addresses back and forth across the kernel/user
227  * divide and look after the messy bits.
228  */
229 
230 /**
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
235  *
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.
239  */
240 
241 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
242 {
243 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
244 		return -EINVAL;
245 	if (ulen == 0)
246 		return 0;
247 	if (copy_from_user(kaddr, uaddr, ulen))
248 		return -EFAULT;
249 	return audit_sockaddr(ulen, kaddr);
250 }
251 
252 /**
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
258  *
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
263  *	accessible.
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.
267  */
268 
269 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
270 			     void __user *uaddr, int __user *ulen)
271 {
272 	int err;
273 	int len;
274 
275 	BUG_ON(klen > sizeof(struct sockaddr_storage));
276 	err = get_user(len, ulen);
277 	if (err)
278 		return err;
279 	if (len > klen)
280 		len = klen;
281 	if (len < 0)
282 		return -EINVAL;
283 	if (len) {
284 		if (audit_sockaddr(klen, kaddr))
285 			return -ENOMEM;
286 		if (copy_to_user(uaddr, kaddr, len))
287 			return -EFAULT;
288 	}
289 	/*
290 	 *      "fromlen shall refer to the value before truncation.."
291 	 *                      1003.1g
292 	 */
293 	return __put_user(klen, ulen);
294 }
295 
296 static struct kmem_cache *sock_inode_cachep __ro_after_init;
297 
298 static struct inode *sock_alloc_inode(struct super_block *sb)
299 {
300 	struct socket_alloc *ei;
301 
302 	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
303 	if (!ei)
304 		return NULL;
305 	init_waitqueue_head(&ei->socket.wq.wait);
306 	ei->socket.wq.fasync_list = NULL;
307 	ei->socket.wq.flags = 0;
308 
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;
314 
315 	return &ei->vfs_inode;
316 }
317 
318 static void sock_free_inode(struct inode *inode)
319 {
320 	struct socket_alloc *ei;
321 
322 	ei = container_of(inode, struct socket_alloc, vfs_inode);
323 	kmem_cache_free(sock_inode_cachep, ei);
324 }
325 
326 static void init_once(void *foo)
327 {
328 	struct socket_alloc *ei = (struct socket_alloc *)foo;
329 
330 	inode_init_once(&ei->vfs_inode);
331 }
332 
333 static void init_inodecache(void)
334 {
335 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
336 					      sizeof(struct socket_alloc),
337 					      0,
338 					      (SLAB_HWCACHE_ALIGN |
339 					       SLAB_RECLAIM_ACCOUNT |
340 					       SLAB_MEM_SPREAD | SLAB_ACCOUNT),
341 					      init_once);
342 	BUG_ON(sock_inode_cachep == NULL);
343 }
344 
345 static const struct super_operations sockfs_ops = {
346 	.alloc_inode	= sock_alloc_inode,
347 	.free_inode	= sock_free_inode,
348 	.statfs		= simple_statfs,
349 };
350 
351 /*
352  * sockfs_dname() is called from d_path().
353  */
354 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
355 {
356 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
357 				d_inode(dentry)->i_ino);
358 }
359 
360 static const struct dentry_operations sockfs_dentry_operations = {
361 	.d_dname  = sockfs_dname,
362 };
363 
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)
367 {
368 	if (value) {
369 		if (dentry->d_name.len + 1 > size)
370 			return -ERANGE;
371 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
372 	}
373 	return dentry->d_name.len + 1;
374 }
375 
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)
379 
380 static const struct xattr_handler sockfs_xattr_handler = {
381 	.name = XATTR_NAME_SOCKPROTONAME,
382 	.get = sockfs_xattr_get,
383 };
384 
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)
390 {
391 	/* Handled by LSM. */
392 	return -EAGAIN;
393 }
394 
395 static const struct xattr_handler sockfs_security_xattr_handler = {
396 	.prefix = XATTR_SECURITY_PREFIX,
397 	.set = sockfs_security_xattr_set,
398 };
399 
400 static const struct xattr_handler *sockfs_xattr_handlers[] = {
401 	&sockfs_xattr_handler,
402 	&sockfs_security_xattr_handler,
403 	NULL
404 };
405 
406 static int sockfs_init_fs_context(struct fs_context *fc)
407 {
408 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
409 	if (!ctx)
410 		return -ENOMEM;
411 	ctx->ops = &sockfs_ops;
412 	ctx->dops = &sockfs_dentry_operations;
413 	ctx->xattr = sockfs_xattr_handlers;
414 	return 0;
415 }
416 
417 static struct vfsmount *sock_mnt __read_mostly;
418 
419 static struct file_system_type sock_fs_type = {
420 	.name =		"sockfs",
421 	.init_fs_context = sockfs_init_fs_context,
422 	.kill_sb =	kill_anon_super,
423 };
424 
425 /*
426  *	Obtains the first available file descriptor and sets it up for use.
427  *
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.
435  *
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.
440  */
441 
442 /**
443  *	sock_alloc_file - Bind a &socket to a &file
444  *	@sock: socket
445  *	@flags: file status flags
446  *	@dname: protocol name
447  *
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.
452  */
453 
454 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
455 {
456 	struct file *file;
457 
458 	if (!dname)
459 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
460 
461 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
462 				O_RDWR | (flags & O_NONBLOCK),
463 				&socket_file_ops);
464 	if (IS_ERR(file)) {
465 		sock_release(sock);
466 		return file;
467 	}
468 
469 	sock->file = file;
470 	file->private_data = sock;
471 	stream_open(SOCK_INODE(sock), file);
472 	return file;
473 }
474 EXPORT_SYMBOL(sock_alloc_file);
475 
476 static int sock_map_fd(struct socket *sock, int flags)
477 {
478 	struct file *newfile;
479 	int fd = get_unused_fd_flags(flags);
480 	if (unlikely(fd < 0)) {
481 		sock_release(sock);
482 		return fd;
483 	}
484 
485 	newfile = sock_alloc_file(sock, flags, NULL);
486 	if (!IS_ERR(newfile)) {
487 		fd_install(fd, newfile);
488 		return fd;
489 	}
490 
491 	put_unused_fd(fd);
492 	return PTR_ERR(newfile);
493 }
494 
495 /**
496  *	sock_from_file - Return the &socket bounded to @file.
497  *	@file: file
498  *
499  *	On failure returns %NULL.
500  */
501 
502 struct socket *sock_from_file(struct file *file)
503 {
504 	if (file->f_op == &socket_file_ops)
505 		return file->private_data;	/* set in sock_map_fd */
506 
507 	return NULL;
508 }
509 EXPORT_SYMBOL(sock_from_file);
510 
511 /**
512  *	sockfd_lookup - Go from a file number to its socket slot
513  *	@fd: file handle
514  *	@err: pointer to an error code return
515  *
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.
520  *
521  *	On a success the socket object pointer is returned.
522  */
523 
524 struct socket *sockfd_lookup(int fd, int *err)
525 {
526 	struct file *file;
527 	struct socket *sock;
528 
529 	file = fget(fd);
530 	if (!file) {
531 		*err = -EBADF;
532 		return NULL;
533 	}
534 
535 	sock = sock_from_file(file);
536 	if (!sock) {
537 		*err = -ENOTSOCK;
538 		fput(file);
539 	}
540 	return sock;
541 }
542 EXPORT_SYMBOL(sockfd_lookup);
543 
544 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
545 {
546 	struct fd f = fdget(fd);
547 	struct socket *sock;
548 
549 	*err = -EBADF;
550 	if (f.file) {
551 		sock = sock_from_file(f.file);
552 		if (likely(sock)) {
553 			*fput_needed = f.flags & FDPUT_FPUT;
554 			return sock;
555 		}
556 		*err = -ENOTSOCK;
557 		fdput(f);
558 	}
559 	return NULL;
560 }
561 
562 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
563 				size_t size)
564 {
565 	ssize_t len;
566 	ssize_t used = 0;
567 
568 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
569 	if (len < 0)
570 		return len;
571 	used += len;
572 	if (buffer) {
573 		if (size < used)
574 			return -ERANGE;
575 		buffer += len;
576 	}
577 
578 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
579 	used += len;
580 	if (buffer) {
581 		if (size < used)
582 			return -ERANGE;
583 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
584 		buffer += len;
585 	}
586 
587 	return used;
588 }
589 
590 static int sockfs_setattr(struct user_namespace *mnt_userns,
591 			  struct dentry *dentry, struct iattr *iattr)
592 {
593 	int err = simple_setattr(&init_user_ns, dentry, iattr);
594 
595 	if (!err && (iattr->ia_valid & ATTR_UID)) {
596 		struct socket *sock = SOCKET_I(d_inode(dentry));
597 
598 		if (sock->sk)
599 			sock->sk->sk_uid = iattr->ia_uid;
600 		else
601 			err = -ENOENT;
602 	}
603 
604 	return err;
605 }
606 
607 static const struct inode_operations sockfs_inode_ops = {
608 	.listxattr = sockfs_listxattr,
609 	.setattr = sockfs_setattr,
610 };
611 
612 /**
613  *	sock_alloc - allocate a socket
614  *
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.
618  */
619 
620 struct socket *sock_alloc(void)
621 {
622 	struct inode *inode;
623 	struct socket *sock;
624 
625 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
626 	if (!inode)
627 		return NULL;
628 
629 	sock = SOCKET_I(inode);
630 
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;
636 
637 	return sock;
638 }
639 EXPORT_SYMBOL(sock_alloc);
640 
641 static void __sock_release(struct socket *sock, struct inode *inode)
642 {
643 	if (sock->ops) {
644 		struct module *owner = sock->ops->owner;
645 
646 		if (inode)
647 			inode_lock(inode);
648 		sock->ops->release(sock);
649 		sock->sk = NULL;
650 		if (inode)
651 			inode_unlock(inode);
652 		sock->ops = NULL;
653 		module_put(owner);
654 	}
655 
656 	if (sock->wq.fasync_list)
657 		pr_err("%s: fasync list not empty!\n", __func__);
658 
659 	if (!sock->file) {
660 		iput(SOCK_INODE(sock));
661 		return;
662 	}
663 	sock->file = NULL;
664 }
665 
666 /**
667  *	sock_release - close a socket
668  *	@sock: socket to close
669  *
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.
