xref: /openbmc/linux/net/socket.c (revision ee7da21a)
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 	[PF_MCTP]	= "PF_MCTP",
216 };
217 
218 /*
219  *	The protocol list. Each protocol is registered in here.
220  */
221 
222 static DEFINE_SPINLOCK(net_family_lock);
223 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
224 
225 /*
226  * Support routines.
227  * Move socket addresses back and forth across the kernel/user
228  * divide and look after the messy bits.
229  */
230 
231 /**
232  *	move_addr_to_kernel	-	copy a socket address into kernel space
233  *	@uaddr: Address in user space
234  *	@kaddr: Address in kernel space
235  *	@ulen: Length in user space
236  *
237  *	The address is copied into kernel space. If the provided address is
238  *	too long an error code of -EINVAL is returned. If the copy gives
239  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
240  */
241 
242 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
243 {
244 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
245 		return -EINVAL;
246 	if (ulen == 0)
247 		return 0;
248 	if (copy_from_user(kaddr, uaddr, ulen))
249 		return -EFAULT;
250 	return audit_sockaddr(ulen, kaddr);
251 }
252 
253 /**
254  *	move_addr_to_user	-	copy an address to user space
255  *	@kaddr: kernel space address
256  *	@klen: length of address in kernel
257  *	@uaddr: user space address
258  *	@ulen: pointer to user length field
259  *
260  *	The value pointed to by ulen on entry is the buffer length available.
261  *	This is overwritten with the buffer space used. -EINVAL is returned
262  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
263  *	is returned if either the buffer or the length field are not
264  *	accessible.
265  *	After copying the data up to the limit the user specifies, the true
266  *	length of the data is written over the length limit the user
267  *	specified. Zero is returned for a success.
268  */
269 
270 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
271 			     void __user *uaddr, int __user *ulen)
272 {
273 	int err;
274 	int len;
275 
276 	BUG_ON(klen > sizeof(struct sockaddr_storage));
277 	err = get_user(len, ulen);
278 	if (err)
279 		return err;
280 	if (len > klen)
281 		len = klen;
282 	if (len < 0)
283 		return -EINVAL;
284 	if (len) {
285 		if (audit_sockaddr(klen, kaddr))
286 			return -ENOMEM;
287 		if (copy_to_user(uaddr, kaddr, len))
288 			return -EFAULT;
289 	}
290 	/*
291 	 *      "fromlen shall refer to the value before truncation.."
292 	 *                      1003.1g
293 	 */
294 	return __put_user(klen, ulen);
295 }
296 
297 static struct kmem_cache *sock_inode_cachep __ro_after_init;
298 
299 static struct inode *sock_alloc_inode(struct super_block *sb)
300 {
301 	struct socket_alloc *ei;
302 
303 	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
304 	if (!ei)
305 		return NULL;
306 	init_waitqueue_head(&ei->socket.wq.wait);
307 	ei->socket.wq.fasync_list = NULL;
308 	ei->socket.wq.flags = 0;
309 
310 	ei->socket.state = SS_UNCONNECTED;
311 	ei->socket.flags = 0;
312 	ei->socket.ops = NULL;
313 	ei->socket.sk = NULL;
314 	ei->socket.file = NULL;
315 
316 	return &ei->vfs_inode;
317 }
318 
319 static void sock_free_inode(struct inode *inode)
320 {
321 	struct socket_alloc *ei;
322 
323 	ei = container_of(inode, struct socket_alloc, vfs_inode);
324 	kmem_cache_free(sock_inode_cachep, ei);
325 }
326 
327 static void init_once(void *foo)
328 {
329 	struct socket_alloc *ei = (struct socket_alloc *)foo;
330 
331 	inode_init_once(&ei->vfs_inode);
332 }
333 
334 static void init_inodecache(void)
335 {
336 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
337 					      sizeof(struct socket_alloc),
338 					      0,
339 					      (SLAB_HWCACHE_ALIGN |
340 					       SLAB_RECLAIM_ACCOUNT |
341 					       SLAB_MEM_SPREAD | SLAB_ACCOUNT),
342 					      init_once);
343 	BUG_ON(sock_inode_cachep == NULL);
344 }
345 
346 static const struct super_operations sockfs_ops = {
347 	.alloc_inode	= sock_alloc_inode,
348 	.free_inode	= sock_free_inode,
349 	.statfs		= simple_statfs,
350 };
351 
352 /*
353  * sockfs_dname() is called from d_path().
354  */
355 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
356 {
357 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
358 				d_inode(dentry)->i_ino);
359 }
360 
361 static const struct dentry_operations sockfs_dentry_operations = {
362 	.d_dname  = sockfs_dname,
363 };
364 
365 static int sockfs_xattr_get(const struct xattr_handler *handler,
366 			    struct dentry *dentry, struct inode *inode,
367 			    const char *suffix, void *value, size_t size)
368 {
369 	if (value) {
370 		if (dentry->d_name.len + 1 > size)
371 			return -ERANGE;
372 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
373 	}
374 	return dentry->d_name.len + 1;
375 }
376 
377 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
378 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
379 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
380 
381 static const struct xattr_handler sockfs_xattr_handler = {
382 	.name = XATTR_NAME_SOCKPROTONAME,
383 	.get = sockfs_xattr_get,
384 };
385 
386 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
387 				     struct user_namespace *mnt_userns,
388 				     struct dentry *dentry, struct inode *inode,
389 				     const char *suffix, const void *value,
390 				     size_t size, int flags)
391 {
392 	/* Handled by LSM. */
393 	return -EAGAIN;
394 }
395 
396 static const struct xattr_handler sockfs_security_xattr_handler = {
397 	.prefix = XATTR_SECURITY_PREFIX,
398 	.set = sockfs_security_xattr_set,
399 };
400 
401 static const struct xattr_handler *sockfs_xattr_handlers[] = {
402 	&sockfs_xattr_handler,
403 	&sockfs_security_xattr_handler,
404 	NULL
405 };
406 
407 static int sockfs_init_fs_context(struct fs_context *fc)
408 {
409 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
410 	if (!ctx)
411 		return -ENOMEM;
412 	ctx->ops = &sockfs_ops;
413 	ctx->dops = &sockfs_dentry_operations;
414 	ctx->xattr = sockfs_xattr_handlers;
415 	return 0;
416 }
417 
418 static struct vfsmount *sock_mnt __read_mostly;
419 
420 static struct file_system_type sock_fs_type = {
421 	.name =		"sockfs",
422 	.init_fs_context = sockfs_init_fs_context,
423 	.kill_sb =	kill_anon_super,
424 };
425 
426 /*
427  *	Obtains the first available file descriptor and sets it up for use.
428  *
429  *	These functions create file structures and maps them to fd space
430  *	of the current process. On success it returns file descriptor
431  *	and file struct implicitly stored in sock->file.
432  *	Note that another thread may close file descriptor before we return
433  *	from this function. We use the fact that now we do not refer
434  *	to socket after mapping. If one day we will need it, this
435  *	function will increment ref. count on file by 1.
436  *
437  *	In any case returned fd MAY BE not valid!
438  *	This race condition is unavoidable
439  *	with shared fd spaces, we cannot solve it inside kernel,
440  *	but we take care of internal coherence yet.
441  */
442 
443 /**
444  *	sock_alloc_file - Bind a &socket to a &file
445  *	@sock: socket
446  *	@flags: file status flags
447  *	@dname: protocol name
448  *
449  *	Returns the &file bound with @sock, implicitly storing it
450  *	in sock->file. If dname is %NULL, sets to "".
451  *	On failure the return is a ERR pointer (see linux/err.h).
452  *	This function uses GFP_KERNEL internally.
453  */
454 
455 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
456 {
457 	struct file *file;
458 
459 	if (!dname)
460 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
461 
462 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
463 				O_RDWR | (flags & O_NONBLOCK),
464 				&socket_file_ops);
465 	if (IS_ERR(file)) {
466 		sock_release(sock);
467 		return file;
468 	}
469 
470 	sock->file = file;
471 	file->private_data = sock;
472 	stream_open(SOCK_INODE(sock), file);
473 	return file;
474 }
475 EXPORT_SYMBOL(sock_alloc_file);
476 
477 static int sock_map_fd(struct socket *sock, int flags)
478 {
479 	struct file *newfile;
480 	int fd = get_unused_fd_flags(flags);
481 	if (unlikely(fd < 0)) {
482 		sock_release(sock);
483 		return fd;
484 	}
485 
486 	newfile = sock_alloc_file(sock, flags, NULL);
487 	if (!IS_ERR(newfile)) {
488 		fd_install(fd, newfile);
489 		return fd;
490 	}
491 
492 	put_unused_fd(fd);
493 	return PTR_ERR(newfile);
494 }
495 
496 /**
497  *	sock_from_file - Return the &socket bounded to @file.
498  *	@file: file
499  *
500  *	On failure returns %NULL.
501  */
502 
503 struct socket *sock_from_file(struct file *file)
504 {
505 	if (file->f_op == &socket_file_ops)
506 		return file->private_data;	/* set in sock_map_fd */
507 
508 	return NULL;
509 }
510 EXPORT_SYMBOL(sock_from_file);
511 
512 /**
513  *	sockfd_lookup - Go from a file number to its socket slot
514  *	@fd: file handle
515  *	@err: pointer to an error code return
516  *
517  *	The file handle passed in is locked and the socket it is bound
518  *	to is returned. If an error occurs the err pointer is overwritten
519  *	with a negative errno code and NULL is returned. The function checks
520  *	for both invalid handles and passing a handle which is not a socket.
521  *
522  *	On a success the socket object pointer is returned.
523  */
524 
525 struct socket *sockfd_lookup(int fd, int *err)
526 {
527 	struct file *file;
528 	struct socket *sock;
529 
530 	file = fget(fd);
531 	if (!file) {
532 		*err = -EBADF;
533 		return NULL;
534 	}
535 
536 	sock = sock_from_file(file);
537 	if (!sock) {
538 		*err = -ENOTSOCK;
539 		fput(file);
540 	}
541 	return sock;
542 }
543 EXPORT_SYMBOL(sockfd_lookup);
544 
545 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
546 {
547 	struct fd f = fdget(fd);
548 	struct socket *sock;
549 
550 	*err = -EBADF;
551 	if (f.file) {
552 		sock = sock_from_file(f.file);
553 		if (likely(sock)) {
554 			*fput_needed = f.flags & FDPUT_FPUT;
555 			return sock;
556 		}
557 		*err = -ENOTSOCK;
558 		fdput(f);
559 	}
560 	return NULL;
561 }
562 
563 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
564 				size_t size)
565 {
566 	ssize_t len;
567 	ssize_t used = 0;
568 
569 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
570 	if (len < 0)
571 		return len;
572 	used += len;
573 	if (buffer) {
574 		if (size < used)
575 			return -ERANGE;
576 		buffer += len;
577 	}
578 
579 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
580 	used += len;
581 	if (buffer) {
582 		if (size < used)
583 			return -ERANGE;
584 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
585 		buffer += len;
586 	}
587 
588 	return used;
589 }
590 
591 static int sockfs_setattr(struct user_namespace *mnt_userns,
592 			  struct dentry *dentry, struct iattr *iattr)
593 {
594 	int err = simple_setattr(&init_user_ns, dentry, iattr);
595 
596 	if (!err && (iattr->ia_valid & ATTR_UID)) {
597 		struct socket *sock = SOCKET_I(d_inode(dentry));
598 
599 		if (sock->sk)
600 			sock->sk->sk_uid = iattr->ia_uid;
601 		else
602 			err = -ENOENT;
603 	}
604 
605 	return err;
606 }
607 
608 static const struct inode_operations sockfs_inode_ops = {
609 	.listxattr = sockfs_listxattr,
610 	.setattr = sockfs_setattr,
611 };
612 
613 /**
614  *	sock_alloc - allocate a socket
615  *
616  *	Allocate a new inode and socket object. The two are bound together
617  *	and initialised. The socket is then returned. If we are out of inodes
618  *	NULL is returned. This functions uses GFP_KERNEL internally.