673  */
674 void sock_release(struct socket *sock)
675 {
676 	__sock_release(sock, NULL);
677 }
678 EXPORT_SYMBOL(sock_release);
679 
680 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
681 {
682 	u8 flags = *tx_flags;
683 
684 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
685 		flags |= SKBTX_HW_TSTAMP;
686 
687 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
688 		flags |= SKBTX_SW_TSTAMP;
689 
690 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
691 		flags |= SKBTX_SCHED_TSTAMP;
692 
693 	*tx_flags = flags;
694 }
695 EXPORT_SYMBOL(__sock_tx_timestamp);
696 
697 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
698 					   size_t));
699 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
700 					    size_t));
701 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
702 {
703 	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
704 				     inet_sendmsg, sock, msg,
705 				     msg_data_left(msg));
706 	BUG_ON(ret == -EIOCBQUEUED);
707 	return ret;
708 }
709 
710 /**
711  *	sock_sendmsg - send a message through @sock
712  *	@sock: socket
713  *	@msg: message to send
714  *
715  *	Sends @msg through @sock, passing through LSM.
716  *	Returns the number of bytes sent, or an error code.
717  */
718 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
719 {
720 	int err = security_socket_sendmsg(sock, msg,
721 					  msg_data_left(msg));
722 
723 	return err ?: sock_sendmsg_nosec(sock, msg);
724 }
725 EXPORT_SYMBOL(sock_sendmsg);
726 
727 /**
728  *	kernel_sendmsg - send a message through @sock (kernel-space)
729  *	@sock: socket
730  *	@msg: message header
731  *	@vec: kernel vec
732  *	@num: vec array length
733  *	@size: total message data size
734  *
735  *	Builds the message data with @vec and sends it through @sock.
736  *	Returns the number of bytes sent, or an error code.
737  */
738 
739 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
740 		   struct kvec *vec, size_t num, size_t size)
741 {
742 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
743 	return sock_sendmsg(sock, msg);
744 }
745 EXPORT_SYMBOL(kernel_sendmsg);
746 
747 /**
748  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
749  *	@sk: sock
750  *	@msg: message header
751  *	@vec: output s/g array
752  *	@num: output s/g array length
753  *	@size: total message data size
754  *
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.
758  */
759 
760 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
761 			  struct kvec *vec, size_t num, size_t size)
762 {
763 	struct socket *sock = sk->sk_socket;
764 
765 	if (!sock->ops->sendmsg_locked)
766 		return sock_no_sendmsg_locked(sk, msg, size);
767 
768 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
769 
770 	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
771 }
772 EXPORT_SYMBOL(kernel_sendmsg_locked);
773 
774 static bool skb_is_err_queue(const struct sk_buff *skb)
775 {
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.
780 	 */
781 	return skb->pkt_type == PACKET_OUTGOING;
782 }
783 
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.
791  */
792 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
793 {
794 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
795 }
796 
797 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
798 {
799 	struct scm_ts_pktinfo ts_pktinfo;
800 	struct net_device *orig_dev;
801 
802 	if (!skb_mac_header_was_set(skb))
803 		return;
804 
805 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
806 
807 	rcu_read_lock();
808 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
809 	if (orig_dev)
810 		ts_pktinfo.if_index = orig_dev->ifindex;
811 	rcu_read_unlock();
812 
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);
816 }
817 
818 /*
819  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
820  */
821 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
822 	struct sk_buff *skb)
823 {
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;
827 
828 	int empty = 1, false_tstamp = 0;
829 	struct skb_shared_hwtstamps *shhwtstamps =
830 		skb_hwtstamps(skb);
831 
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);
836 		false_tstamp = 1;
837 	}
838 
839 	if (need_software_tstamp) {
840 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
841 			if (new_tstamp) {
842 				struct __kernel_sock_timeval tv;
843 
844 				skb_get_new_timestamp(skb, &tv);
845 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
846 					 sizeof(tv), &tv);
847 			} else {
848 				struct __kernel_old_timeval tv;
849 
850 				skb_get_timestamp(skb, &tv);
851 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
852 					 sizeof(tv), &tv);
853 			}
854 		} else {
855 			if (new_tstamp) {
856 				struct __kernel_timespec ts;
857 
858 				skb_get_new_timestampns(skb, &ts);
859 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
860 					 sizeof(ts), &ts);
861 			} else {
862 				struct __kernel_old_timespec ts;
863 
864 				skb_get_timestampns(skb, &ts);
865 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
866 					 sizeof(ts), &ts);
867 			}
868 		}
869 	}
870 
871 	memset(&tss, 0, sizeof(tss));
872 	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
873 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
874 		empty = 0;
875 	if (shhwtstamps &&
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);
880 
881 		if (ktime_to_timespec64_cond(shhwtstamps->hwtstamp,
882 					     tss.ts + 2)) {
883 			empty = 0;
884 
885 			if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
886 			    !skb_is_err_queue(skb))
887 				put_ts_pktinfo(msg, skb);
888 		}
889 	}
890 	if (!empty) {
891 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
892 			put_cmsg_scm_timestamping64(msg, &tss);
893 		else
894 			put_cmsg_scm_timestamping(msg, &tss);
895 
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);
900 	}
901 }
902 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
903 
904 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
905 	struct sk_buff *skb)
906 {
907 	int ack;
908 
909 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
910 		return;
911 	if (!skb->wifi_acked_valid)
912 		return;
913 
914 	ack = skb->wifi_acked;
915 
916 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
917 }
918 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
919 
920 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
921 				   struct sk_buff *skb)
922 {
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);
926 }
927 
928 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
929 	struct sk_buff *skb)
930 {
931 	sock_recv_timestamp(msg, sk, skb);
932 	sock_recv_drops(msg, sk, skb);
933 }
934 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
935 
936 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
937 					   size_t, int));
938 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
939 					    size_t, int));
940 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
941 				     int flags)
942 {
943 	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
944 				  inet_recvmsg, sock, msg, msg_data_left(msg),
945 				  flags);
946 }
947 
948 /**
949  *	sock_recvmsg - receive a message from @sock
950  *	@sock: socket
951  *	@msg: message to receive
952  *	@flags: message flags
953  *
954  *	Receives @msg from @sock, passing through LSM. Returns the total number
955  *	of bytes received, or an error.
956  */
957 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
958 {
959 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
960 
961 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
962 }
963 EXPORT_SYMBOL(sock_recvmsg);
964 
965 /**
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...)
973  *
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.
977  *
978  *	The returned value is the total number of bytes received, or an error.
979  */
980 
981 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
982 		   struct kvec *vec, size_t num, size_t size, int flags)
983 {
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);
987 }
988 EXPORT_SYMBOL(kernel_recvmsg);
989 
990 static ssize_t sock_sendpage(struct file *file, struct page *page,
991 			     int offset, size_t size, loff_t *ppos, int more)
992 {
993 	struct socket *sock;
994 	int flags;
995 
996 	sock = file->private_data;
997 
998 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
999 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1000 	flags |= more;
1001 
1002 	return kernel_sendpage(sock, page, offset, size, flags);
1003 }
1004 
1005 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1006 				struct pipe_inode_info *pipe, size_t len,
1007 				unsigned int flags)
1008 {
1009 	struct socket *sock = file->private_data;
1010 
1011 	if (unlikely(!sock->ops->splice_read))
1012 		return generic_file_splice_read(file, ppos, pipe, len, flags);
1013 
1014 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1015 }
1016 
1017 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1018 {
1019 	struct file *file = iocb->ki_filp;
1020 	struct socket *sock = file->private_data;
1021 	struct msghdr msg = {.msg_iter = *to,
1022 			     .msg_iocb = iocb};
1023 	ssize_t res;
1024 
1025 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1026 		msg.msg_flags = MSG_DONTWAIT;
1027 
1028 	if (iocb->ki_pos != 0)
1029 		return -ESPIPE;
1030 
1031 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1032 		return 0;
1033 
1034 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1035 	*to = msg.msg_iter;
1036 	return res;
1037 }
1038 
1039 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1040 {
1041 	struct file *file = iocb->ki_filp;
1042 	struct socket *sock = file->private_data;
1043 	struct msghdr msg = {.msg_iter = *from,
1044 			     .msg_iocb = iocb};
1045 	ssize_t res;
1046 
1047 	if (iocb->ki_pos != 0)
1048 		return -ESPIPE;
1049 
1050 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1051 		msg.msg_flags = MSG_DONTWAIT;
1052 
1053 	if (sock->type == SOCK_SEQPACKET)
1054 		msg.msg_flags |= MSG_EOR;
1055 
1056 	res = sock_sendmsg(sock, &msg);
1057 	*from = msg.msg_iter;
1058 	return res;
1059 }
1060 
1061 /*
1062  * Atomic setting of ioctl hooks to avoid race
1063  * with module unload.
1064  */
1065 
1066 static DEFINE_MUTEX(br_ioctl_mutex);
1067 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1068 
1069 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1070 {
1071 	mutex_lock(&br_ioctl_mutex);
1072 	br_ioctl_hook = hook;
1073 	mutex_unlock(&br_ioctl_mutex);
1074 }
1075 EXPORT_SYMBOL(brioctl_set);
1076 
1077 static DEFINE_MUTEX(vlan_ioctl_mutex);
1078 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1079 
1080 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1081 {
1082 	mutex_lock(&vlan_ioctl_mutex);
1083 	vlan_ioctl_hook = hook;
1084 	mutex_unlock(&vlan_ioctl_mutex);
1085 }
1086 EXPORT_SYMBOL(vlan_ioctl_set);
1087 
1088 static long sock_do_ioctl(struct net *net, struct socket *sock,
1089 			  unsigned int cmd, unsigned long arg)
1090 {
1091 	int err;
1092 	void __user *argp = (void __user *)arg;
1093 
1094 	err = sock->ops->ioctl(sock, cmd, arg);
1095 
1096 	/*
1097 	 * If this ioctl is unknown try to hand it down
1098 	 * to the NIC driver.
1099 	 */
1100 	if (err != -ENOIOCTLCMD)
1101 		return err;
1102 
1103 	if (cmd == SIOCGIFCONF) {
1104 		struct ifconf ifc;
1105 		if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1106 			return -EFAULT;
1107 		rtnl_lock();
1108 		err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1109 		rtnl_unlock();
1110 		if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1111 			err = -EFAULT;
1112 	} else {
1113 		struct ifreq ifr;
1114 		bool need_copyout;
1115 		if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1116 			return -EFAULT;
1117 		err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1118 		if (!err && need_copyout)
1119 			if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1120 				return -EFAULT;
1121 	}
1122 	return err;
1123 }
1124 
1125 /*
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.
1128  */
1129 
1130 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1131 {
1132 	struct socket *sock;
1133 	struct sock *sk;
1134 	void __user *argp = (void __user *)arg;
1135 	int pid, err;
1136 	struct net *net;
1137 
1138 	sock = file->private_data;
1139 	sk = sock->sk;
1140 	net = sock_net(sk);
1141 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1142 		struct ifreq ifr;
1143 		bool need_copyout;
1144 		if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1145 			return -EFAULT;
1146 		err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1147 		if (!err && need_copyout)
1148 			if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1149 				return -EFAULT;
1150 	} else
1151 #ifdef CONFIG_WEXT_CORE
1152 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1153 		err = wext_handle_ioctl(net, cmd, argp);
1154 	} else
1155 #endif
1156 		switch (cmd) {
1157 		case FIOSETOWN:
1158 		case SIOCSPGRP:
1159 			err = -EFAULT;
1160 			if (get_user(pid, (int __user *)argp))
1161 				break;
1162 			err = f_setown(sock->file, pid, 1);
1163 			break;
1164 		case FIOGETOWN:
1165 		case SIOCGPGRP:
1166 			err = put_user(f_getown(sock->file),
1167 				       (int __user *)argp);
1168 			break;
1169 		case SIOCGIFBR:
1170 		case SIOCSIFBR:
1171 		case SIOCBRADDBR:
1172 		case SIOCBRDELBR:
1173 			err = -ENOPKG;
1174 			if (!br_ioctl_hook)
1175 				request_module("bridge");
1176 
1177 			mutex_lock(&br_ioctl_mutex);
1178 			if (br_ioctl_hook)
1179 				err = br_ioctl_hook(net, cmd, argp);
1180 			mutex_unlock(&br_ioctl_mutex);
1181 			break;
1182 		case SIOCGIFVLAN:
1183 		case SIOCSIFVLAN:
1184 			err = -ENOPKG;
1185 			if (!vlan_ioctl_hook)
1186 				request_module("8021q");
1187 
1188 			mutex_lock(&vlan_ioctl_mutex);
1189 			if (vlan_ioctl_hook)
1190 				err = vlan_ioctl_hook(net, argp);
1191 			mutex_unlock(&vlan_ioctl_mutex);
1192 			break;
1193 		case SIOCGSKNS:
1194 			err = -EPERM;
1195 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1196 				break;
1197 
1198 			err = open_related_ns(&net->ns, get_net_ns);
1199 			break;
1200 		case SIOCGSTAMP_OLD:
1201 		case SIOCGSTAMPNS_OLD:
1202 			if (!sock->ops->gettstamp) {
1203 				err = -ENOIOCTLCMD;
1204 				break;
1205 			}
1206 			err = sock->ops->gettstamp(sock, argp,
1207 						   cmd == SIOCGSTAMP_OLD,
1208 						   !IS_ENABLED(CONFIG_64BIT));
1209 			break;
1210 		case SIOCGSTAMP_NEW:
1211 		case SIOCGSTAMPNS_NEW:
1212 			if (!sock->ops->gettstamp) {
1213 				err = -ENOIOCTLCMD;
1214 				break;
1215 			}
1216 			err = sock->ops->gettstamp(sock, argp,
1217 						   cmd == SIOCGSTAMP_NEW,
1218 						   false);
1219 			break;
1220 		default:
1221 			err = sock_do_ioctl(net, sock, cmd, arg);
1222 			break;
1223 		}
1224 	return err;
1225 }
1226 
1227 /**
1228  *	sock_create_lite - creates a socket
1229  *	@family: protocol family (AF_INET, ...)
1230  *	@type: communication type (SOCK_STREAM, ...)
1231  *	@protocol: protocol (0, ...)
1232  *	@res: new socket
1233  *
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.
1238  */
1239 
1240 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1241 {
1242 	int err;
1243 	struct socket *sock = NULL;
1244 
1245 	err = security_socket_create(family, type, protocol, 1);
1246 	if (err)
1247 		goto out;
1248 
1249 	sock = sock_alloc();
1250 	if (!sock) {
1251 		err = -ENOMEM;
1252 		goto out;
1253 	}
1254 
1255 	sock->type = type;
1256 	err = security_socket_post_create(sock, family, type, protocol, 1);
1257 	if (err)
1258 		goto out_release;
1259 
1260 out:
1261 	*res = sock;
1262 	return err;
1263 out_release:
1264 	sock_release(sock);
1265 	sock = NULL;
1266 	goto out;
1267 }
1268 EXPORT_SYMBOL(sock_create_lite);
1269 
1270 /* No kernel lock held - perfect */
1271 static __poll_t sock_poll(struct file *file, poll_table *wait)
1272 {
1273 	struct socket *sock = file->private_data;
1274 	__poll_t events = poll_requested_events(wait), flag = 0;
1275 
1276 	if (!sock->ops->poll)
1277 		return 0;
1278 
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);
1283 
1284 		/* if this socket can poll_ll, tell the system call */
1285 		flag = POLL_BUSY_LOOP;
1286 	}
1287 
1288 	return sock->ops->poll(file, sock, wait) | flag;
1289 }
1290 
1291 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1292 {
1293 	struct socket *sock = file->private_data;
1294 
1295 	return sock->ops->mmap(file, sock, vma);
1296 }
1297 
1298 static int sock_close(struct inode *inode, struct file *filp)
1299 {
1300 	__sock_release(SOCKET_I(inode), inode);
1301 	return 0;
1302 }
1303 
1304 /*
1305  *	Update the socket async list
1306  *
1307  *	Fasync_list locking strategy.
1308  *
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
1313  */
1314 
1315 static int sock_fasync(int fd, struct file *filp, int on)
1316 {
1317 	struct socket *sock = filp->private_data;
1318 	struct sock *sk = sock->sk;
1319 	struct socket_wq *wq = &sock->wq;
1320 
1321 	if (sk == NULL)
1322 		return -EINVAL;
1323 
1324 	lock_sock(sk);
1325 	fasync_helper(fd, filp, on, &wq->fasync_list);
1326 
1327 	if (!wq->fasync_list)
1328 		sock_reset_flag(sk, SOCK_FASYNC);
1329 	else
1330 		sock_set_flag(sk, SOCK_FASYNC);
1331 
1332 	release_sock(sk);
1333 	return 0;
1334 }
1335 
1336 /* This function may be called only under rcu_lock */
1337 
1338 int sock_wake_async(struct socket_wq *wq, int how, int band)
1339 {
1340 	if (!wq || !wq->fasync_list)
1341 		return -1;
1342 
1343 	switch (how) {
1344 	case SOCK_WAKE_WAITD:
1345 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1346 			break;
1347 		goto call_kill;
1348 	case SOCK_WAKE_SPACE:
1349 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1350 			break;
1351 		fallthrough;
1352 	case SOCK_WAKE_IO:
1353 call_kill:
1354 		kill_fasync(&wq->fasync_list, SIGIO, band);
1355 		break;
1356 	case SOCK_WAKE_URG:
1357 		kill_fasync(&wq->fasync_list, SIGURG, band);
1358 	}
1359 
1360 	return 0;
1361 }
1362 EXPORT_SYMBOL(sock_wake_async);
1363 
1364 /**
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, ...)
1370  *	@res: new socket
1371  *	@kern: boolean for kernel space sockets
1372  *
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.
1377  */
1378 
1379 int __sock_create(struct net *net, int family, int type, int protocol,
1380 			 struct socket **res, int kern)
1381 {
1382 	int err;
1383 	struct socket *sock;
1384 	const struct net_proto_family *pf;
1385 
1386 	/*
1387 	 *      Check protocol is in range
1388 	 */
1389 	if (family < 0 || family >= NPROTO)
1390 		return -EAFNOSUPPORT;
1391 	if (type < 0 || type >= SOCK_MAX)
1392 		return -EINVAL;
1393 
1394 	/* Compatibility.
1395 
1396 	   This uglymoron is moved from INET layer to here to avoid
1397 	   deadlock in module load.
1398 	 */
1399 	if (family == PF_INET && type == SOCK_PACKET) {
1400 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1401 			     current->comm);
1402 		family = PF_PACKET;
1403 	}
1404 
1405 	err = security_socket_create(family, type, protocol, kern);
1406 	if (err)
1407 		return err;
1408 
1409 	/*
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
1412 	 *	default.
1413 	 */
1414 	sock = sock_alloc();
1415 	if (!sock) {
1416 		net_warn_ratelimited("socket: no more sockets\n");
1417 		return -ENFILE;	/* Not exactly a match, but its the
1418 				   closest posix thing */
1419 	}
1420 
1421 	sock->type = type;
1422 
1423 #ifdef CONFIG_MODULES
1424 	/* Attempt to load a protocol module if the find failed.
1425 	 *
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!
1429 	 */
1430 	if (rcu_access_pointer(net_families[family]) == NULL)
1431 		request_module("net-pf-%d", family);
1432 #endif
1433 
1434 	rcu_read_lock();
1435 	pf = rcu_dereference(net_families[family]);
1436 	err = -EAFNOSUPPORT;
1437 	if (!pf)
1438 		goto out_release;
1439 
1440 	/*
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.
1443 	 */
1444 	if (!try_module_get(pf->owner))
1445 		goto out_release;
1446 
1447 	/* Now protected by module ref count */
1448 	rcu_read_unlock();
1449 
1450 	err = pf->create(net, sock, protocol, kern);
1451 	if (err < 0)
1452 		goto out_module_put;
1453 
1454 	/*
1455 	 * Now to bump the refcnt of the [loadable] module that owns this
1456 	 * socket at sock_release time we decrement its refcnt.
1457 	 */
1458 	if (!try_module_get(sock->ops->owner))
1459 		goto out_module_busy;
1460 
1461 	/*
1462 	 * Now that we're done with the ->create function, the [loadable]
1463 	 * module can have its refcnt decremented
1464 	 */
1465 	module_put(pf->owner);
1466 	err = security_socket_post_create(sock, family, type, protocol, kern);
1467 	if (err)
1468 		goto out_sock_release;
1469 	*res = sock;
1470 
1471 	return 0;
1472 
1473 out_module_busy:
1474 	err = -EAFNOSUPPORT;
1475 out_module_put:
1476 	sock->ops = NULL;
1477 	module_put(pf->owner);
1478 out_sock_release:
1479 	sock_release(sock);
1480 	return err;
1481 
1482 out_release:
1483 	rcu_read_unlock();
1484 	goto out_sock_release;
1485 }
1486 EXPORT_SYMBOL(__sock_create);
1487 
1488 /**
1489  *	sock_create - creates a socket
1490  *	@family: protocol family (AF_INET, ...)
1491  *	@type: communication type (SOCK_STREAM, ...)
1492  *	@protocol: protocol (0, ...)
1493  *	@res: new socket
1494  *
1495  *	A wrapper around __sock_create().
1496  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1497  */
1498 
1499 int sock_create(int family, int type, int protocol, struct socket **res)
1500 {
1501 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1502 }
1503 EXPORT_SYMBOL(sock_create);
1504 
1505 /**
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, ...)
1511  *	@res: new socket
1512  *
1513  *	A wrapper around __sock_create().