619  */
620 
621 struct socket *sock_alloc(void)
622 {
623 	struct inode *inode;
624 	struct socket *sock;
625 
626 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
627 	if (!inode)
628 		return NULL;
629 
630 	sock = SOCKET_I(inode);
631 
632 	inode->i_ino = get_next_ino();
633 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
634 	inode->i_uid = current_fsuid();
635 	inode->i_gid = current_fsgid();
636 	inode->i_op = &sockfs_inode_ops;
637 
638 	return sock;
639 }
640 EXPORT_SYMBOL(sock_alloc);
641 
642 static void __sock_release(struct socket *sock, struct inode *inode)
643 {
644 	if (sock->ops) {
645 		struct module *owner = sock->ops->owner;
646 
647 		if (inode)
648 			inode_lock(inode);
649 		sock->ops->release(sock);
650 		sock->sk = NULL;
651 		if (inode)
652 			inode_unlock(inode);
653 		sock->ops = NULL;
654 		module_put(owner);
655 	}
656 
657 	if (sock->wq.fasync_list)
658 		pr_err("%s: fasync list not empty!\n", __func__);
659 
660 	if (!sock->file) {
661 		iput(SOCK_INODE(sock));
662 		return;
663 	}
664 	sock->file = NULL;
665 }
666 
667 /**
668  *	sock_release - close a socket
669  *	@sock: socket to close
670  *
671  *	The socket is released from the protocol stack if it has a release
672  *	callback, and the inode is then released if the socket is bound to
673  *	an inode not a file.
674  */
675 void sock_release(struct socket *sock)
676 {
677 	__sock_release(sock, NULL);
678 }
679 EXPORT_SYMBOL(sock_release);
680 
681 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
682 {
683 	u8 flags = *tx_flags;
684 
685 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
686 		flags |= SKBTX_HW_TSTAMP;
687 
688 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
689 		flags |= SKBTX_SW_TSTAMP;
690 
691 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
692 		flags |= SKBTX_SCHED_TSTAMP;
693 
694 	*tx_flags = flags;
695 }
696 EXPORT_SYMBOL(__sock_tx_timestamp);
697 
698 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
699 					   size_t));
700 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
701 					    size_t));
702 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
703 {
704 	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
705 				     inet_sendmsg, sock, msg,
706 				     msg_data_left(msg));
707 	BUG_ON(ret == -EIOCBQUEUED);
708 	return ret;
709 }
710 
711 /**
712  *	sock_sendmsg - send a message through @sock
713  *	@sock: socket
714  *	@msg: message to send
715  *
716  *	Sends @msg through @sock, passing through LSM.
717  *	Returns the number of bytes sent, or an error code.
718  */
719 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
720 {
721 	int err = security_socket_sendmsg(sock, msg,
722 					  msg_data_left(msg));
723 
724 	return err ?: sock_sendmsg_nosec(sock, msg);
725 }
726 EXPORT_SYMBOL(sock_sendmsg);
727 
728 /**
729  *	kernel_sendmsg - send a message through @sock (kernel-space)
730  *	@sock: socket
731  *	@msg: message header
732  *	@vec: kernel vec
733  *	@num: vec array length
734  *	@size: total message data size
735  *
736  *	Builds the message data with @vec and sends it through @sock.
737  *	Returns the number of bytes sent, or an error code.
738  */
739 
740 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
741 		   struct kvec *vec, size_t num, size_t size)
742 {
743 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
744 	return sock_sendmsg(sock, msg);
745 }
746 EXPORT_SYMBOL(kernel_sendmsg);
747 
748 /**
749  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
750  *	@sk: sock
751  *	@msg: message header
752  *	@vec: output s/g array
753  *	@num: output s/g array length
754  *	@size: total message data size
755  *
756  *	Builds the message data with @vec and sends it through @sock.
757  *	Returns the number of bytes sent, or an error code.
758  *	Caller must hold @sk.
759  */
760 
761 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
762 			  struct kvec *vec, size_t num, size_t size)
763 {
764 	struct socket *sock = sk->sk_socket;
765 
766 	if (!sock->ops->sendmsg_locked)
767 		return sock_no_sendmsg_locked(sk, msg, size);
768 
769 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
770 
771 	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
772 }
773 EXPORT_SYMBOL(kernel_sendmsg_locked);
774 
775 static bool skb_is_err_queue(const struct sk_buff *skb)
776 {
777 	/* pkt_type of skbs enqueued on the error queue are set to
778 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
779 	 * in recvmsg, since skbs received on a local socket will never
780 	 * have a pkt_type of PACKET_OUTGOING.
781 	 */
782 	return skb->pkt_type == PACKET_OUTGOING;
783 }
784 
785 /* On transmit, software and hardware timestamps are returned independently.
786  * As the two skb clones share the hardware timestamp, which may be updated
787  * before the software timestamp is received, a hardware TX timestamp may be
788  * returned only if there is no software TX timestamp. Ignore false software
789  * timestamps, which may be made in the __sock_recv_timestamp() call when the
790  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
791  * hardware timestamp.
792  */
793 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
794 {
795 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
796 }
797 
798 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
799 {
800 	struct scm_ts_pktinfo ts_pktinfo;
801 	struct net_device *orig_dev;
802 
803 	if (!skb_mac_header_was_set(skb))
804 		return;
805 
806 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
807 
808 	rcu_read_lock();
809 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
810 	if (orig_dev)
811 		ts_pktinfo.if_index = orig_dev->ifindex;
812 	rcu_read_unlock();
813 
814 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
815 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
816 		 sizeof(ts_pktinfo), &ts_pktinfo);
817 }
818 
819 /*
820  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
821  */
822 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
823 	struct sk_buff *skb)
824 {
825 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
826 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
827 	struct scm_timestamping_internal tss;
828 
829 	int empty = 1, false_tstamp = 0;
830 	struct skb_shared_hwtstamps *shhwtstamps =
831 		skb_hwtstamps(skb);
832 
833 	/* Race occurred between timestamp enabling and packet
834 	   receiving.  Fill in the current time for now. */
835 	if (need_software_tstamp && skb->tstamp == 0) {
836 		__net_timestamp(skb);
837 		false_tstamp = 1;
838 	}
839 
840 	if (need_software_tstamp) {
841 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
842 			if (new_tstamp) {
843 				struct __kernel_sock_timeval tv;
844 
845 				skb_get_new_timestamp(skb, &tv);
846 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
847 					 sizeof(tv), &tv);
848 			} else {
849 				struct __kernel_old_timeval tv;
850 
851 				skb_get_timestamp(skb, &tv);
852 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
853 					 sizeof(tv), &tv);
854 			}
855 		} else {
856 			if (new_tstamp) {
857 				struct __kernel_timespec ts;
858 
859 				skb_get_new_timestampns(skb, &ts);
860 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
861 					 sizeof(ts), &ts);
862 			} else {
863 				struct __kernel_old_timespec ts;
864 
865 				skb_get_timestampns(skb, &ts);
866 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
867 					 sizeof(ts), &ts);
868 			}
869 		}
870 	}
871 
872 	memset(&tss, 0, sizeof(tss));
873 	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
874 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
875 		empty = 0;
876 	if (shhwtstamps &&
877 	    (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
878 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
879 		if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
880 			ptp_convert_timestamp(shhwtstamps, sk->sk_bind_phc);
881 
882 		if (ktime_to_timespec64_cond(shhwtstamps->hwtstamp,
883 					     tss.ts + 2)) {
884 			empty = 0;
885 
886 			if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
887 			    !skb_is_err_queue(skb))
888 				put_ts_pktinfo(msg, skb);
889 		}
890 	}
891 	if (!empty) {
892 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
893 			put_cmsg_scm_timestamping64(msg, &tss);
894 		else
895 			put_cmsg_scm_timestamping(msg, &tss);
896 
897 		if (skb_is_err_queue(skb) && skb->len &&
898 		    SKB_EXT_ERR(skb)->opt_stats)
899 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
900 				 skb->len, skb->data);
901 	}
902 }
903 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
904 
905 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
906 	struct sk_buff *skb)
907 {
908 	int ack;
909 
910 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
911 		return;
912 	if (!skb->wifi_acked_valid)
913 		return;
914 
915 	ack = skb->wifi_acked;
916 
917 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
918 }
919 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
920 
921 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
922 				   struct sk_buff *skb)
923 {
924 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
925 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
926 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
927 }
928 
929 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
930 	struct sk_buff *skb)
931 {
932 	sock_recv_timestamp(msg, sk, skb);
933 	sock_recv_drops(msg, sk, skb);
934 }
935 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
936 
937 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
938 					   size_t, int));
939 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
940 					    size_t, int));
941 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
942 				     int flags)
943 {
944 	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
945 				  inet_recvmsg, sock, msg, msg_data_left(msg),
946 				  flags);
947 }
948 
949 /**
950  *	sock_recvmsg - receive a message from @sock
951  *	@sock: socket
952  *	@msg: message to receive
953  *	@flags: message flags
954  *
955  *	Receives @msg from @sock, passing through LSM. Returns the total number
956  *	of bytes received, or an error.
957  */
958 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
959 {
960 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
961 
962 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
963 }
964 EXPORT_SYMBOL(sock_recvmsg);
965 
966 /**
967  *	kernel_recvmsg - Receive a message from a socket (kernel space)
968  *	@sock: The socket to receive the message from
969  *	@msg: Received message
970  *	@vec: Input s/g array for message data
971  *	@num: Size of input s/g array
972  *	@size: Number of bytes to read
973  *	@flags: Message flags (MSG_DONTWAIT, etc...)
974  *
975  *	On return the msg structure contains the scatter/gather array passed in the
976  *	vec argument. The array is modified so that it consists of the unfilled
977  *	portion of the original array.
978  *
979  *	The returned value is the total number of bytes received, or an error.
980  */
981 
982 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
983 		   struct kvec *vec, size_t num, size_t size, int flags)
984 {
985 	msg->msg_control_is_user = false;
986 	iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
987 	return sock_recvmsg(sock, msg, flags);
988 }
989 EXPORT_SYMBOL(kernel_recvmsg);
990 
991 static ssize_t sock_sendpage(struct file *file, struct page *page,
992 			     int offset, size_t size, loff_t *ppos, int more)
993 {
994 	struct socket *sock;
995 	int flags;
996 
997 	sock = file->private_data;
998 
999 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1000 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1001 	flags |= more;
1002 
1003 	return kernel_sendpage(sock, page, offset, size, flags);
1004 }
1005 
1006 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1007 				struct pipe_inode_info *pipe, size_t len,
1008 				unsigned int flags)
1009 {
1010 	struct socket *sock = file->private_data;
1011 
1012 	if (unlikely(!sock->ops->splice_read))
1013 		return generic_file_splice_read(file, ppos, pipe, len, flags);
1014 
1015 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1016 }
1017 
1018 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1019 {
1020 	struct file *file = iocb->ki_filp;
1021 	struct socket *sock = file->private_data;
1022 	struct msghdr msg = {.msg_iter = *to,
1023 			     .msg_iocb = iocb};
1024 	ssize_t res;
1025 
1026 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1027 		msg.msg_flags = MSG_DONTWAIT;
1028 
1029 	if (iocb->ki_pos != 0)
1030 		return -ESPIPE;
1031 
1032 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1033 		return 0;
1034 
1035 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1036 	*to = msg.msg_iter;
1037 	return res;
1038 }
1039 
1040 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1041 {
1042 	struct file *file = iocb->ki_filp;
1043 	struct socket *sock = file->private_data;
1044 	struct msghdr msg = {.msg_iter = *from,
1045 			     .msg_iocb = iocb};
1046 	ssize_t res;
1047 
1048 	if (iocb->ki_pos != 0)
1049 		return -ESPIPE;
1050 
1051 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1052 		msg.msg_flags = MSG_DONTWAIT;
1053 
1054 	if (sock->type == SOCK_SEQPACKET)
1055 		msg.msg_flags |= MSG_EOR;
1056 
1057 	res = sock_sendmsg(sock, &msg);
1058 	*from = msg.msg_iter;
1059 	return res;
1060 }
1061 
1062 /*
1063  * Atomic setting of ioctl hooks to avoid race
1064  * with module unload.