1514  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1515  */
1516 
1517 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1518 {
1519 	return __sock_create(net, family, type, protocol, res, 1);
1520 }
1521 EXPORT_SYMBOL(sock_create_kern);
1522 
1523 int __sys_socket(int family, int type, int protocol)
1524 {
1525 	int retval;
1526 	struct socket *sock;
1527 	int flags;
1528 
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);
1534 
1535 	flags = type & ~SOCK_TYPE_MASK;
1536 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1537 		return -EINVAL;
1538 	type &= SOCK_TYPE_MASK;
1539 
1540 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1541 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1542 
1543 	retval = sock_create(family, type, protocol, &sock);
1544 	if (retval < 0)
1545 		return retval;
1546 
1547 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1548 }
1549 
1550 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1551 {
1552 	return __sys_socket(family, type, protocol);
1553 }
1554 
1555 /*
1556  *	Create a pair of connected sockets.
1557  */
1558 
1559 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1560 {
1561 	struct socket *sock1, *sock2;
1562 	int fd1, fd2, err;
1563 	struct file *newfile1, *newfile2;
1564 	int flags;
1565 
1566 	flags = type & ~SOCK_TYPE_MASK;
1567 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1568 		return -EINVAL;
1569 	type &= SOCK_TYPE_MASK;
1570 
1571 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1572 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1573 
1574 	/*
1575 	 * reserve descriptors and make sure we won't fail
1576 	 * to return them to userland.
1577 	 */
1578 	fd1 = get_unused_fd_flags(flags);
1579 	if (unlikely(fd1 < 0))
1580 		return fd1;
1581 
1582 	fd2 = get_unused_fd_flags(flags);
1583 	if (unlikely(fd2 < 0)) {
1584 		put_unused_fd(fd1);
1585 		return fd2;
1586 	}
1587 
1588 	err = put_user(fd1, &usockvec[0]);
1589 	if (err)
1590 		goto out;
1591 
1592 	err = put_user(fd2, &usockvec[1]);
1593 	if (err)
1594 		goto out;
1595 
1596 	/*
1597 	 * Obtain the first socket and check if the underlying protocol
1598 	 * supports the socketpair call.
1599 	 */
1600 
1601 	err = sock_create(family, type, protocol, &sock1);
1602 	if (unlikely(err < 0))
1603 		goto out;
1604 
1605 	err = sock_create(family, type, protocol, &sock2);
1606 	if (unlikely(err < 0)) {
1607 		sock_release(sock1);
1608 		goto out;
1609 	}
1610 
1611 	err = security_socket_socketpair(sock1, sock2);
1612 	if (unlikely(err)) {
1613 		sock_release(sock2);
1614 		sock_release(sock1);
1615 		goto out;
1616 	}
1617 
1618 	err = sock1->ops->socketpair(sock1, sock2);
1619 	if (unlikely(err < 0)) {
1620 		sock_release(sock2);
1621 		sock_release(sock1);
1622 		goto out;
1623 	}
1624 
1625 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1626 	if (IS_ERR(newfile1)) {
1627 		err = PTR_ERR(newfile1);
1628 		sock_release(sock2);
1629 		goto out;
1630 	}
1631 
1632 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1633 	if (IS_ERR(newfile2)) {
1634 		err = PTR_ERR(newfile2);
1635 		fput(newfile1);
1636 		goto out;
1637 	}
1638 
1639 	audit_fd_pair(fd1, fd2);
1640 
1641 	fd_install(fd1, newfile1);
1642 	fd_install(fd2, newfile2);
1643 	return 0;
1644 
1645 out:
1646 	put_unused_fd(fd2);
1647 	put_unused_fd(fd1);
1648 	return err;
1649 }
1650 
1651 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1652 		int __user *, usockvec)
1653 {
1654 	return __sys_socketpair(family, type, protocol, usockvec);
1655 }
1656 
1657 /*
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.
1660  *
1661  *	We move the socket address to kernel space before we call
1662  *	the protocol layer (having also checked the address is ok).
1663  */
1664 
1665 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1666 {
1667 	struct socket *sock;
1668 	struct sockaddr_storage address;
1669 	int err, fput_needed;
1670 
1671 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672 	if (sock) {
1673 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1674 		if (!err) {
1675 			err = security_socket_bind(sock,
1676 						   (struct sockaddr *)&address,
1677 						   addrlen);
1678 			if (!err)
1679 				err = sock->ops->bind(sock,
1680 						      (struct sockaddr *)
1681 						      &address, addrlen);
1682 		}
1683 		fput_light(sock->file, fput_needed);
1684 	}
1685 	return err;
1686 }
1687 
1688 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1689 {
1690 	return __sys_bind(fd, umyaddr, addrlen);
1691 }
1692 
1693 /*
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.
1697  */
1698 
1699 int __sys_listen(int fd, int backlog)
1700 {
1701 	struct socket *sock;
1702 	int err, fput_needed;
1703 	int somaxconn;
1704 
1705 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706 	if (sock) {
1707 		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1708 		if ((unsigned int)backlog > somaxconn)
1709 			backlog = somaxconn;
1710 
1711 		err = security_socket_listen(sock, backlog);
1712 		if (!err)
1713 			err = sock->ops->listen(sock, backlog);
1714 
1715 		fput_light(sock->file, fput_needed);
1716 	}
1717 	return err;
1718 }
1719 
1720 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1721 {
1722 	return __sys_listen(fd, backlog);
1723 }
1724 
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)
1729 {
1730 	struct socket *sock, *newsock;
1731 	struct file *newfile;
1732 	int err, len, newfd;
1733 	struct sockaddr_storage address;
1734 
1735 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1736 		return -EINVAL;
1737 
1738 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1739 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1740 
1741 	sock = sock_from_file(file);
1742 	if (!sock) {
1743 		err = -ENOTSOCK;
1744 		goto out;
1745 	}
1746 
1747 	err = -ENFILE;
1748 	newsock = sock_alloc();
1749 	if (!newsock)
1750 		goto out;
1751 
1752 	newsock->type = sock->type;
1753 	newsock->ops = sock->ops;
1754 
1755 	/*
1756 	 * We don't need try_module_get here, as the listening socket (sock)
1757 	 * has the protocol module (sock->ops->owner) held.
1758 	 */
1759 	__module_get(newsock->ops->owner);
1760 
1761 	newfd = __get_unused_fd_flags(flags, nofile);
1762 	if (unlikely(newfd < 0)) {
1763 		err = newfd;
1764 		sock_release(newsock);
1765 		goto out;
1766 	}
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);
1771 		goto out;
1772 	}
1773 
1774 	err = security_socket_accept(sock, newsock);
1775 	if (err)
1776 		goto out_fd;
1777 
1778 	err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1779 					false);
1780 	if (err < 0)
1781 		goto out_fd;
1782 
1783 	if (upeer_sockaddr) {
1784 		len = newsock->ops->getname(newsock,
1785 					(struct sockaddr *)&address, 2);
1786 		if (len < 0) {
1787 			err = -ECONNABORTED;
1788 			goto out_fd;
1789 		}
1790 		err = move_addr_to_user(&address,
1791 					len, upeer_sockaddr, upeer_addrlen);
1792 		if (err < 0)
1793 			goto out_fd;
1794 	}
1795 
1796 	/* File flags are not inherited via accept() unlike another OSes. */
1797 
1798 	fd_install(newfd, newfile);
1799 	err = newfd;
1800 out:
1801 	return err;
1802 out_fd:
1803 	fput(newfile);
1804 	put_unused_fd(newfd);
1805 	goto out;
1806 
1807 }
1808 
1809 /*
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.
1815  *
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.
1819  */
1820 
1821 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1822 		  int __user *upeer_addrlen, int flags)
1823 {
1824 	int ret = -EBADF;
1825 	struct fd f;
1826 
1827 	f = fdget(fd);
1828 	if (f.file) {
1829 		ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1830 						upeer_addrlen, flags,
1831 						rlimit(RLIMIT_NOFILE));
1832 		fdput(f);
1833 	}
1834 
1835 	return ret;
1836 }
1837 
1838 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1839 		int __user *, upeer_addrlen, int, flags)
1840 {
1841 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1842 }
1843 
1844 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1845 		int __user *, upeer_addrlen)
1846 {
1847 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1848 }
1849 
1850 /*
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.
1853  *
1854  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1855  *	break bindings
1856  *
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.
1860  */
1861 
1862 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1863 		       int addrlen, int file_flags)
1864 {
1865 	struct socket *sock;
1866 	int err;
1867 
1868 	sock = sock_from_file(file);
1869 	if (!sock) {
1870 		err = -ENOTSOCK;
1871 		goto out;
1872 	}
1873 
1874 	err =
1875 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1876 	if (err)
1877 		goto out;
1878 
1879 	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1880 				 sock->file->f_flags | file_flags);
1881 out:
1882 	return err;
1883 }
1884 
1885 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1886 {
1887 	int ret = -EBADF;
1888 	struct fd f;
1889 
1890 	f = fdget(fd);
1891 	if (f.file) {
1892 		struct sockaddr_storage address;
1893 
1894 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1895 		if (!ret)
1896 			ret = __sys_connect_file(f.file, &address, addrlen, 0);
1897 		fdput(f);
1898 	}
1899 
1900 	return ret;
1901 }
1902 
1903 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1904 		int, addrlen)
1905 {
1906 	return __sys_connect(fd, uservaddr, addrlen);
1907 }
1908 
1909 /*
1910  *	Get the local address ('name') of a socket object. Move the obtained
1911  *	name to user space.
1912  */
1913 
1914 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1915 		      int __user *usockaddr_len)
1916 {
1917 	struct socket *sock;
1918 	struct sockaddr_storage address;
1919 	int err, fput_needed;
1920 
1921 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1922 	if (!sock)
1923 		goto out;
1924 
1925 	err = security_socket_getsockname(sock);
1926 	if (err)
1927 		goto out_put;
1928 
1929 	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1930 	if (err < 0)
1931 		goto out_put;
1932         /* "err" is actually length in this case */
1933 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1934 
1935 out_put:
1936 	fput_light(sock->file, fput_needed);
1937 out:
1938 	return err;
1939 }
1940 
1941 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1942 		int __user *, usockaddr_len)
1943 {
1944 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1945 }
1946 
1947 /*
1948  *	Get the remote address ('name') of a socket object. Move the obtained
1949  *	name to user space.
1950  */
1951 
1952 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1953 		      int __user *usockaddr_len)
1954 {
1955 	struct socket *sock;
1956 	struct sockaddr_storage address;
1957 	int err, fput_needed;
1958 
1959 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1960 	if (sock != NULL) {
1961 		err = security_socket_getpeername(sock);
1962 		if (err) {
1963 			fput_light(sock->file, fput_needed);
1964 			return err;
1965 		}
1966 
1967 		err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1968 		if (err >= 0)
1969 			/* "err" is actually length in this case */
1970 			err = move_addr_to_user(&address, err, usockaddr,
1971 						usockaddr_len);
1972 		fput_light(sock->file, fput_needed);
1973 	}
1974 	return err;
1975 }
1976 
1977 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1978 		int __user *, usockaddr_len)
1979 {
1980 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1981 }
1982 
1983 /*
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
1986  *	the protocol.