1065  */
1066 
1067 static DEFINE_MUTEX(br_ioctl_mutex);
1068 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1069 			    unsigned int cmd, struct ifreq *ifr,
1070 			    void __user *uarg);
1071 
1072 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1073 			     unsigned int cmd, struct ifreq *ifr,
1074 			     void __user *uarg))
1075 {
1076 	mutex_lock(&br_ioctl_mutex);
1077 	br_ioctl_hook = hook;
1078 	mutex_unlock(&br_ioctl_mutex);
1079 }
1080 EXPORT_SYMBOL(brioctl_set);
1081 
1082 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1083 		  struct ifreq *ifr, void __user *uarg)
1084 {
1085 	int err = -ENOPKG;
1086 
1087 	if (!br_ioctl_hook)
1088 		request_module("bridge");
1089 
1090 	mutex_lock(&br_ioctl_mutex);
1091 	if (br_ioctl_hook)
1092 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1093 	mutex_unlock(&br_ioctl_mutex);
1094 
1095 	return err;
1096 }
1097 
1098 static DEFINE_MUTEX(vlan_ioctl_mutex);
1099 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1100 
1101 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1102 {
1103 	mutex_lock(&vlan_ioctl_mutex);
1104 	vlan_ioctl_hook = hook;
1105 	mutex_unlock(&vlan_ioctl_mutex);
1106 }
1107 EXPORT_SYMBOL(vlan_ioctl_set);
1108 
1109 static long sock_do_ioctl(struct net *net, struct socket *sock,
1110 			  unsigned int cmd, unsigned long arg)
1111 {
1112 	struct ifreq ifr;
1113 	bool need_copyout;
1114 	int err;
1115 	void __user *argp = (void __user *)arg;
1116 	void __user *data;
1117 
1118 	err = sock->ops->ioctl(sock, cmd, arg);
1119 
1120 	/*
1121 	 * If this ioctl is unknown try to hand it down
1122 	 * to the NIC driver.
1123 	 */
1124 	if (err != -ENOIOCTLCMD)
1125 		return err;
1126 
1127 	if (get_user_ifreq(&ifr, &data, argp))
1128 		return -EFAULT;
1129 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1130 	if (!err && need_copyout)
1131 		if (put_user_ifreq(&ifr, argp))
1132 			return -EFAULT;
1133 
1134 	return err;
1135 }
1136 
1137 /*
1138  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1139  *	what to do with it - that's up to the protocol still.
1140  */
1141 
1142 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1143 {
1144 	struct socket *sock;
1145 	struct sock *sk;
1146 	void __user *argp = (void __user *)arg;
1147 	int pid, err;
1148 	struct net *net;
1149 
1150 	sock = file->private_data;
1151 	sk = sock->sk;
1152 	net = sock_net(sk);
1153 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1154 		struct ifreq ifr;
1155 		void __user *data;
1156 		bool need_copyout;
1157 		if (get_user_ifreq(&ifr, &data, argp))
1158 			return -EFAULT;
1159 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1160 		if (!err && need_copyout)
1161 			if (put_user_ifreq(&ifr, argp))
1162 				return -EFAULT;
1163 	} else
1164 #ifdef CONFIG_WEXT_CORE
1165 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1166 		err = wext_handle_ioctl(net, cmd, argp);
1167 	} else
1168 #endif
1169 		switch (cmd) {
1170 		case FIOSETOWN:
1171 		case SIOCSPGRP:
1172 			err = -EFAULT;
1173 			if (get_user(pid, (int __user *)argp))
1174 				break;
1175 			err = f_setown(sock->file, pid, 1);
1176 			break;
1177 		case FIOGETOWN:
1178 		case SIOCGPGRP:
1179 			err = put_user(f_getown(sock->file),
1180 				       (int __user *)argp);
1181 			break;
1182 		case SIOCGIFBR:
1183 		case SIOCSIFBR:
1184 		case SIOCBRADDBR:
1185 		case SIOCBRDELBR:
1186 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1187 			break;
1188 		case SIOCGIFVLAN:
1189 		case SIOCSIFVLAN:
1190 			err = -ENOPKG;
1191 			if (!vlan_ioctl_hook)
1192 				request_module("8021q");
1193 
1194 			mutex_lock(&vlan_ioctl_mutex);
1195 			if (vlan_ioctl_hook)
1196 				err = vlan_ioctl_hook(net, argp);
1197 			mutex_unlock(&vlan_ioctl_mutex);
1198 			break;
1199 		case SIOCGSKNS:
1200 			err = -EPERM;
1201 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1202 				break;
1203 
1204 			err = open_related_ns(&net->ns, get_net_ns);
1205 			break;
1206 		case SIOCGSTAMP_OLD:
1207 		case SIOCGSTAMPNS_OLD:
1208 			if (!sock->ops->gettstamp) {
1209 				err = -ENOIOCTLCMD;
1210 				break;
1211 			}
1212 			err = sock->ops->gettstamp(sock, argp,
1213 						   cmd == SIOCGSTAMP_OLD,
1214 						   !IS_ENABLED(CONFIG_64BIT));
1215 			break;
1216 		case SIOCGSTAMP_NEW:
1217 		case SIOCGSTAMPNS_NEW:
1218 			if (!sock->ops->gettstamp) {
1219 				err = -ENOIOCTLCMD;
1220 				break;
1221 			}
1222 			err = sock->ops->gettstamp(sock, argp,
1223 						   cmd == SIOCGSTAMP_NEW,
1224 						   false);
1225 			break;
1226 
1227 		case SIOCGIFCONF:
1228 			err = dev_ifconf(net, argp);
1229 			break;
1230 
1231 		default:
1232 			err = sock_do_ioctl(net, sock, cmd, arg);
1233 			break;
1234 		}
1235 	return err;
1236 }
1237 
1238 /**
1239  *	sock_create_lite - creates a socket
1240  *	@family: protocol family (AF_INET, ...)
1241  *	@type: communication type (SOCK_STREAM, ...)
1242  *	@protocol: protocol (0, ...)
1243  *	@res: new socket
1244  *
1245  *	Creates a new socket and assigns it to @res, passing through LSM.
1246  *	The new socket initialization is not complete, see kernel_accept().
1247  *	Returns 0 or an error. On failure @res is set to %NULL.
1248  *	This function internally uses GFP_KERNEL.
1249  */
1250 
1251 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1252 {
1253 	int err;
1254 	struct socket *sock = NULL;
1255 
1256 	err = security_socket_create(family, type, protocol, 1);
1257 	if (err)
1258 		goto out;
1259 
1260 	sock = sock_alloc();
1261 	if (!sock) {
1262 		err = -ENOMEM;
1263 		goto out;
1264 	}
1265 
1266 	sock->type = type;
1267 	err = security_socket_post_create(sock, family, type, protocol, 1);
1268 	if (err)
1269 		goto out_release;
1270 
1271 out:
1272 	*res = sock;
1273 	return err;
1274 out_release:
1275 	sock_release(sock);
1276 	sock = NULL;
1277 	goto out;
1278 }
1279 EXPORT_SYMBOL(sock_create_lite);
1280 
1281 /* No kernel lock held - perfect */
1282 static __poll_t sock_poll(struct file *file, poll_table *wait)
1283 {
1284 	struct socket *sock = file->private_data;
1285 	__poll_t events = poll_requested_events(wait), flag = 0;
1286 
1287 	if (!sock->ops->poll)
1288 		return 0;
1289 
1290 	if (sk_can_busy_loop(sock->sk)) {
1291 		/* poll once if requested by the syscall */
1292 		if (events & POLL_BUSY_LOOP)
1293 			sk_busy_loop(sock->sk, 1);
1294 
1295 		/* if this socket can poll_ll, tell the system call */
1296 		flag = POLL_BUSY_LOOP;
1297 	}
1298 
1299 	return sock->ops->poll(file, sock, wait) | flag;
1300 }
1301 
1302 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1303 {
1304 	struct socket *sock = file->private_data;
1305 
1306 	return sock->ops->mmap(file, sock, vma);
1307 }
1308 
1309 static int sock_close(struct inode *inode, struct file *filp)
1310 {
1311 	__sock_release(SOCKET_I(inode), inode);
1312 	return 0;
1313 }
1314 
1315 /*
1316  *	Update the socket async list
1317  *
1318  *	Fasync_list locking strategy.
1319  *
1320  *	1. fasync_list is modified only under process context socket lock
1321  *	   i.e. under semaphore.
1322  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1323  *	   or under socket lock
1324  */
1325 
1326 static int sock_fasync(int fd, struct file *filp, int on)
1327 {
1328 	struct socket *sock = filp->private_data;
1329 	struct sock *sk = sock->sk;
1330 	struct socket_wq *wq = &sock->wq;
1331 
1332 	if (sk == NULL)
1333 		return -EINVAL;
1334 
1335 	lock_sock(sk);
1336 	fasync_helper(fd, filp, on, &wq->fasync_list);
1337 
1338 	if (!wq->fasync_list)
1339 		sock_reset_flag(sk, SOCK_FASYNC);
1340 	else
1341 		sock_set_flag(sk, SOCK_FASYNC);
1342 
1343 	release_sock(sk);
1344 	return 0;
1345 }
1346 
1347 /* This function may be called only under rcu_lock */
1348 
1349 int sock_wake_async(struct socket_wq *wq, int how, int band)
1350 {
1351 	if (!wq || !wq->fasync_list)
1352 		return -1;
1353 
1354 	switch (how) {
1355 	case SOCK_WAKE_WAITD:
1356 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1357 			break;
1358 		goto call_kill;
1359 	case SOCK_WAKE_SPACE:
1360 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1361 			break;
1362 		fallthrough;
1363 	case SOCK_WAKE_IO:
1364 call_kill:
1365 		kill_fasync(&wq->fasync_list, SIGIO, band);
1366 		break;
1367 	case SOCK_WAKE_URG:
1368 		kill_fasync(&wq->fasync_list, SIGURG, band);
1369 	}
1370 
1371 	return 0;
1372 }
1373 EXPORT_SYMBOL(sock_wake_async);
1374 
1375 /**
1376  *	__sock_create - creates a socket
1377  *	@net: net namespace
1378  *	@family: protocol family (AF_INET, ...)
1379  *	@type: communication type (SOCK_STREAM, ...)
1380  *	@protocol: protocol (0, ...)
1381  *	@res: new socket
1382  *	@kern: boolean for kernel space sockets
1383  *
1384  *	Creates a new socket and assigns it to @res, passing through LSM.