1987  */
1988 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1989 		 struct sockaddr __user *addr,  int addr_len)
1990 {
1991 	struct socket *sock;
1992 	struct sockaddr_storage address;
1993 	int err;
1994 	struct msghdr msg;
1995 	struct iovec iov;
1996 	int fput_needed;
1997 
1998 	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1999 	if (unlikely(err))
2000 		return err;
2001 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2002 	if (!sock)
2003 		goto out;
2004 
2005 	msg.msg_name = NULL;
2006 	msg.msg_control = NULL;
2007 	msg.msg_controllen = 0;
2008 	msg.msg_namelen = 0;
2009 	if (addr) {
2010 		err = move_addr_to_kernel(addr, addr_len, &address);
2011 		if (err < 0)
2012 			goto out_put;
2013 		msg.msg_name = (struct sockaddr *)&address;
2014 		msg.msg_namelen = addr_len;
2015 	}
2016 	if (sock->file->f_flags & O_NONBLOCK)
2017 		flags |= MSG_DONTWAIT;
2018 	msg.msg_flags = flags;
2019 	err = sock_sendmsg(sock, &msg);
2020 
2021 out_put:
2022 	fput_light(sock->file, fput_needed);
2023 out:
2024 	return err;
2025 }
2026 
2027 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2028 		unsigned int, flags, struct sockaddr __user *, addr,
2029 		int, addr_len)
2030 {
2031 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2032 }
2033 
2034 /*
2035  *	Send a datagram down a socket.
2036  */
2037 
2038 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2039 		unsigned int, flags)
2040 {
2041 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2042 }
2043 
2044 /*
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.
2048  */
2049 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2050 		   struct sockaddr __user *addr, int __user *addr_len)
2051 {
2052 	struct socket *sock;
2053 	struct iovec iov;
2054 	struct msghdr msg;
2055 	struct sockaddr_storage address;
2056 	int err, err2;
2057 	int fput_needed;
2058 
2059 	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2060 	if (unlikely(err))
2061 		return err;
2062 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2063 	if (!sock)
2064 		goto out;
2065 
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;
2073 	msg.msg_flags = 0;
2074 	if (sock->file->f_flags & O_NONBLOCK)
2075 		flags |= MSG_DONTWAIT;
2076 	err = sock_recvmsg(sock, &msg, flags);
2077 
2078 	if (err >= 0 && addr != NULL) {
2079 		err2 = move_addr_to_user(&address,
2080 					 msg.msg_namelen, addr, addr_len);
2081 		if (err2 < 0)
2082 			err = err2;
2083 	}
2084 
2085 	fput_light(sock->file, fput_needed);
2086 out:
2087 	return err;
2088 }
2089 
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)
2093 {
2094 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2095 }
2096 
2097 /*
2098  *	Receive a datagram from a socket.
2099  */
2100 
2101 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2102 		unsigned int, flags)
2103 {
2104 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2105 }
2106 
2107 static bool sock_use_custom_sol_socket(const struct socket *sock)
2108 {
2109 	const struct sock *sk = sock->sk;
2110 
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);
2116 }
2117 
2118 /*
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.
2121  */
2122 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2123 		int optlen)
2124 {
2125 	sockptr_t optval = USER_SOCKPTR(user_optval);
2126 	char *kernel_optval = NULL;
2127 	int err, fput_needed;
2128 	struct socket *sock;
2129 
2130 	if (optlen < 0)
2131 		return -EINVAL;
2132 
2133 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2134 	if (!sock)
2135 		return err;
2136 
2137 	err = security_socket_setsockopt(sock, level, optname);
2138 	if (err)
2139 		goto out_put;
2140 
2141 	if (!in_compat_syscall())
2142 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2143 						     user_optval, &optlen,
2144 						     &kernel_optval);
2145 	if (err < 0)
2146 		goto out_put;
2147 	if (err > 0) {
2148 		err = 0;
2149 		goto out_put;
2150 	}
2151 
2152 	if (kernel_optval)
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))
2157 		err = -EOPNOTSUPP;
2158 	else
2159 		err = sock->ops->setsockopt(sock, level, optname, optval,
2160 					    optlen);
2161 	kfree(kernel_optval);
2162 out_put:
2163 	fput_light(sock->file, fput_needed);
2164 	return err;
2165 }
2166 
2167 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2168 		char __user *, optval, int, optlen)
2169 {
2170 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2171 }
2172 
2173 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2174 							 int optname));
2175 
2176 /*
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.
2179  */
2180 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2181 		int __user *optlen)
2182 {
2183 	int err, fput_needed;
2184 	struct socket *sock;
2185 	int max_optlen;
2186 
2187 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2188 	if (!sock)
2189 		return err;
2190 
2191 	err = security_socket_getsockopt(sock, level, optname);
2192 	if (err)
2193 		goto out_put;
2194 
2195 	if (!in_compat_syscall())
2196 		max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2197 
2198 	if (level == SOL_SOCKET)
2199 		err = sock_getsockopt(sock, level, optname, optval, optlen);
2200 	else if (unlikely(!sock->ops->getsockopt))
2201 		err = -EOPNOTSUPP;
2202 	else
2203 		err = sock->ops->getsockopt(sock, level, optname, optval,
2204 					    optlen);
2205 
2206 	if (!in_compat_syscall())
2207 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2208 						     optval, optlen, max_optlen,
2209 						     err);
2210 out_put:
2211 	fput_light(sock->file, fput_needed);
2212 	return err;
2213 }
2214 
2215 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2216 		char __user *, optval, int __user *, optlen)
2217 {
2218 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2219 }
2220 
2221 /*
2222  *	Shutdown a socket.
2223  */
2224 
2225 int __sys_shutdown_sock(struct socket *sock, int how)
2226 {
2227 	int err;
2228 
2229 	err = security_socket_shutdown(sock, how);
2230 	if (!err)
2231 		err = sock->ops->shutdown(sock, how);
2232 
2233 	return err;
2234 }
2235 
2236 int __sys_shutdown(int fd, int how)
2237 {
2238 	int err, fput_needed;
2239 	struct socket *sock;
2240 
2241 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2242 	if (sock != NULL) {
2243 		err = __sys_shutdown_sock(sock, how);
2244 		fput_light(sock->file, fput_needed);
2245 	}
2246 	return err;
2247 }
2248 
2249 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2250 {
2251 	return __sys_shutdown(fd, how);
2252 }
2253 
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.
2256  */
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)
2260 
2261 struct used_address {
2262 	struct sockaddr_storage name;
2263 	unsigned int name_len;
2264 };
2265 
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)
2270 {
2271 	struct user_msghdr msg;
2272 	ssize_t err;
2273 
2274 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2275 		return -EFAULT;
2276 
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;
2281 
2282 	kmsg->msg_namelen = msg.msg_namelen;
2283 	if (!msg.msg_name)
2284 		kmsg->msg_namelen = 0;
2285 
2286 	if (kmsg->msg_namelen < 0)
2287 		return -EINVAL;
2288 
2289 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2290 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2291 
2292 	if (save_addr)
2293 		*save_addr = msg.msg_name;
2294 
2295 	if (msg.msg_name && kmsg->msg_namelen) {
2296 		if (!save_addr) {
2297 			err = move_addr_to_kernel(msg.msg_name,
2298 						  kmsg->msg_namelen,
2299 						  kmsg->msg_name);
2300 			if (err < 0)
2301 				return err;
2302 		}
2303 	} else {
2304 		kmsg->msg_name = NULL;
2305 		kmsg->msg_namelen = 0;
2306 	}
2307 
2308 	if (msg.msg_iovlen > UIO_MAXIOV)
2309 		return -EMSGSIZE;
2310 
2311 	kmsg->msg_iocb = NULL;
2312 	*uiov = msg.msg_iov;
2313 	*nsegs = msg.msg_iovlen;
2314 	return 0;
2315 }
2316 
2317 static int copy_msghdr_from_user(struct msghdr *kmsg,
2318 				 struct user_msghdr __user *umsg,
2319 				 struct sockaddr __user **save_addr,
2320 				 struct iovec **iov)
2321 {
2322 	struct user_msghdr msg;
2323 	ssize_t err;
2324 
2325 	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2326 					&msg.msg_iovlen);
2327 	if (err)
2328 		return err;
2329 
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;
2334 }
2335 
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)
2339 {
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;
2344 	int ctl_len;
2345 	ssize_t err;
2346 
2347 	err = -ENOBUFS;
2348 
2349 	if (msg_sys->msg_controllen > INT_MAX)
2350 		goto out;
2351 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2352 	ctl_len = msg_sys->msg_controllen;
2353 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2354 		err =
2355 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2356 						     sizeof(ctl));
2357 		if (err)
2358 			goto out;
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)
2367 				goto out;
2368 		}
2369 		err = -EFAULT;
2370 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2371 			goto out_freectl;
2372 		msg_sys->msg_control = ctl_buf;
2373 		msg_sys->msg_control_is_user = false;
2374 	}
2375 	msg_sys->msg_flags = flags;
2376 
2377 	if (sock->file->f_flags & O_NONBLOCK)
2378 		msg_sys->msg_flags |= MSG_DONTWAIT;
2379 	/*
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.
2384 	 */
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);
2390 		goto out_freectl;
2391 	}
2392 	err = sock_sendmsg(sock, msg_sys);
2393 	/*
2394 	 * If this is sendmmsg() and sending to current destination address was
2395 	 * successful, remember it.