1385  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1386  *	be set to true if the socket resides in kernel space.
1387  *	This function internally uses GFP_KERNEL.
1388  */
1389 
1390 int __sock_create(struct net *net, int family, int type, int protocol,
1391 			 struct socket **res, int kern)
1392 {
1393 	int err;
1394 	struct socket *sock;
1395 	const struct net_proto_family *pf;
1396 
1397 	/*
1398 	 *      Check protocol is in range
1399 	 */
1400 	if (family < 0 || family >= NPROTO)
1401 		return -EAFNOSUPPORT;
1402 	if (type < 0 || type >= SOCK_MAX)
1403 		return -EINVAL;
1404 
1405 	/* Compatibility.
1406 
1407 	   This uglymoron is moved from INET layer to here to avoid
1408 	   deadlock in module load.
1409 	 */
1410 	if (family == PF_INET && type == SOCK_PACKET) {
1411 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1412 			     current->comm);
1413 		family = PF_PACKET;
1414 	}
1415 
1416 	err = security_socket_create(family, type, protocol, kern);
1417 	if (err)
1418 		return err;
1419 
1420 	/*
1421 	 *	Allocate the socket and allow the family to set things up. if
1422 	 *	the protocol is 0, the family is instructed to select an appropriate
1423 	 *	default.
1424 	 */
1425 	sock = sock_alloc();
1426 	if (!sock) {
1427 		net_warn_ratelimited("socket: no more sockets\n");
1428 		return -ENFILE;	/* Not exactly a match, but its the
1429 				   closest posix thing */
1430 	}
1431 
1432 	sock->type = type;
1433 
1434 #ifdef CONFIG_MODULES
1435 	/* Attempt to load a protocol module if the find failed.
1436 	 *
1437 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1438 	 * requested real, full-featured networking support upon configuration.
1439 	 * Otherwise module support will break!
1440 	 */
1441 	if (rcu_access_pointer(net_families[family]) == NULL)
1442 		request_module("net-pf-%d", family);
1443 #endif
1444 
1445 	rcu_read_lock();
1446 	pf = rcu_dereference(net_families[family]);
1447 	err = -EAFNOSUPPORT;
1448 	if (!pf)
1449 		goto out_release;
1450 
1451 	/*
1452 	 * We will call the ->create function, that possibly is in a loadable
1453 	 * module, so we have to bump that loadable module refcnt first.
1454 	 */
1455 	if (!try_module_get(pf->owner))
1456 		goto out_release;
1457 
1458 	/* Now protected by module ref count */
1459 	rcu_read_unlock();
1460 
1461 	err = pf->create(net, sock, protocol, kern);
1462 	if (err < 0)
1463 		goto out_module_put;
1464 
1465 	/*
1466 	 * Now to bump the refcnt of the [loadable] module that owns this
1467 	 * socket at sock_release time we decrement its refcnt.
1468 	 */
1469 	if (!try_module_get(sock->ops->owner))
1470 		goto out_module_busy;
1471 
1472 	/*
1473 	 * Now that we're done with the ->create function, the [loadable]
1474 	 * module can have its refcnt decremented
1475 	 */
1476 	module_put(pf->owner);
1477 	err = security_socket_post_create(sock, family, type, protocol, kern);
1478 	if (err)
1479 		goto out_sock_release;
1480 	*res = sock;
1481 
1482 	return 0;
1483 
1484 out_module_busy:
1485 	err = -EAFNOSUPPORT;
1486 out_module_put:
1487 	sock->ops = NULL;
1488 	module_put(pf->owner);
1489 out_sock_release:
1490 	sock_release(sock);
1491 	return err;
1492 
1493 out_release:
1494 	rcu_read_unlock();
1495 	goto out_sock_release;
1496 }
1497 EXPORT_SYMBOL(__sock_create);
1498 
1499 /**
1500  *	sock_create - creates a socket
1501  *	@family: protocol family (AF_INET, ...)
1502  *	@type: communication type (SOCK_STREAM, ...)
1503  *	@protocol: protocol (0, ...)
1504  *	@res: new socket
1505  *
1506  *	A wrapper around __sock_create().
1507  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1508  */
1509 
1510 int sock_create(int family, int type, int protocol, struct socket **res)
1511 {
1512 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1513 }
1514 EXPORT_SYMBOL(sock_create);
1515 
1516 /**
1517  *	sock_create_kern - creates a socket (kernel space)
1518  *	@net: net namespace
1519  *	@family: protocol family (AF_INET, ...)
1520  *	@type: communication type (SOCK_STREAM, ...)
1521  *	@protocol: protocol (0, ...)
1522  *	@res: new socket
1523  *
1524  *	A wrapper around __sock_create().
1525  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1526  */
1527 
1528 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1529 {
1530 	return __sock_create(net, family, type, protocol, res, 1);
1531 }
1532 EXPORT_SYMBOL(sock_create_kern);
1533 
1534 int __sys_socket(int family, int type, int protocol)
1535 {
1536 	int retval;
1537 	struct socket *sock;
1538 	int flags;
1539 
1540 	/* Check the SOCK_* constants for consistency.  */
1541 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1542 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1543 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1544 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1545 
1546 	flags = type & ~SOCK_TYPE_MASK;
1547 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1548 		return -EINVAL;
1549 	type &= SOCK_TYPE_MASK;
1550 
1551 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1552 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1553 
1554 	retval = sock_create(family, type, protocol, &sock);
1555 	if (retval < 0)
1556 		return retval;
1557 
1558 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1559 }
1560 
1561 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1562 {
1563 	return __sys_socket(family, type, protocol);
1564 }
1565 
1566 /*
1567  *	Create a pair of connected sockets.
1568  */
1569 
1570 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1571 {
1572 	struct socket *sock1, *sock2;
1573 	int fd1, fd2, err;
1574 	struct file *newfile1, *newfile2;
1575 	int flags;
1576 
1577 	flags = type & ~SOCK_TYPE_MASK;
1578 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1579 		return -EINVAL;
1580 	type &= SOCK_TYPE_MASK;
1581 
1582 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1583 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1584 
1585 	/*
1586 	 * reserve descriptors and make sure we won't fail
1587 	 * to return them to userland.
1588 	 */
1589 	fd1 = get_unused_fd_flags(flags);
1590 	if (unlikely(fd1 < 0))
1591 		return fd1;
1592 
1593 	fd2 = get_unused_fd_flags(flags);
1594 	if (unlikely(fd2 < 0)) {
1595 		put_unused_fd(fd1);
1596 		return fd2;
1597 	}
1598 
1599 	err = put_user(fd1, &usockvec[0]);
1600 	if (err)
1601 		goto out;
1602 
1603 	err = put_user(fd2, &usockvec[1]);
1604 	if (err)
1605 		goto out;
1606 
1607 	/*
1608 	 * Obtain the first socket and check if the underlying protocol
1609 	 * supports the socketpair call.
1610 	 */
1611 
1612 	err = sock_create(family, type, protocol, &sock1);
1613 	if (unlikely(err < 0))
1614 		goto out;
1615 
1616 	err = sock_create(family, type, protocol, &sock2);
1617 	if (unlikely(err < 0)) {
1618 		sock_release(sock1);
1619 		goto out;
1620 	}
1621 
1622 	err = security_socket_socketpair(sock1, sock2);
1623 	if (unlikely(err)) {
1624 		sock_release(sock2);
1625 		sock_release(sock1);
1626 		goto out;
1627 	}
1628 
1629 	err = sock1->ops->socketpair(sock1, sock2);
1630 	if (unlikely(err < 0)) {
1631 		sock_release(sock2);
1632 		sock_release(sock1);
1633 		goto out;
1634 	}
1635 
1636 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1637 	if (IS_ERR(newfile1)) {
1638 		err = PTR_ERR(newfile1);
1639 		sock_release(sock2);
1640 		goto out;
1641 	}
1642 
1643 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1644 	if (IS_ERR(newfile2)) {
1645 		err = PTR_ERR(newfile2);
1646 		fput(newfile1);
1647 		goto out;
1648 	}
1649 
1650 	audit_fd_pair(fd1, fd2);
1651 
1652 	fd_install(fd1, newfile1);
1653 	fd_install(fd2, newfile2);
1654 	return 0;
1655 
1656 out:
1657 	put_unused_fd(fd2);
1658 	put_unused_fd(fd1);
1659 	return err;
1660 }
1661 
1662 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1663 		int __user *, usockvec)
1664 {
1665 	return __sys_socketpair(family, type, protocol, usockvec);
1666 }
1667 
1668 /*
1669  *	Bind a name to a socket. Nothing much to do here since it's
1670  *	the protocol's responsibility to handle the local address.
1671  *
1672  *	We move the socket address to kernel space before we call
1673  *	the protocol layer (having also checked the address is ok).
1674  */
1675 
1676 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1677 {
1678 	struct socket *sock;
1679 	struct sockaddr_storage address;
1680 	int err, fput_needed;
1681 
1682 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1683 	if (sock) {
1684 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1685 		if (!err) {
1686 			err = security_socket_bind(sock,
1687 						   (struct sockaddr *)&address,
1688 						   addrlen);
1689 			if (!err)
1690 				err = sock->ops->bind(sock,
1691 						      (struct sockaddr *)
1692 						      &address, addrlen);
1693 		}
1694 		fput_light(sock->file, fput_needed);
1695 	}
1696 	return err;
1697 }
1698 
1699 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1700 {
1701 	return __sys_bind(fd, umyaddr, addrlen);
1702 }
1703 
1704 /*
1705  *	Perform a listen. Basically, we allow the protocol to do anything
1706  *	necessary for a listen, and if that works, we mark the socket as
1707  *	ready for listening.
1708  */
1709 
1710 int __sys_listen(int fd, int backlog)
1711 {
1712 	struct socket *sock;
1713 	int err, fput_needed;
1714 	int somaxconn;
1715 
1716 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1717 	if (sock) {
1718 		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1719 		if ((unsigned int)backlog > somaxconn)
1720 			backlog = somaxconn;
1721 
1722 		err = security_socket_listen(sock, backlog);
1723 		if (!err)
1724 			err = sock->ops->listen(sock, backlog);
1725 
1726 		fput_light(sock->file, fput_needed);
1727 	}
1728 	return err;
1729 }
1730 
1731 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1732 {
1733 	return __sys_listen(fd, backlog);
1734 }
1735 
1736 int __sys_accept4_file(struct file *file, unsigned file_flags,
1737 		       struct sockaddr __user *upeer_sockaddr,
1738 		       int __user *upeer_addrlen, int flags,
1739 		       unsigned long nofile)
1740 {
1741 	struct socket *sock, *newsock;
1742 	struct file *newfile;
1743 	int err, len, newfd;
1744 	struct sockaddr_storage address;
1745 
1746 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1747 		return -EINVAL;
1748 
1749 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1750 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1751 
1752 	sock = sock_from_file(file);
1753 	if (!sock) {
1754 		err = -ENOTSOCK;
1755 		goto out;
1756 	}
1757 
1758 	err = -ENFILE;
1759 	newsock = sock_alloc();
1760 	if (!newsock)
1761 		goto out;
1762 
1763 	newsock->type = sock->type;
1764 	newsock->ops = sock->ops;
1765 
1766 	/*
1767 	 * We don't need try_module_get here, as the listening socket (sock)
1768 	 * has the protocol module (sock->ops->owner) held.
1769 	 */
1770 	__module_get(newsock->ops->owner);
1771 
1772 	newfd = __get_unused_fd_flags(flags, nofile);
1773 	if (unlikely(newfd < 0)) {
1774 		err = newfd;
1775 		sock_release(newsock);
1776 		goto out;
1777 	}
1778 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1779 	if (IS_ERR(newfile)) {
1780 		err = PTR_ERR(newfile);
1781 		put_unused_fd(newfd);
1782 		goto out;
1783 	}
1784 
1785 	err = security_socket_accept(sock, newsock);
1786 	if (err)
1787 		goto out_fd;
1788 
1789 	err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1790 					false);
1791 	if (err < 0)
1792 		goto out_fd;
1793 
1794 	if (upeer_sockaddr) {
1795 		len = newsock->ops->getname(newsock,
1796 					(struct sockaddr *)&address, 2);
1797 		if (len < 0) {
1798 			err = -ECONNABORTED;
1799 			goto out_fd;
1800 		}
1801 		err = move_addr_to_user(&address,
1802 					len, upeer_sockaddr, upeer_addrlen);
1803 		if (err < 0)
1804 			goto out_fd;
1805 	}
1806 
1807 	/* File flags are not inherited via accept() unlike another OSes. */
1808 
1809 	fd_install(newfd, newfile);
1810 	err = newfd;
1811 out:
1812 	return err;
1813 out_fd:
1814 	fput(newfile);
1815 	put_unused_fd(newfd);
1816 	goto out;
1817 
1818 }
1819 
1820 /*
1821  *	For accept, we attempt to create a new socket, set up the link
1822  *	with the client, wake up the client, then return the new
1823  *	connected fd. We collect the address of the connector in kernel
1824  *	space and move it to user at the very end. This is unclean because
1825  *	we open the socket then return an error.
1826  *
1827  *	1003.1g adds the ability to recvmsg() to query connection pending
1828  *	status to recvmsg. We need to add that support in a way thats
1829  *	clean when we restructure accept also.
1830  */
1831 
1832 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1833 		  int __user *upeer_addrlen, int flags)
1834 {
1835 	int ret = -EBADF;
1836 	struct fd f;
1837 
1838 	f = fdget(fd);
1839 	if (f.file) {
1840 		ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1841 						upeer_addrlen, flags,
1842 						rlimit(RLIMIT_NOFILE));
1843 		fdput(f);
1844 	}
1845 
1846 	return ret;
1847 }
1848 
1849 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1850 		int __user *, upeer_addrlen, int, flags)
1851 {
1852 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1853 }
1854 
1855 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1856 		int __user *, upeer_addrlen)
1857 {
1858 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1859 }
1860 
1861 /*
1862  *	Attempt to connect to a socket with the server address.  The address
1863  *	is in user space so we verify it is OK and move it to kernel space.
1864  *
1865  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1866  *	break bindings
1867  *
1868  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1869  *	other SEQPACKET protocols that take time to connect() as it doesn't
1870  *	include the -EINPROGRESS status for such sockets.
1871  */
1872 
1873 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1874 		       int addrlen, int file_flags)
1875 {
1876 	struct socket *sock;
1877 	int err;
1878 
1879 	sock = sock_from_file(file);
1880 	if (!sock) {
1881 		err = -ENOTSOCK;
1882 		goto out;
1883 	}
1884 
1885 	err =
1886 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1887 	if (err)
1888 		goto out;
1889 
1890 	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1891 				 sock->file->f_flags | file_flags);
1892 out:
1893 	return err;
1894 }
1895 
1896 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1897 {
1898 	int ret = -EBADF;
1899 	struct fd f;
1900 
1901 	f = fdget(fd);
1902 	if (f.file) {
1903 		struct sockaddr_storage address;
1904 
1905 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1906 		if (!ret)
1907 			ret = __sys_connect_file(f.file, &address, addrlen, 0);
1908 		fdput(f);
1909 	}
1910 
1911 	return ret;
1912 }
1913 
1914 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1915 		int, addrlen)
1916 {
1917 	return __sys_connect(fd, uservaddr, addrlen);
1918 }
1919 
1920 /*
1921  *	Get the local address ('name') of a socket object. Move the obtained
1922  *	name to user space.
1923  */
1924 
1925 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1926 		      int __user *usockaddr_len)
1927 {
1928 	struct socket *sock;
1929 	struct sockaddr_storage address;
1930 	int err, fput_needed;
1931 
1932 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1933 	if (!sock)
1934 		goto out;
1935 
1936 	err = security_socket_getsockname(sock);
1937 	if (err)
1938 		goto out_put;
1939 
1940 	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1941 	if (err < 0)
1942 		goto out_put;
1943         /* "err" is actually length in this case */
1944 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1945 
1946 out_put:
1947 	fput_light(sock->file, fput_needed);
1948 out:
1949 	return err;
1950 }
1951 
1952 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1953 		int __user *, usockaddr_len)
1954 {
1955 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1956 }
1957 
1958 /*
1959  *	Get the remote address ('name') of a socket object. Move the obtained
1960  *	name to user space.
1961  */
1962 
1963 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1964 		      int __user *usockaddr_len)
1965 {
1966 	struct socket *sock;
1967 	struct sockaddr_storage address;
1968 	int err, fput_needed;
1969 
1970 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1971 	if (sock != NULL) {
1972 		err = security_socket_getpeername(sock);
1973 		if (err) {
1974 			fput_light(sock->file, fput_needed);
1975 			return err;
1976 		}
1977 
1978 		err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1979 		if (err >= 0)
1980 			/* "err" is actually length in this case */
1981 			err = move_addr_to_user(&address, err, usockaddr,
1982 						usockaddr_len);
1983 		fput_light(sock->file, fput_needed);
1984 	}
1985 	return err;
1986 }
1987 
1988 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1989 		int __user *, usockaddr_len)
1990 {
1991 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1992 }
1993 
1994 /*
1995  *	Send a datagram to a given address. We move the address into kernel
1996  *	space and check the user space data area is readable before invoking
1997  *	the protocol.
1998  */
1999 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2000 		 struct sockaddr __user *addr,  int addr_len)
2001 {
2002 	struct socket *sock;
2003 	struct sockaddr_storage address;
2004 	int err;
2005 	struct msghdr msg;
2006 	struct iovec iov;
2007 	int fput_needed;
2008 
2009 	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
2010 	if (unlikely(err))
2011 		return err;
2012 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2013 	if (!sock)
2014 		goto out;
2015 
2016 	msg.msg_name = NULL;
2017 	msg.msg_control = NULL;
2018 	msg.msg_controllen = 0;
2019 	msg.msg_namelen = 0;
2020 	if (addr) {
2021 		err = move_addr_to_kernel(addr, addr_len, &address);
2022 		if (err < 0)
2023 			goto out_put;
2024 		msg.msg_name = (struct sockaddr *)&address;
2025 		msg.msg_namelen = addr_len;
2026 	}
2027 	if (sock->file->f_flags & O_NONBLOCK)
2028 		flags |= MSG_DONTWAIT;
2029 	msg.msg_flags = flags;
2030 	err = sock_sendmsg(sock, &msg);
2031 
2032 out_put:
2033 	fput_light(sock->file, fput_needed);
2034 out:
2035 	return err;
2036 }
2037 
2038 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2039 		unsigned int, flags, struct sockaddr __user *, addr,
2040 		int, addr_len)
2041 {
2042 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2043 }
2044 
2045 /*
2046  *	Send a datagram down a socket.
2047  */
2048 
2049 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2050 		unsigned int, flags)
2051 {
2052 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2053 }
2054 
2055 /*
2056  *	Receive a frame from the socket and optionally record the address of the
2057  *	sender. We verify the buffers are writable and if needed move the
2058  *	sender address from kernel to user space.
2059  */
2060 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2061 		   struct sockaddr __user *addr, int __user *addr_len)
2062 {
2063 	struct socket *sock;
2064 	struct iovec iov;
2065 	struct msghdr msg;
2066 	struct sockaddr_storage address;
2067 	int err, err2;
2068 	int fput_needed;
2069 
2070 	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2071 	if (unlikely(err))
2072 		return err;
2073 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2074 	if (!sock)
2075 		goto out;
2076 
2077 	msg.msg_control = NULL;
2078 	msg.msg_controllen = 0;
2079 	/* Save some cycles and don't copy the address if not needed */
2080 	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2081 	/* We assume all kernel code knows the size of sockaddr_storage */
2082 	msg.msg_namelen = 0;
2083 	msg.msg_iocb = NULL;
2084 	msg.msg_flags = 0;
2085 	if (sock->file->f_flags & O_NONBLOCK)
2086 		flags |= MSG_DONTWAIT;
2087 	err = sock_recvmsg(sock, &msg, flags);
2088 
2089 	if (err >= 0 && addr != NULL) {
2090 		err2 = move_addr_to_user(&address,
2091 					 msg.msg_namelen, addr, addr_len);
2092 		if (err2 < 0)
2093 			err = err2;
2094 	}
2095 
2096 	fput_light(sock->file, fput_needed);
2097 out:
2098 	return err;
2099 }
2100 
2101 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2102 		unsigned int, flags, struct sockaddr __user *, addr,
2103 		int __user *, addr_len)
2104 {
2105 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2106 }
2107 
2108 /*
2109  *	Receive a datagram from a socket.
2110  */
2111 
2112 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2113 		unsigned int, flags)
2114 {
2115 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2116 }
2117 
2118 static bool sock_use_custom_sol_socket(const struct socket *sock)
2119 {
2120 	const struct sock *sk = sock->sk;
2121 
2122 	/* Use sock->ops->setsockopt() for MPTCP */
2123 	return IS_ENABLED(CONFIG_MPTCP) &&
2124 	       sk->sk_protocol == IPPROTO_MPTCP &&
2125 	       sk->sk_type == SOCK_STREAM &&
2126 	       (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2127 }
2128 
2129 /*
2130  *	Set a socket option. Because we don't know the option lengths we have
2131  *	to pass the user mode parameter for the protocols to sort out.
2132  */
2133 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2134 		int optlen)
2135 {
2136 	sockptr_t optval = USER_SOCKPTR(user_optval);
2137 	char *kernel_optval = NULL;
2138 	int err, fput_needed;
2139 	struct socket *sock;
2140 
2141 	if (optlen < 0)
2142 		return -EINVAL;
2143 
2144 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2145 	if (!sock)
2146 		return err;
2147 
2148 	err = security_socket_setsockopt(sock, level, optname);
2149 	if (err)
2150 		goto out_put;
2151 
2152 	if (!in_compat_syscall())
2153 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2154 						     user_optval, &optlen,
2155 						     &kernel_optval);
2156 	if (err < 0)
2157 		goto out_put;
2158 	if (err > 0) {
2159 		err = 0;
2160 		goto out_put;
2161 	}
2162 
2163 	if (kernel_optval)
2164 		optval = KERNEL_SOCKPTR(kernel_optval);
2165 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2166 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2167 	else if (unlikely(!sock->ops->setsockopt))
2168 		err = -EOPNOTSUPP;
2169 	else
2170 		err = sock->ops->setsockopt(sock, level, optname, optval,
2171 					    optlen);
2172 	kfree(kernel_optval);
2173 out_put:
2174 	fput_light(sock->file, fput_needed);
2175 	return err;
2176 }
2177 
2178 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2179 		char __user *, optval, int, optlen)
2180 {
2181 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2182 }
2183 
2184 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2185 							 int optname));
2186 
2187 /*
2188  *	Get a socket option. Because we don't know the option lengths we have
2189  *	to pass a user mode parameter for the protocols to sort out.