2396 	 */
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);
2402 	}
2403 
2404 out_freectl:
2405 	if (ctl_buf != ctl)
2406 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2407 out:
2408 	return err;
2409 }
2410 
2411 int sendmsg_copy_msghdr(struct msghdr *msg,
2412 			struct user_msghdr __user *umsg, unsigned flags,
2413 			struct iovec **iov)
2414 {
2415 	int err;
2416 
2417 	if (flags & MSG_CMSG_COMPAT) {
2418 		struct compat_msghdr __user *msg_compat;
2419 
2420 		msg_compat = (struct compat_msghdr __user *) umsg;
2421 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2422 	} else {
2423 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2424 	}
2425 	if (err < 0)
2426 		return err;
2427 
2428 	return 0;
2429 }
2430 
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)
2435 {
2436 	struct sockaddr_storage address;
2437 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2438 	ssize_t err;
2439 
2440 	msg_sys->msg_name = &address;
2441 
2442 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2443 	if (err < 0)
2444 		return err;
2445 
2446 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2447 				allowed_msghdr_flags);
2448 	kfree(iov);
2449 	return err;
2450 }
2451 
2452 /*
2453  *	BSD sendmsg interface
2454  */
2455 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2456 			unsigned int flags)
2457 {
2458 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2459 }
2460 
2461 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2462 		   bool forbid_cmsg_compat)
2463 {
2464 	int fput_needed, err;
2465 	struct msghdr msg_sys;
2466 	struct socket *sock;
2467 
2468 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2469 		return -EINVAL;
2470 
2471 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2472 	if (!sock)
2473 		goto out;
2474 
2475 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2476 
2477 	fput_light(sock->file, fput_needed);
2478 out:
2479 	return err;
2480 }
2481 
2482 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2483 {
2484 	return __sys_sendmsg(fd, msg, flags, true);
2485 }
2486 
2487 /*
2488  *	Linux sendmmsg interface
2489  */
2490 
2491 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2492 		   unsigned int flags, bool forbid_cmsg_compat)
2493 {
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;
2501 
2502 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2503 		return -EINVAL;
2504 
2505 	if (vlen > UIO_MAXIOV)
2506 		vlen = UIO_MAXIOV;
2507 
2508 	datagrams = 0;
2509 
2510 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2511 	if (!sock)
2512 		return err;
2513 
2514 	used_address.name_len = UINT_MAX;
2515 	entry = mmsg;
2516 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2517 	err = 0;
2518 	flags |= MSG_BATCH;
2519 
2520 	while (datagrams < vlen) {
2521 		if (datagrams == vlen - 1)
2522 			flags = oflags;
2523 
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);
2527 			if (err < 0)
2528 				break;
2529 			err = __put_user(err, &compat_entry->msg_len);
2530 			++compat_entry;
2531 		} else {
2532 			err = ___sys_sendmsg(sock,
2533 					     (struct user_msghdr __user *)entry,
2534 					     &msg_sys, flags, &used_address, MSG_EOR);
2535 			if (err < 0)
2536 				break;
2537 			err = put_user(err, &entry->msg_len);
2538 			++entry;
2539 		}
2540 
2541 		if (err)
2542 			break;
2543 		++datagrams;
2544 		if (msg_data_left(&msg_sys))
2545 			break;
2546 		cond_resched();
2547 	}
2548 
2549 	fput_light(sock->file, fput_needed);
2550 
2551 	/* We only return an error if no datagrams were able to be sent */
2552 	if (datagrams != 0)
2553 		return datagrams;
2554 
2555 	return err;
2556 }
2557 
2558 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2559 		unsigned int, vlen, unsigned int, flags)
2560 {
2561 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2562 }
2563 
2564 int recvmsg_copy_msghdr(struct msghdr *msg,
2565 			struct user_msghdr __user *umsg, unsigned flags,
2566 			struct sockaddr __user **uaddr,
2567 			struct iovec **iov)
2568 {
2569 	ssize_t err;
2570 
2571 	if (MSG_CMSG_COMPAT & flags) {
2572 		struct compat_msghdr __user *msg_compat;
2573 
2574 		msg_compat = (struct compat_msghdr __user *) umsg;
2575 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2576 	} else {
2577 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2578 	}
2579 	if (err < 0)
2580 		return err;
2581 
2582 	return 0;
2583 }
2584 
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)
2589 {
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;
2595 	int len;
2596 	ssize_t err;
2597 
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);
2601 
2602 	/* We assume all kernel code knows the size of sockaddr_storage */
2603 	msg_sys->msg_namelen = 0;
2604 
2605 	if (sock->file->f_flags & O_NONBLOCK)
2606 		flags |= MSG_DONTWAIT;
2607 
2608 	if (unlikely(nosec))
2609 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2610 	else
2611 		err = sock_recvmsg(sock, msg_sys, flags);
2612 
2613 	if (err < 0)
2614 		goto out;
2615 	len = err;
2616 
2617 	if (uaddr != NULL) {
2618 		err = move_addr_to_user(&addr,
2619 					msg_sys->msg_namelen, uaddr,
2620 					uaddr_len);
2621 		if (err < 0)
2622 			goto out;
2623 	}
2624 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2625 			 COMPAT_FLAGS(msg));
2626 	if (err)
2627 		goto out;
2628 	if (MSG_CMSG_COMPAT & flags)
2629 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2630 				 &msg_compat->msg_controllen);
2631 	else
2632 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2633 				 &msg->msg_controllen);
2634 	if (err)
2635 		goto out;
2636 	err = len;
2637 out:
2638 	return err;
2639 }
2640 
2641 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2642 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2643 {
2644 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2645 	/* user mode address pointers */
2646 	struct sockaddr __user *uaddr;
2647 	ssize_t err;
2648 
2649 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2650 	if (err < 0)
2651 		return err;
2652 
2653 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2654 	kfree(iov);
2655 	return err;
2656 }
2657 
2658 /*
2659  *	BSD recvmsg interface
2660  */
2661 
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)
2665 {
2666 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2667 }
2668 
2669 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2670 		   bool forbid_cmsg_compat)
2671 {
2672 	int fput_needed, err;
2673 	struct msghdr msg_sys;
2674 	struct socket *sock;
2675 
2676 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2677 		return -EINVAL;
2678 
2679 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2680 	if (!sock)
2681 		goto out;
2682 
2683 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2684 
2685 	fput_light(sock->file, fput_needed);
2686 out:
2687 	return err;
2688 }
2689 
2690 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2691 		unsigned int, flags)
2692 {
2693 	return __sys_recvmsg(fd, msg, flags, true);
2694 }
2695 
2696 /*
2697  *     Linux recvmmsg interface
2698  */
2699 
2700 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2701 			  unsigned int vlen, unsigned int flags,
2702 			  struct timespec64 *timeout)
2703 {
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;
2711 
2712 	if (timeout &&
2713 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2714 				    timeout->tv_nsec))
2715 		return -EINVAL;
2716 
2717 	datagrams = 0;
2718 
2719 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2720 	if (!sock)
2721 		return err;
2722 
2723 	if (likely(!(flags & MSG_ERRQUEUE))) {
2724 		err = sock_error(sock->sk);
2725 		if (err) {
2726 			datagrams = err;
2727 			goto out_put;
2728 		}
2729 	}
2730 
2731 	entry = mmsg;
2732 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2733 
2734 	while (datagrams < vlen) {
2735 		/*
2736 		 * No need to ask LSM for more than the first datagram.
2737 		 */
2738 		if (MSG_CMSG_COMPAT & flags) {
2739 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2740 					     &msg_sys, flags & ~MSG_WAITFORONE,
2741 					     datagrams);
2742 			if (err < 0)
2743 				break;
2744 			err = __put_user(err, &compat_entry->msg_len);
2745 			++compat_entry;
2746 		} else {
2747 			err = ___sys_recvmsg(sock,
2748 					     (struct user_msghdr __user *)entry,
2749 					     &msg_sys, flags & ~MSG_WAITFORONE,
2750 					     datagrams);
2751 			if (err < 0)
2752 				break;
2753 			err = put_user(err, &entry->msg_len);
2754 			++entry;
2755 		}
2756 
2757 		if (err)
2758 			break;
2759 		++datagrams;
2760 
2761 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2762 		if (flags & MSG_WAITFORONE)
2763 			flags |= MSG_DONTWAIT;
2764 
2765 		if (timeout) {
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;
2770 				break;
2771 			}
2772 
2773 			/* Timeout, return less than vlen datagrams */
2774 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2775 				break;
2776 		}
2777 
2778 		/* Out of band data, return right away */
2779 		if (msg_sys.msg_flags & MSG_OOB)
2780 			break;
2781 		cond_resched();
2782 	}
2783 
2784 	if (err == 0)
2785 		goto out_put;
2786 
2787 	if (datagrams == 0) {
2788 		datagrams = err;
2789 		goto out_put;
2790 	}
2791 
2792 	/*
2793 	 * We may return less entries than requested (vlen) if the
2794 	 * sock is non block and there aren't enough datagrams...
2795 	 */
2796 	if (err != -EAGAIN) {
2797 		/*
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).
2802 		 */
2803 		sock->sk->sk_err = -err;
2804 	}
2805 out_put:
2806 	fput_light(sock->file, fput_needed);
2807 
2808 	return datagrams;
2809 }
2810 
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)
2815 {
2816 	int datagrams;
2817 	struct timespec64 timeout_sys;
2818 
2819 	if (timeout && get_timespec64(&timeout_sys, timeout))
2820 		return -EFAULT;
2821 
2822 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2823 		return -EFAULT;
2824 
2825 	if (!timeout && !timeout32)
2826 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2827 
2828 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2829 
2830 	if (datagrams <= 0)
2831 		return datagrams;
2832 
2833 	if (timeout && put_timespec64(&timeout_sys, timeout))
2834 		datagrams = -EFAULT;
2835 
2836 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2837 		datagrams = -EFAULT;
2838 
2839 	return datagrams;
2840 }
2841 
2842 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2843 		unsigned int, vlen, unsigned int, flags,
2844 		struct __kernel_timespec __user *, timeout)
2845 {
2846 	if (flags & MSG_CMSG_COMPAT)
2847 		return -EINVAL;
2848 
2849 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2850 }
2851 
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)
2856 {
2857 	if (flags & MSG_CMSG_COMPAT)
2858 		return -EINVAL;
2859 
2860 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2861 }
2862 #endif
2863 
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),
2871 	AL(4), AL(5), AL(4)
2872 };
2873 
2874 #undef AL
2875 
2876 /*
2877  *	System call vectors.
2878  *
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.