2190  */
2191 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2192 		int __user *optlen)
2193 {
2194 	int err, fput_needed;
2195 	struct socket *sock;
2196 	int max_optlen;
2197 
2198 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2199 	if (!sock)
2200 		return err;
2201 
2202 	err = security_socket_getsockopt(sock, level, optname);
2203 	if (err)
2204 		goto out_put;
2205 
2206 	if (!in_compat_syscall())
2207 		max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2208 
2209 	if (level == SOL_SOCKET)
2210 		err = sock_getsockopt(sock, level, optname, optval, optlen);
2211 	else if (unlikely(!sock->ops->getsockopt))
2212 		err = -EOPNOTSUPP;
2213 	else
2214 		err = sock->ops->getsockopt(sock, level, optname, optval,
2215 					    optlen);
2216 
2217 	if (!in_compat_syscall())
2218 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2219 						     optval, optlen, max_optlen,
2220 						     err);
2221 out_put:
2222 	fput_light(sock->file, fput_needed);
2223 	return err;
2224 }
2225 
2226 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2227 		char __user *, optval, int __user *, optlen)
2228 {
2229 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2230 }
2231 
2232 /*
2233  *	Shutdown a socket.
2234  */
2235 
2236 int __sys_shutdown_sock(struct socket *sock, int how)
2237 {
2238 	int err;
2239 
2240 	err = security_socket_shutdown(sock, how);
2241 	if (!err)
2242 		err = sock->ops->shutdown(sock, how);
2243 
2244 	return err;
2245 }
2246 
2247 int __sys_shutdown(int fd, int how)
2248 {
2249 	int err, fput_needed;
2250 	struct socket *sock;
2251 
2252 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2253 	if (sock != NULL) {
2254 		err = __sys_shutdown_sock(sock, how);
2255 		fput_light(sock->file, fput_needed);
2256 	}
2257 	return err;
2258 }
2259 
2260 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2261 {
2262 	return __sys_shutdown(fd, how);
2263 }
2264 
2265 /* A couple of helpful macros for getting the address of the 32/64 bit
2266  * fields which are the same type (int / unsigned) on our platforms.
2267  */
2268 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2269 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2270 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2271 
2272 struct used_address {
2273 	struct sockaddr_storage name;
2274 	unsigned int name_len;
2275 };
2276 
2277 int __copy_msghdr_from_user(struct msghdr *kmsg,
2278 			    struct user_msghdr __user *umsg,
2279 			    struct sockaddr __user **save_addr,
2280 			    struct iovec __user **uiov, size_t *nsegs)
2281 {
2282 	struct user_msghdr msg;
2283 	ssize_t err;
2284 
2285 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2286 		return -EFAULT;
2287 
2288 	kmsg->msg_control_is_user = true;
2289 	kmsg->msg_control_user = msg.msg_control;
2290 	kmsg->msg_controllen = msg.msg_controllen;
2291 	kmsg->msg_flags = msg.msg_flags;
2292 
2293 	kmsg->msg_namelen = msg.msg_namelen;
2294 	if (!msg.msg_name)
2295 		kmsg->msg_namelen = 0;
2296 
2297 	if (kmsg->msg_namelen < 0)
2298 		return -EINVAL;
2299 
2300 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2301 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2302 
2303 	if (save_addr)
2304 		*save_addr = msg.msg_name;
2305 
2306 	if (msg.msg_name && kmsg->msg_namelen) {
2307 		if (!save_addr) {
2308 			err = move_addr_to_kernel(msg.msg_name,
2309 						  kmsg->msg_namelen,
2310 						  kmsg->msg_name);
2311 			if (err < 0)
2312 				return err;
2313 		}
2314 	} else {
2315 		kmsg->msg_name = NULL;
2316 		kmsg->msg_namelen = 0;
2317 	}
2318 
2319 	if (msg.msg_iovlen > UIO_MAXIOV)
2320 		return -EMSGSIZE;
2321 
2322 	kmsg->msg_iocb = NULL;
2323 	*uiov = msg.msg_iov;
2324 	*nsegs = msg.msg_iovlen;
2325 	return 0;
2326 }
2327 
2328 static int copy_msghdr_from_user(struct msghdr *kmsg,
2329 				 struct user_msghdr __user *umsg,
2330 				 struct sockaddr __user **save_addr,
2331 				 struct iovec **iov)
2332 {
2333 	struct user_msghdr msg;
2334 	ssize_t err;
2335 
2336 	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2337 					&msg.msg_iovlen);
2338 	if (err)
2339 		return err;
2340 
2341 	err = import_iovec(save_addr ? READ : WRITE,
2342 			    msg.msg_iov, msg.msg_iovlen,
2343 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2344 	return err < 0 ? err : 0;
2345 }
2346 
2347 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2348 			   unsigned int flags, struct used_address *used_address,
2349 			   unsigned int allowed_msghdr_flags)
2350 {
2351 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2352 				__aligned(sizeof(__kernel_size_t));
2353 	/* 20 is size of ipv6_pktinfo */
2354 	unsigned char *ctl_buf = ctl;
2355 	int ctl_len;
2356 	ssize_t err;
2357 
2358 	err = -ENOBUFS;
2359 
2360 	if (msg_sys->msg_controllen > INT_MAX)
2361 		goto out;
2362 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2363 	ctl_len = msg_sys->msg_controllen;
2364 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2365 		err =
2366 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2367 						     sizeof(ctl));
2368 		if (err)
2369 			goto out;
2370 		ctl_buf = msg_sys->msg_control;
2371 		ctl_len = msg_sys->msg_controllen;
2372 	} else if (ctl_len) {
2373 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2374 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2375 		if (ctl_len > sizeof(ctl)) {
2376 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2377 			if (ctl_buf == NULL)
2378 				goto out;
2379 		}
2380 		err = -EFAULT;
2381 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2382 			goto out_freectl;
2383 		msg_sys->msg_control = ctl_buf;
2384 		msg_sys->msg_control_is_user = false;
2385 	}
2386 	msg_sys->msg_flags = flags;
2387 
2388 	if (sock->file->f_flags & O_NONBLOCK)
2389 		msg_sys->msg_flags |= MSG_DONTWAIT;
2390 	/*
2391 	 * If this is sendmmsg() and current destination address is same as
2392 	 * previously succeeded address, omit asking LSM's decision.
2393 	 * used_address->name_len is initialized to UINT_MAX so that the first
2394 	 * destination address never matches.
2395 	 */
2396 	if (used_address && msg_sys->msg_name &&
2397 	    used_address->name_len == msg_sys->msg_namelen &&
2398 	    !memcmp(&used_address->name, msg_sys->msg_name,
2399 		    used_address->name_len)) {
2400 		err = sock_sendmsg_nosec(sock, msg_sys);
2401 		goto out_freectl;
2402 	}
2403 	err = sock_sendmsg(sock, msg_sys);
2404 	/*
2405 	 * If this is sendmmsg() and sending to current destination address was
2406 	 * successful, remember it.
2407 	 */
2408 	if (used_address && err >= 0) {
2409 		used_address->name_len = msg_sys->msg_namelen;
2410 		if (msg_sys->msg_name)
2411 			memcpy(&used_address->name, msg_sys->msg_name,
2412 			       used_address->name_len);
2413 	}
2414 
2415 out_freectl:
2416 	if (ctl_buf != ctl)
2417 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2418 out:
2419 	return err;
2420 }
2421 
2422 int sendmsg_copy_msghdr(struct msghdr *msg,
2423 			struct user_msghdr __user *umsg, unsigned flags,
2424 			struct iovec **iov)
2425 {
2426 	int err;
2427 
2428 	if (flags & MSG_CMSG_COMPAT) {
2429 		struct compat_msghdr __user *msg_compat;
2430 
2431 		msg_compat = (struct compat_msghdr __user *) umsg;
2432 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2433 	} else {
2434 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2435 	}
2436 	if (err < 0)
2437 		return err;
2438 
2439 	return 0;
2440 }
2441 
2442 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2443 			 struct msghdr *msg_sys, unsigned int flags,
2444 			 struct used_address *used_address,
2445 			 unsigned int allowed_msghdr_flags)
2446 {
2447 	struct sockaddr_storage address;
2448 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2449 	ssize_t err;
2450 
2451 	msg_sys->msg_name = &address;
2452 
2453 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2454 	if (err < 0)
2455 		return err;
2456 
2457 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2458 				allowed_msghdr_flags);
2459 	kfree(iov);
2460 	return err;
2461 }
2462 
2463 /*
2464  *	BSD sendmsg interface
2465  */
2466 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2467 			unsigned int flags)
2468 {
2469 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2470 }
2471 
2472 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2473 		   bool forbid_cmsg_compat)
2474 {
2475 	int fput_needed, err;
2476 	struct msghdr msg_sys;
2477 	struct socket *sock;
2478 
2479 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2480 		return -EINVAL;
2481 
2482 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2483 	if (!sock)
2484 		goto out;
2485 
2486 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2487 
2488 	fput_light(sock->file, fput_needed);
2489 out:
2490 	return err;
2491 }
2492 
2493 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2494 {
2495 	return __sys_sendmsg(fd, msg, flags, true);
2496 }
2497 
2498 /*
2499  *	Linux sendmmsg interface
2500  */
2501 
2502 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2503 		   unsigned int flags, bool forbid_cmsg_compat)
2504 {
2505 	int fput_needed, err, datagrams;
2506 	struct socket *sock;
2507 	struct mmsghdr __user *entry;
2508 	struct compat_mmsghdr __user *compat_entry;
2509 	struct msghdr msg_sys;
2510 	struct used_address used_address;
2511 	unsigned int oflags = flags;
2512 
2513 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2514 		return -EINVAL;
2515 
2516 	if (vlen > UIO_MAXIOV)
2517 		vlen = UIO_MAXIOV;
2518 
2519 	datagrams = 0;
2520 
2521 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2522 	if (!sock)
2523 		return err;
2524 
2525 	used_address.name_len = UINT_MAX;
2526 	entry = mmsg;
2527 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2528 	err = 0;
2529 	flags |= MSG_BATCH;
2530 
2531 	while (datagrams < vlen) {
2532 		if (datagrams == vlen - 1)
2533 			flags = oflags;
2534 
2535 		if (MSG_CMSG_COMPAT & flags) {
2536 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2537 					     &msg_sys, flags, &used_address, MSG_EOR);
2538 			if (err < 0)
2539 				break;
2540 			err = __put_user(err, &compat_entry->msg_len);
2541 			++compat_entry;
2542 		} else {
2543 			err = ___sys_sendmsg(sock,
2544 					     (struct user_msghdr __user *)entry,
2545 					     &msg_sys, flags, &used_address, MSG_EOR);
2546 			if (err < 0)
2547 				break;
2548 			err = put_user(err, &entry->msg_len);
2549 			++entry;
2550 		}
2551 
2552 		if (err)
2553 			break;
2554 		++datagrams;
2555 		if (msg_data_left(&msg_sys))
2556 			break;
2557 		cond_resched();
2558 	}
2559 
2560 	fput_light(sock->file, fput_needed);
2561 
2562 	/* We only return an error if no datagrams were able to be sent */
2563 	if (datagrams != 0)
2564 		return datagrams;
2565 
2566 	return err;
2567 }
2568 
2569 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2570 		unsigned int, vlen, unsigned int, flags)
2571 {
2572 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2573 }
2574 
2575 int recvmsg_copy_msghdr(struct msghdr *msg,
2576 			struct user_msghdr __user *umsg, unsigned flags,
2577 			struct sockaddr __user **uaddr,
2578 			struct iovec **iov)
2579 {
2580 	ssize_t err;
2581 
2582 	if (MSG_CMSG_COMPAT & flags) {
2583 		struct compat_msghdr __user *msg_compat;
2584 
2585 		msg_compat = (struct compat_msghdr __user *) umsg;
2586 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2587 	} else {
2588 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2589 	}
2590 	if (err < 0)
2591 		return err;
2592 
2593 	return 0;
2594 }
2595 
2596 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2597 			   struct user_msghdr __user *msg,
2598 			   struct sockaddr __user *uaddr,
2599 			   unsigned int flags, int nosec)
2600 {
2601 	struct compat_msghdr __user *msg_compat =
2602 					(struct compat_msghdr __user *) msg;
2603 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2604 	struct sockaddr_storage addr;
2605 	unsigned long cmsg_ptr;
2606 	int len;
2607 	ssize_t err;
2608 
2609 	msg_sys->msg_name = &addr;
2610 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2611 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2612 
2613 	/* We assume all kernel code knows the size of sockaddr_storage */
2614 	msg_sys->msg_namelen = 0;
2615 
2616 	if (sock->file->f_flags & O_NONBLOCK)
2617 		flags |= MSG_DONTWAIT;
2618 
2619 	if (unlikely(nosec))
2620 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2621 	else
2622 		err = sock_recvmsg(sock, msg_sys, flags);
2623 
2624 	if (err < 0)
2625 		goto out;
2626 	len = err;
2627 
2628 	if (uaddr != NULL) {
2629 		err = move_addr_to_user(&addr,
2630 					msg_sys->msg_namelen, uaddr,
2631 					uaddr_len);
2632 		if (err < 0)
2633 			goto out;
2634 	}
2635 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2636 			 COMPAT_FLAGS(msg));
2637 	if (err)
2638 		goto out;
2639 	if (MSG_CMSG_COMPAT & flags)
2640 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2641 				 &msg_compat->msg_controllen);
2642 	else
2643 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2644 				 &msg->msg_controllen);
2645 	if (err)
2646 		goto out;
2647 	err = len;
2648 out:
2649 	return err;
2650 }
2651 
2652 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2653 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2654 {
2655 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2656 	/* user mode address pointers */
2657 	struct sockaddr __user *uaddr;
2658 	ssize_t err;
2659 
2660 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2661 	if (err < 0)
2662 		return err;
2663 
2664 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2665 	kfree(iov);
2666 	return err;
2667 }
2668 
2669 /*
2670  *	BSD recvmsg interface
2671  */
2672 
2673 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2674 			struct user_msghdr __user *umsg,
2675 			struct sockaddr __user *uaddr, unsigned int flags)
2676 {
2677 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2678 }
2679 
2680 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2681 		   bool forbid_cmsg_compat)
2682 {
2683 	int fput_needed, err;
2684 	struct msghdr msg_sys;
2685 	struct socket *sock;
2686 
2687 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2688 		return -EINVAL;
2689 
2690 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2691 	if (!sock)
2692 		goto out;
2693 
2694 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2695 
2696 	fput_light(sock->file, fput_needed);
2697 out:
2698 	return err;
2699 }
2700 
2701 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2702 		unsigned int, flags)
2703 {
2704 	return __sys_recvmsg(fd, msg, flags, true);
2705 }
2706 
2707 /*
2708  *     Linux recvmmsg interface
2709  */
2710 
2711 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2712 			  unsigned int vlen, unsigned int flags,
2713 			  struct timespec64 *timeout)
2714 {
2715 	int fput_needed, err, datagrams;
2716 	struct socket *sock;
2717 	struct mmsghdr __user *entry;
2718 	struct compat_mmsghdr __user *compat_entry;
2719 	struct msghdr msg_sys;
2720 	struct timespec64 end_time;
2721 	struct timespec64 timeout64;
2722 
2723 	if (timeout &&
2724 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2725 				    timeout->tv_nsec))
2726 		return -EINVAL;
2727 
2728 	datagrams = 0;
2729 
2730 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2731 	if (!sock)
2732 		return err;
2733 
2734 	if (likely(!(flags & MSG_ERRQUEUE))) {
2735 		err = sock_error(sock->sk);
2736 		if (err) {
2737 			datagrams = err;
2738 			goto out_put;
2739 		}
2740 	}
2741 
2742 	entry = mmsg;
2743 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2744 
2745 	while (datagrams < vlen) {
2746 		/*
2747 		 * No need to ask LSM for more than the first datagram.
2748 		 */
2749 		if (MSG_CMSG_COMPAT & flags) {
2750 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2751 					     &msg_sys, flags & ~MSG_WAITFORONE,
2752 					     datagrams);
2753 			if (err < 0)
2754 				break;
2755 			err = __put_user(err, &compat_entry->msg_len);
2756 			++compat_entry;
2757 		} else {
2758 			err = ___sys_recvmsg(sock,
2759 					     (struct user_msghdr __user *)entry,
2760 					     &msg_sys, flags & ~MSG_WAITFORONE,
2761 					     datagrams);
2762 			if (err < 0)
2763 				break;
2764 			err = put_user(err, &entry->msg_len);
2765 			++entry;
2766 		}
2767 
2768 		if (err)
2769 			break;
2770 		++datagrams;
2771 
2772 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2773 		if (flags & MSG_WAITFORONE)
2774 			flags |= MSG_DONTWAIT;
2775 
2776 		if (timeout) {
2777 			ktime_get_ts64(&timeout64);
2778 			*timeout = timespec64_sub(end_time, timeout64);
2779 			if (timeout->tv_sec < 0) {
2780 				timeout->tv_sec = timeout->tv_nsec = 0;
2781 				break;
2782 			}
2783 
2784 			/* Timeout, return less than vlen datagrams */
2785 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2786 				break;
2787 		}
2788 
2789 		/* Out of band data, return right away */
2790 		if (msg_sys.msg_flags & MSG_OOB)
2791 			break;
2792 		cond_resched();
2793 	}
2794 
2795 	if (err == 0)
2796 		goto out_put;
2797 
2798 	if (datagrams == 0) {
2799 		datagrams = err;
2800 		goto out_put;
2801 	}
2802 
2803 	/*
2804 	 * We may return less entries than requested (vlen) if the
2805 	 * sock is non block and there aren't enough datagrams...
2806 	 */
2807 	if (err != -EAGAIN) {
2808 		/*
2809 		 * ... or  if recvmsg returns an error after we
2810 		 * received some datagrams, where we record the
2811 		 * error to return on the next call or if the
2812 		 * app asks about it using getsockopt(SO_ERROR).
2813 		 */
2814 		sock->sk->sk_err = -err;
2815 	}
2816 out_put:
2817 	fput_light(sock->file, fput_needed);
2818 
2819 	return datagrams;
2820 }
2821 
2822 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2823 		   unsigned int vlen, unsigned int flags,
2824 		   struct __kernel_timespec __user *timeout,
2825 		   struct old_timespec32 __user *timeout32)
2826 {
2827 	int datagrams;
2828 	struct timespec64 timeout_sys;
2829 
2830 	if (timeout && get_timespec64(&timeout_sys, timeout))
2831 		return -EFAULT;
2832 
2833 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2834 		return -EFAULT;
2835 
2836 	if (!timeout && !timeout32)
2837 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2838 
2839 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2840 
2841 	if (datagrams <= 0)
2842 		return datagrams;
2843 
2844 	if (timeout && put_timespec64(&timeout_sys, timeout))
2845 		datagrams = -EFAULT;
2846 
2847 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2848 		datagrams = -EFAULT;
2849 
2850 	return datagrams;
2851 }
2852 
2853 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2854 		unsigned int, vlen, unsigned int, flags,
2855 		struct __kernel_timespec __user *, timeout)
2856 {
2857 	if (flags & MSG_CMSG_COMPAT)
2858 		return -EINVAL;
2859 
2860 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2861 }
2862 
2863 #ifdef CONFIG_COMPAT_32BIT_TIME
2864 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2865 		unsigned int, vlen, unsigned int, flags,
2866 		struct old_timespec32 __user *, timeout)
2867 {
2868 	if (flags & MSG_CMSG_COMPAT)
2869 		return -EINVAL;
2870 
2871 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2872 }
2873 #endif
2874 
2875 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2876 /* Argument list sizes for sys_socketcall */
2877 #define AL(x) ((x) * sizeof(unsigned long))
2878 static const unsigned char nargs[21] = {
2879 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2880 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2881 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2882 	AL(4), AL(5), AL(4)
2883 };
2884 
2885 #undef AL
2886 
2887 /*
2888  *	System call vectors.
2889  *
2890  *	Argument checking cleaned up. Saved 20% in size.
2891  *  This function doesn't need to set the kernel lock because
2892  *  it is set by the callees.
2893  */
2894 
2895 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2896 {
2897 	unsigned long a[AUDITSC_ARGS];
2898 	unsigned long a0, a1;
2899 	int err;
2900 	unsigned int len;
2901 
2902 	if (call < 1 || call > SYS_SENDMMSG)
2903 		return -EINVAL;
2904 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2905 
2906 	len = nargs[call];
2907 	if (len > sizeof(a))
2908 		return -EINVAL;
2909 
2910 	/* copy_from_user should be SMP safe. */
2911 	if (copy_from_user(a, args, len))
2912 		return -EFAULT;
2913 
2914 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2915 	if (err)
2916 		return err;
2917 
2918 	a0 = a[0];
2919 	a1 = a[1];
2920 
2921 	switch (call) {
2922 	case SYS_SOCKET:
2923 		err = __sys_socket(a0, a1, a[2]);
2924 		break;
2925 	case SYS_BIND:
2926 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2927 		break;
2928 	case SYS_CONNECT:
2929 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2930 		break;
2931 	case SYS_LISTEN:
2932 		err = __sys_listen(a0, a1);
2933 		break;
2934 	case SYS_ACCEPT:
2935 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2936 				    (int __user *)a[2], 0);
2937 		break;
2938 	case SYS_GETSOCKNAME:
2939 		err =
2940 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
2941 				      (int __user *)a[2]);
2942 		break;
2943 	case SYS_GETPEERNAME:
2944 		err =
2945 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
2946 				      (int __user *)a[2]);
2947 		break;
2948 	case SYS_SOCKETPAIR:
2949 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2950 		break;
2951 	case SYS_SEND:
2952 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2953 				   NULL, 0);
2954 		break;
2955 	case SYS_SENDTO:
2956 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2957 				   (struct sockaddr __user *)a[4], a[5]);
2958 		break;
2959 	case SYS_RECV:
2960 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2961 				     NULL, NULL);
2962 		break;
2963 	case SYS_RECVFROM:
2964 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2965 				     (struct sockaddr __user *)a[4],
2966 				     (int __user *)a[5]);
2967 		break;
2968 	case SYS_SHUTDOWN:
2969 		err = __sys_shutdown(a0, a1);
2970 		break;
2971 	case SYS_SETSOCKOPT:
2972 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2973 				       a[4]);
2974 		break;
2975 	case SYS_GETSOCKOPT:
2976 		err =
2977 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2978 				     (int __user *)a[4]);
2979 		break;
2980 	case SYS_SENDMSG:
2981 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2982 				    a[2], true);
2983 		break;
2984 	case SYS_SENDMMSG:
2985 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2986 				     a[3], true);
2987 		break;
2988 	case SYS_RECVMSG:
2989 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2990 				    a[2], true);
2991 		break;
2992 	case SYS_RECVMMSG:
2993 		if (IS_ENABLED(CONFIG_64BIT))
2994 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2995 					     a[2], a[3],
2996 					     (struct __kernel_timespec __user *)a[4],
2997 					     NULL);
2998 		else
2999 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3000 					     a[2], a[3], NULL,
3001 					     (struct old_timespec32 __user *)a[4]);
3002 		break;
3003 	case SYS_ACCEPT4:
3004 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3005 				    (int __user *)a[2], a[3]);
3006 		break;
3007 	default:
3008 		err = -EINVAL;
3009 		break;
3010 	}
3011 	return err;
3012 }
3013 
3014 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3015 
3016 /**
3017  *	sock_register - add a socket protocol handler
3018  *	@ops: description of protocol
3019  *
3020  *	This function is called by a protocol handler that wants to
3021  *	advertise its address family, and have it linked into the
3022  *	socket interface. The value ops->family corresponds to the
3023  *	socket system call protocol family.