2882  */
2883 
2884 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2885 {
2886 	unsigned long a[AUDITSC_ARGS];
2887 	unsigned long a0, a1;
2888 	int err;
2889 	unsigned int len;
2890 
2891 	if (call < 1 || call > SYS_SENDMMSG)
2892 		return -EINVAL;
2893 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2894 
2895 	len = nargs[call];
2896 	if (len > sizeof(a))
2897 		return -EINVAL;
2898 
2899 	/* copy_from_user should be SMP safe. */
2900 	if (copy_from_user(a, args, len))
2901 		return -EFAULT;
2902 
2903 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2904 	if (err)
2905 		return err;
2906 
2907 	a0 = a[0];
2908 	a1 = a[1];
2909 
2910 	switch (call) {
2911 	case SYS_SOCKET:
2912 		err = __sys_socket(a0, a1, a[2]);
2913 		break;
2914 	case SYS_BIND:
2915 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2916 		break;
2917 	case SYS_CONNECT:
2918 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2919 		break;
2920 	case SYS_LISTEN:
2921 		err = __sys_listen(a0, a1);
2922 		break;
2923 	case SYS_ACCEPT:
2924 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2925 				    (int __user *)a[2], 0);
2926 		break;
2927 	case SYS_GETSOCKNAME:
2928 		err =
2929 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
2930 				      (int __user *)a[2]);
2931 		break;
2932 	case SYS_GETPEERNAME:
2933 		err =
2934 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
2935 				      (int __user *)a[2]);
2936 		break;
2937 	case SYS_SOCKETPAIR:
2938 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2939 		break;
2940 	case SYS_SEND:
2941 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2942 				   NULL, 0);
2943 		break;
2944 	case SYS_SENDTO:
2945 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2946 				   (struct sockaddr __user *)a[4], a[5]);
2947 		break;
2948 	case SYS_RECV:
2949 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2950 				     NULL, NULL);
2951 		break;
2952 	case SYS_RECVFROM:
2953 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2954 				     (struct sockaddr __user *)a[4],
2955 				     (int __user *)a[5]);
2956 		break;
2957 	case SYS_SHUTDOWN:
2958 		err = __sys_shutdown(a0, a1);
2959 		break;
2960 	case SYS_SETSOCKOPT:
2961 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2962 				       a[4]);
2963 		break;
2964 	case SYS_GETSOCKOPT:
2965 		err =
2966 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2967 				     (int __user *)a[4]);
2968 		break;
2969 	case SYS_SENDMSG:
2970 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2971 				    a[2], true);
2972 		break;
2973 	case SYS_SENDMMSG:
2974 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2975 				     a[3], true);
2976 		break;
2977 	case SYS_RECVMSG:
2978 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2979 				    a[2], true);
2980 		break;
2981 	case SYS_RECVMMSG:
2982 		if (IS_ENABLED(CONFIG_64BIT))
2983 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2984 					     a[2], a[3],
2985 					     (struct __kernel_timespec __user *)a[4],
2986 					     NULL);
2987 		else
2988 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2989 					     a[2], a[3], NULL,
2990 					     (struct old_timespec32 __user *)a[4]);
2991 		break;
2992 	case SYS_ACCEPT4:
2993 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2994 				    (int __user *)a[2], a[3]);
2995 		break;
2996 	default:
2997 		err = -EINVAL;
2998 		break;
2999 	}
3000 	return err;
3001 }
3002 
3003 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3004 
3005 /**
3006  *	sock_register - add a socket protocol handler
3007  *	@ops: description of protocol
3008  *
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.
3013  */
3014 int sock_register(const struct net_proto_family *ops)
3015 {
3016 	int err;
3017 
3018 	if (ops->family >= NPROTO) {
3019 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3020 		return -ENOBUFS;
3021 	}
3022 
3023 	spin_lock(&net_family_lock);
3024 	if (rcu_dereference_protected(net_families[ops->family],
3025 				      lockdep_is_held(&net_family_lock)))
3026 		err = -EEXIST;
3027 	else {
3028 		rcu_assign_pointer(net_families[ops->family], ops);
3029 		err = 0;
3030 	}
3031 	spin_unlock(&net_family_lock);
3032 
3033 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3034 	return err;
3035 }
3036 EXPORT_SYMBOL(sock_register);
3037 
3038 /**
3039  *	sock_unregister - remove a protocol handler
3040  *	@family: protocol family to remove
3041  *
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.
3045  *
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.
3050  */
3051 void sock_unregister(int family)
3052 {
3053 	BUG_ON(family < 0 || family >= NPROTO);
3054 
3055 	spin_lock(&net_family_lock);
3056 	RCU_INIT_POINTER(net_families[family], NULL);
3057 	spin_unlock(&net_family_lock);
3058 
3059 	synchronize_rcu();
3060 
3061 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3062 }
3063 EXPORT_SYMBOL(sock_unregister);
3064 
3065 bool sock_is_registered(int family)
3066 {
3067 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3068 }
3069 
3070 static int __init sock_init(void)
3071 {
3072 	int err;
3073 	/*
3074 	 *      Initialize the network sysctl infrastructure.
3075 	 */
3076 	err = net_sysctl_init();
3077 	if (err)
3078 		goto out;
3079 
3080 	/*
3081 	 *      Initialize skbuff SLAB cache
3082 	 */
3083 	skb_init();
3084 
3085 	/*
3086 	 *      Initialize the protocols module.
3087 	 */
3088 
3089 	init_inodecache();
3090 
3091 	err = register_filesystem(&sock_fs_type);
3092 	if (err)
3093 		goto out;
3094 	sock_mnt = kern_mount(&sock_fs_type);
3095 	if (IS_ERR(sock_mnt)) {
3096 		err = PTR_ERR(sock_mnt);
3097 		goto out_mount;
3098 	}
3099 
3100 	/* The real protocol initialization is performed in later initcalls.
3101 	 */
3102 
3103 #ifdef CONFIG_NETFILTER
3104 	err = netfilter_init();
3105 	if (err)
3106 		goto out;
3107 #endif
3108 
3109 	ptp_classifier_init();
3110 
3111 out:
3112 	return err;
3113 
3114 out_mount:
3115 	unregister_filesystem(&sock_fs_type);
3116 	goto out;
3117 }
3118 
3119 core_initcall(sock_init);	/* early initcall */
3120 
3121 #ifdef CONFIG_PROC_FS
3122 void socket_seq_show(struct seq_file *seq)
3123 {
3124 	seq_printf(seq, "sockets: used %d\n",
3125 		   sock_inuse_get(seq->private));
3126 }
3127 #endif				/* CONFIG_PROC_FS */
3128 
3129 #ifdef CONFIG_COMPAT
3130 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
3131 {
3132 	struct compat_ifconf ifc32;
3133 	struct ifconf ifc;
3134 	int err;
3135 
3136 	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3137 		return -EFAULT;
3138 
3139 	ifc.ifc_len = ifc32.ifc_len;
3140 	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3141 
3142 	rtnl_lock();
3143 	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3144 	rtnl_unlock();
3145 	if (err)
3146 		return err;
3147 
3148 	ifc32.ifc_len = ifc.ifc_len;
3149 	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3150 		return -EFAULT;
3151 
3152 	return 0;
3153 }
3154 
3155 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3156 {
3157 	compat_uptr_t uptr32;
3158 	struct ifreq ifr;
3159 	void __user *saved;
3160 	int err;
3161 
3162 	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3163 		return -EFAULT;
3164 
3165 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3166 		return -EFAULT;
3167 
3168 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3169 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3170 
3171 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3172 	if (!err) {
3173 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3174 		if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3175 			err = -EFAULT;
3176 	}
3177 	return err;
3178 }
3179 
3180 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3181 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3182 				 struct compat_ifreq __user *u_ifreq32)
3183 {
3184 	struct ifreq ifreq;
3185 	u32 data32;
3186 
3187 	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3188 		return -EFAULT;
3189 	if (get_user(data32, &u_ifreq32->ifr_data))
3190 		return -EFAULT;
3191 	ifreq.ifr_data = compat_ptr(data32);
3192 
3193 	return dev_ioctl(net, cmd, &ifreq, NULL);
3194 }
3195 
3196 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3197 			      unsigned int cmd,
3198 			      struct compat_ifreq __user *uifr32)
3199 {
3200 	struct ifreq __user *uifr;
3201 	int err;
3202 
3203 	/* Handle the fact that while struct ifreq has the same *layout* on
3204 	 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3205 	 * which are handled elsewhere, it still has different *size* due to
3206 	 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3207 	 * resulting in struct ifreq being 32 and 40 bytes respectively).
3208 	 * As a result, if the struct happens to be at the end of a page and
3209 	 * the next page isn't readable/writable, we get a fault. To prevent
3210 	 * that, copy back and forth to the full size.
3211 	 */
3212 
3213 	uifr = compat_alloc_user_space(sizeof(*uifr));
3214 	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3215 		return -EFAULT;
3216 
3217 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3218 
3219 	if (!err) {
3220 		switch (cmd) {
3221 		case SIOCGIFFLAGS:
3222 		case SIOCGIFMETRIC:
3223 		case SIOCGIFMTU:
3224 		case SIOCGIFMEM:
3225 		case SIOCGIFHWADDR:
3226 		case SIOCGIFINDEX:
3227 		case SIOCGIFADDR:
3228 		case SIOCGIFBRDADDR:
3229 		case SIOCGIFDSTADDR:
3230 		case SIOCGIFNETMASK:
3231 		case SIOCGIFPFLAGS:
3232 		case SIOCGIFTXQLEN:
3233 		case SIOCGMIIPHY:
3234 		case SIOCGMIIREG:
3235 		case SIOCGIFNAME:
3236 			if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3237 				err = -EFAULT;
3238 			break;
3239 		}
3240 	}
3241 	return err;
3242 }
3243 
3244 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3245 			struct compat_ifreq __user *uifr32)
3246 {
3247 	struct ifreq ifr;
3248 	struct compat_ifmap __user *uifmap32;
3249 	int err;
3250 
3251 	uifmap32 = &uifr32->ifr_ifru.ifru_map;
3252 	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3253 	err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3254 	err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3255 	err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3256 	err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3257 	err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3258 	err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3259 	if (err)
3260 		return -EFAULT;
3261 
3262 	err = dev_ioctl(net, cmd, &ifr, NULL);
3263 
3264 	if (cmd == SIOCGIFMAP && !err) {
3265 		err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3266 		err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3267 		err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3268 		err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3269 		err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3270 		err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3271 		err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3272 		if (err)
3273 			err = -EFAULT;
3274 	}
3275 	return err;
3276 }
3277 
3278 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3279  * for some operations; this forces use of the newer bridge-utils that
3280  * use compatible ioctls
3281  */
3282 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3283 {
3284 	compat_ulong_t tmp;
3285 
3286 	if (get_user(tmp, argp))
3287 		return -EFAULT;
3288 	if (tmp == BRCTL_GET_VERSION)
3289 		return BRCTL_VERSION + 1;
3290 	return -EINVAL;
3291 }
3292 
3293 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3294 			 unsigned int cmd, unsigned long arg)
3295 {
3296 	void __user *argp = compat_ptr(arg);
3297 	struct sock *sk = sock->sk;
3298 	struct net *net = sock_net(sk);
3299 
3300 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3301 		return compat_ifr_data_ioctl(net, cmd, argp);
3302 
3303 	switch (cmd) {
3304 	case SIOCSIFBR:
3305 	case SIOCGIFBR:
3306 		return old_bridge_ioctl(argp);
3307 	case SIOCGIFCONF:
3308 		return compat_dev_ifconf(net, argp);
3309 	case SIOCWANDEV:
3310 		return compat_siocwandev(net, argp);
3311 	case SIOCGIFMAP:
3312 	case SIOCSIFMAP:
3313 		return compat_sioc_ifmap(net, cmd, argp);
3314 	case SIOCGSTAMP_OLD:
3315 	case SIOCGSTAMPNS_OLD:
3316 		if (!sock->ops->gettstamp)
3317 			return -ENOIOCTLCMD;
3318 		return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3319 					    !COMPAT_USE_64BIT_TIME);
3320 
3321 	case SIOCETHTOOL:
3322 	case SIOCBONDSLAVEINFOQUERY:
3323 	case SIOCBONDINFOQUERY:
3324 	case SIOCSHWTSTAMP:
3325 	case SIOCGHWTSTAMP:
3326 		return compat_ifr_data_ioctl(net, cmd, argp);
3327 
3328 	case FIOSETOWN:
3329 	case SIOCSPGRP:
3330 	case FIOGETOWN:
3331 	case SIOCGPGRP:
3332 	case SIOCBRADDBR:
3333 	case SIOCBRDELBR:
3334 	case SIOCGIFVLAN:
3335 	case SIOCSIFVLAN:
3336 	case SIOCGSKNS:
3337 	case SIOCGSTAMP_NEW:
3338 	case SIOCGSTAMPNS_NEW:
3339 		return sock_ioctl(file, cmd, arg);
3340 
3341 	case SIOCGIFFLAGS:
3342 	case SIOCSIFFLAGS:
3343 	case SIOCGIFMETRIC:
3344 	case SIOCSIFMETRIC:
3345 	case SIOCGIFMTU:
3346 	case SIOCSIFMTU:
3347 	case SIOCGIFMEM:
3348 	case SIOCSIFMEM:
3349 	case SIOCGIFHWADDR:
3350 	case SIOCSIFHWADDR:
3351 	case SIOCADDMULTI:
3352 	case SIOCDELMULTI:
3353 	case SIOCGIFINDEX:
3354 	case SIOCGIFADDR:
3355 	case SIOCSIFADDR:
3356 	case SIOCSIFHWBROADCAST:
3357 	case SIOCDIFADDR:
3358 	case SIOCGIFBRDADDR:
3359 	case SIOCSIFBRDADDR:
3360 	case SIOCGIFDSTADDR:
3361 	case SIOCSIFDSTADDR:
3362 	case SIOCGIFNETMASK:
3363 	case SIOCSIFNETMASK:
3364 	case SIOCSIFPFLAGS:
3365 	case SIOCGIFPFLAGS:
3366 	case SIOCGIFTXQLEN:
3367 	case SIOCSIFTXQLEN:
3368 	case SIOCBRADDIF:
3369 	case SIOCBRDELIF:
3370 	case SIOCGIFNAME:
3371 	case SIOCSIFNAME:
3372 	case SIOCGMIIPHY:
3373 	case SIOCGMIIREG:
3374 	case SIOCSMIIREG:
3375 	case SIOCBONDENSLAVE:
3376 	case SIOCBONDRELEASE:
3377 	case SIOCBONDSETHWADDR:
3378 	case SIOCBONDCHANGEACTIVE:
3379 		return compat_ifreq_ioctl(net, sock, cmd, argp);
3380 
3381 	case SIOCSARP:
3382 	case SIOCGARP:
3383 	case SIOCDARP:
3384 	case SIOCOUTQ:
3385 	case SIOCOUTQNSD:
3386 	case SIOCATMARK:
3387 		return sock_do_ioctl(net, sock, cmd, arg);
3388 	}
3389 
3390 	return -ENOIOCTLCMD;
3391 }
3392 
3393 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3394 			      unsigned long arg)
3395 {
3396 	struct socket *sock = file->private_data;
3397 	int ret = -ENOIOCTLCMD;
3398 	struct sock *sk;
3399 	struct net *net;
3400 
3401 	sk = sock->sk;
3402 	net = sock_net(sk);
3403 
3404 	if (sock->ops->compat_ioctl)
3405 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3406 
3407 	if (ret == -ENOIOCTLCMD &&
3408 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3409 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3410 
3411 	if (ret == -ENOIOCTLCMD)
3412 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3413 
3414 	return ret;
3415 }
3416 #endif
3417 
3418 /**
3419  *	kernel_bind - bind an address to a socket (kernel space)
3420  *	@sock: socket
3421  *	@addr: address
3422  *	@addrlen: length of address
3423  *
3424  *	Returns 0 or an error.
3425  */
3426 
3427 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3428 {
3429 	return sock->ops->bind(sock, addr, addrlen);
3430 }
3431 EXPORT_SYMBOL(kernel_bind);
3432 
3433 /**
3434  *	kernel_listen - move socket to listening state (kernel space)
3435  *	@sock: socket
3436  *	@backlog: pending connections queue size
3437  *
3438  *	Returns 0 or an error.
3439  */
3440 
3441 int kernel_listen(struct socket *sock, int backlog)
3442 {
3443 	return sock->ops->listen(sock, backlog);
3444 }
3445 EXPORT_SYMBOL(kernel_listen);
3446 
3447 /**
3448  *	kernel_accept - accept a connection (kernel space)
3449  *	@sock: listening socket
3450  *	@newsock: new connected socket
3451  *	@flags: flags
3452  *
3453  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3454  *	If it fails, @newsock is guaranteed to be %NULL.
3455  *	Returns 0 or an error.
3456  */
3457 
3458 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3459 {
3460 	struct sock *sk = sock->sk;
3461 	int err;
3462 
3463 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3464 			       newsock);
3465 	if (err < 0)
3466 		goto done;
3467 
3468 	err = sock->ops->accept(sock, *newsock, flags, true);
3469 	if (err < 0) {
3470 		sock_release(*newsock);
3471 		*newsock = NULL;
3472 		goto done;
3473 	}
3474 
3475 	(*newsock)->ops = sock->ops;
3476 	__module_get((*newsock)->ops->owner);
3477 
3478 done:
3479 	return err;
3480 }
3481 EXPORT_SYMBOL(kernel_accept);
3482 
3483 /**
3484  *	kernel_connect - connect a socket (kernel space)
3485  *	@sock: socket
3486  *	@addr: address
3487  *	@addrlen: address length
3488  *	@flags: flags (O_NONBLOCK, ...)
3489  *
3490  *	For datagram sockets, @addr is the address to which datagrams are sent
3491  *	by default, and the only address from which datagrams are received.
3492  *	For stream sockets, attempts to connect to @addr.
3493  *	Returns 0 or an error code.
3494  */
3495 
3496 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3497 		   int flags)
3498 {
3499 	return sock->ops->connect(sock, addr, addrlen, flags);
3500 }
3501 EXPORT_SYMBOL(kernel_connect);
3502 
3503 /**
3504  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3505  *	@sock: socket
3506  *	@addr: address holder
3507  *
3508  * 	Fills the @addr pointer with the address which the socket is bound.
3509  *	Returns 0 or an error code.
3510  */
3511 
3512 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3513 {
3514 	return sock->ops->getname(sock, addr, 0);
3515 }
3516 EXPORT_SYMBOL(kernel_getsockname);
3517 
3518 /**
3519  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3520  *	@sock: socket
3521  *	@addr: address holder
3522  *
3523  * 	Fills the @addr pointer with the address which the socket is connected.
3524  *	Returns 0 or an error code.
3525  */
3526 
3527 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3528 {
3529 	return sock->ops->getname(sock, addr, 1);
3530 }
3531 EXPORT_SYMBOL(kernel_getpeername);
3532 
3533 /**
3534  *	kernel_sendpage - send a &page through a socket (kernel space)
3535  *	@sock: socket
3536  *	@page: page
3537  *	@offset: page offset
3538  *	@size: total size in bytes
3539  *	@flags: flags (MSG_DONTWAIT, ...)
3540  *
3541  *	Returns the total amount sent in bytes or an error.
3542  */
3543 
3544 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3545 		    size_t size, int flags)
3546 {
3547 	if (sock->ops->sendpage) {
3548 		/* Warn in case the improper page to zero-copy send */
3549 		WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3550 		return sock->ops->sendpage(sock, page, offset, size, flags);
3551 	}
3552 	return sock_no_sendpage(sock, page, offset, size, flags);
3553 }
3554 EXPORT_SYMBOL(kernel_sendpage);
3555 
3556 /**
3557  *	kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3558  *	@sk: sock
3559  *	@page: page
3560  *	@offset: page offset
3561  *	@size: total size in bytes
3562  *	@flags: flags (MSG_DONTWAIT, ...)
3563  *
3564  *	Returns the total amount sent in bytes or an error.
3565  *	Caller must hold @sk.
3566  */
3567 
3568 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3569 			   size_t size, int flags)
3570 {
3571 	struct socket *sock = sk->sk_socket;
3572 
3573 	if (sock->ops->sendpage_locked)
3574 		return sock->ops->sendpage_locked(sk, page, offset, size,
3575 						  flags);
3576 
3577 	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3578 }
3579 EXPORT_SYMBOL(kernel_sendpage_locked);
3580 
3581 /**
3582  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3583  *	@sock: socket
3584  *	@how: connection part
3585  *
3586  *	Returns 0 or an error.
3587  */
3588 
3589 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3590 {
3591 	return sock->ops->shutdown(sock, how);
3592 }
3593 EXPORT_SYMBOL(kernel_sock_shutdown);
3594 
3595 /**
3596  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3597  *	@sk: socket
3598  *
3599  *	This routine returns the IP overhead imposed by a socket i.e.
3600  *	the length of the underlying IP header, depending on whether
3601  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3602  *	on at the socket. Assumes that the caller has a lock on the socket.
3603  */
3604 
3605 u32 kernel_sock_ip_overhead(struct sock *sk)
3606 {
3607 	struct inet_sock *inet;
3608 	struct ip_options_rcu *opt;
3609 	u32 overhead = 0;
3610 #if IS_ENABLED(CONFIG_IPV6)
3611 	struct ipv6_pinfo *np;
3612 	struct ipv6_txoptions *optv6 = NULL;
3613 #endif /* IS_ENABLED(CONFIG_IPV6) */
3614 
3615 	if (!sk)
3616 		return overhead;
3617 
3618 	switch (sk->sk_family) {
3619 	case AF_INET:
3620 		inet = inet_sk(sk);
3621 		overhead += sizeof(struct iphdr);
3622 		opt = rcu_dereference_protected(inet->inet_opt,
3623 						sock_owned_by_user(sk));
3624 		if (opt)
3625 			overhead += opt->opt.optlen;
3626 		return overhead;
3627 #if IS_ENABLED(CONFIG_IPV6)
3628 	case AF_INET6:
3629 		np = inet6_sk(sk);
3630 		overhead += sizeof(struct ipv6hdr);
3631 		if (np)
3632 			optv6 = rcu_dereference_protected(np->opt,
3633 							  sock_owned_by_user(sk));
3634 		if (optv6)
3635 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3636 		return overhead;
3637 #endif /* IS_ENABLED(CONFIG_IPV6) */
3638 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3639 		return overhead;
3640 	}
3641 }
3642 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3643