3024  */
3025 int sock_register(const struct net_proto_family *ops)
3026 {
3027 	int err;
3028 
3029 	if (ops->family >= NPROTO) {
3030 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3031 		return -ENOBUFS;
3032 	}
3033 
3034 	spin_lock(&net_family_lock);
3035 	if (rcu_dereference_protected(net_families[ops->family],
3036 				      lockdep_is_held(&net_family_lock)))
3037 		err = -EEXIST;
3038 	else {
3039 		rcu_assign_pointer(net_families[ops->family], ops);
3040 		err = 0;
3041 	}
3042 	spin_unlock(&net_family_lock);
3043 
3044 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3045 	return err;
3046 }
3047 EXPORT_SYMBOL(sock_register);
3048 
3049 /**
3050  *	sock_unregister - remove a protocol handler
3051  *	@family: protocol family to remove
3052  *
3053  *	This function is called by a protocol handler that wants to
3054  *	remove its address family, and have it unlinked from the
3055  *	new socket creation.
3056  *
3057  *	If protocol handler is a module, then it can use module reference
3058  *	counts to protect against new references. If protocol handler is not
3059  *	a module then it needs to provide its own protection in
3060  *	the ops->create routine.
3061  */
3062 void sock_unregister(int family)
3063 {
3064 	BUG_ON(family < 0 || family >= NPROTO);
3065 
3066 	spin_lock(&net_family_lock);
3067 	RCU_INIT_POINTER(net_families[family], NULL);
3068 	spin_unlock(&net_family_lock);
3069 
3070 	synchronize_rcu();
3071 
3072 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3073 }
3074 EXPORT_SYMBOL(sock_unregister);
3075 
3076 bool sock_is_registered(int family)
3077 {
3078 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3079 }
3080 
3081 static int __init sock_init(void)
3082 {
3083 	int err;
3084 	/*
3085 	 *      Initialize the network sysctl infrastructure.
3086 	 */
3087 	err = net_sysctl_init();
3088 	if (err)
3089 		goto out;
3090 
3091 	/*
3092 	 *      Initialize skbuff SLAB cache
3093 	 */
3094 	skb_init();
3095 
3096 	/*
3097 	 *      Initialize the protocols module.
3098 	 */
3099 
3100 	init_inodecache();
3101 
3102 	err = register_filesystem(&sock_fs_type);
3103 	if (err)
3104 		goto out;
3105 	sock_mnt = kern_mount(&sock_fs_type);
3106 	if (IS_ERR(sock_mnt)) {
3107 		err = PTR_ERR(sock_mnt);
3108 		goto out_mount;
3109 	}
3110 
3111 	/* The real protocol initialization is performed in later initcalls.
3112 	 */
3113 
3114 #ifdef CONFIG_NETFILTER
3115 	err = netfilter_init();
3116 	if (err)
3117 		goto out;
3118 #endif
3119 
3120 	ptp_classifier_init();
3121 
3122 out:
3123 	return err;
3124 
3125 out_mount:
3126 	unregister_filesystem(&sock_fs_type);
3127 	goto out;
3128 }
3129 
3130 core_initcall(sock_init);	/* early initcall */
3131 
3132 #ifdef CONFIG_PROC_FS
3133 void socket_seq_show(struct seq_file *seq)
3134 {
3135 	seq_printf(seq, "sockets: used %d\n",
3136 		   sock_inuse_get(seq->private));
3137 }
3138 #endif				/* CONFIG_PROC_FS */
3139 
3140 /* Handle the fact that while struct ifreq has the same *layout* on
3141  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3142  * which are handled elsewhere, it still has different *size* due to
3143  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3144  * resulting in struct ifreq being 32 and 40 bytes respectively).
3145  * As a result, if the struct happens to be at the end of a page and
3146  * the next page isn't readable/writable, we get a fault. To prevent
3147  * that, copy back and forth to the full size.
3148  */
3149 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3150 {
3151 	if (in_compat_syscall()) {
3152 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3153 
3154 		memset(ifr, 0, sizeof(*ifr));
3155 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3156 			return -EFAULT;
3157 
3158 		if (ifrdata)
3159 			*ifrdata = compat_ptr(ifr32->ifr_data);
3160 
3161 		return 0;
3162 	}
3163 
3164 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3165 		return -EFAULT;
3166 
3167 	if (ifrdata)
3168 		*ifrdata = ifr->ifr_data;
3169 
3170 	return 0;
3171 }
3172 EXPORT_SYMBOL(get_user_ifreq);
3173 
3174 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3175 {
3176 	size_t size = sizeof(*ifr);
3177 
3178 	if (in_compat_syscall())
3179 		size = sizeof(struct compat_ifreq);
3180 
3181 	if (copy_to_user(arg, ifr, size))
3182 		return -EFAULT;
3183 
3184 	return 0;
3185 }
3186 EXPORT_SYMBOL(put_user_ifreq);
3187 
3188 #ifdef CONFIG_COMPAT
3189 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3190 {
3191 	compat_uptr_t uptr32;
3192 	struct ifreq ifr;
3193 	void __user *saved;
3194 	int err;
3195 
3196 	if (get_user_ifreq(&ifr, NULL, uifr32))
3197 		return -EFAULT;
3198 
3199 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3200 		return -EFAULT;
3201 
3202 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3203 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3204 
3205 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3206 	if (!err) {
3207 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3208 		if (put_user_ifreq(&ifr, uifr32))
3209 			err = -EFAULT;
3210 	}
3211 	return err;
3212 }
3213 
3214 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3215 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3216 				 struct compat_ifreq __user *u_ifreq32)
3217 {
3218 	struct ifreq ifreq;
3219 	void __user *data;
3220 
3221 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3222 		return -EFAULT;
3223 	ifreq.ifr_data = data;
3224 
3225 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3226 }
3227 
3228 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3229  * for some operations; this forces use of the newer bridge-utils that
3230  * use compatible ioctls
3231  */
3232 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3233 {
3234 	compat_ulong_t tmp;
3235 
3236 	if (get_user(tmp, argp))
3237 		return -EFAULT;
3238 	if (tmp == BRCTL_GET_VERSION)
3239 		return BRCTL_VERSION + 1;
3240 	return -EINVAL;
3241 }
3242 
3243 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3244 			 unsigned int cmd, unsigned long arg)
3245 {
3246 	void __user *argp = compat_ptr(arg);
3247 	struct sock *sk = sock->sk;
3248 	struct net *net = sock_net(sk);
3249 
3250 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3251 		return sock_ioctl(file, cmd, (unsigned long)argp);
3252 
3253 	switch (cmd) {
3254 	case SIOCSIFBR:
3255 	case SIOCGIFBR:
3256 		return old_bridge_ioctl(argp);
3257 	case SIOCWANDEV:
3258 		return compat_siocwandev(net, argp);
3259 	case SIOCGSTAMP_OLD:
3260 	case SIOCGSTAMPNS_OLD:
3261 		if (!sock->ops->gettstamp)
3262 			return -ENOIOCTLCMD;
3263 		return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3264 					    !COMPAT_USE_64BIT_TIME);
3265 
3266 	case SIOCETHTOOL:
3267 	case SIOCBONDSLAVEINFOQUERY:
3268 	case SIOCBONDINFOQUERY:
3269 	case SIOCSHWTSTAMP:
3270 	case SIOCGHWTSTAMP:
3271 		return compat_ifr_data_ioctl(net, cmd, argp);
3272 
3273 	case FIOSETOWN:
3274 	case SIOCSPGRP:
3275 	case FIOGETOWN:
3276 	case SIOCGPGRP:
3277 	case SIOCBRADDBR:
3278 	case SIOCBRDELBR:
3279 	case SIOCGIFVLAN:
3280 	case SIOCSIFVLAN:
3281 	case SIOCGSKNS:
3282 	case SIOCGSTAMP_NEW:
3283 	case SIOCGSTAMPNS_NEW:
3284 	case SIOCGIFCONF:
3285 		return sock_ioctl(file, cmd, arg);
3286 
3287 	case SIOCGIFFLAGS:
3288 	case SIOCSIFFLAGS:
3289 	case SIOCGIFMAP:
3290 	case SIOCSIFMAP:
3291 	case SIOCGIFMETRIC:
3292 	case SIOCSIFMETRIC:
3293 	case SIOCGIFMTU:
3294 	case SIOCSIFMTU:
3295 	case SIOCGIFMEM:
3296 	case SIOCSIFMEM:
3297 	case SIOCGIFHWADDR:
3298 	case SIOCSIFHWADDR:
3299 	case SIOCADDMULTI:
3300 	case SIOCDELMULTI:
3301 	case SIOCGIFINDEX:
3302 	case SIOCGIFADDR:
3303 	case SIOCSIFADDR:
3304 	case SIOCSIFHWBROADCAST:
3305 	case SIOCDIFADDR:
3306 	case SIOCGIFBRDADDR:
3307 	case SIOCSIFBRDADDR:
3308 	case SIOCGIFDSTADDR:
3309 	case SIOCSIFDSTADDR:
3310 	case SIOCGIFNETMASK:
3311 	case SIOCSIFNETMASK:
3312 	case SIOCSIFPFLAGS:
3313 	case SIOCGIFPFLAGS:
3314 	case SIOCGIFTXQLEN:
3315 	case SIOCSIFTXQLEN:
3316 	case SIOCBRADDIF:
3317 	case SIOCBRDELIF:
3318 	case SIOCGIFNAME:
3319 	case SIOCSIFNAME:
3320 	case SIOCGMIIPHY:
3321 	case SIOCGMIIREG:
3322 	case SIOCSMIIREG:
3323 	case SIOCBONDENSLAVE:
3324 	case SIOCBONDRELEASE:
3325 	case SIOCBONDSETHWADDR:
3326 	case SIOCBONDCHANGEACTIVE:
3327 	case SIOCSARP:
3328 	case SIOCGARP:
3329 	case SIOCDARP:
3330 	case SIOCOUTQ:
3331 	case SIOCOUTQNSD:
3332 	case SIOCATMARK:
3333 		return sock_do_ioctl(net, sock, cmd, arg);
3334 	}
3335 
3336 	return -ENOIOCTLCMD;
3337 }
3338 
3339 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3340 			      unsigned long arg)
3341 {
3342 	struct socket *sock = file->private_data;
3343 	int ret = -ENOIOCTLCMD;
3344 	struct sock *sk;
3345 	struct net *net;
3346 
3347 	sk = sock->sk;
3348 	net = sock_net(sk);
3349 
3350 	if (sock->ops->compat_ioctl)
3351 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3352 
3353 	if (ret == -ENOIOCTLCMD &&
3354 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3355 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3356 
3357 	if (ret == -ENOIOCTLCMD)
3358 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3359 
3360 	return ret;
3361 }
3362 #endif
3363 
3364 /**
3365  *	kernel_bind - bind an address to a socket (kernel space)
3366  *	@sock: socket
3367  *	@addr: address
3368  *	@addrlen: length of address
3369  *
3370  *	Returns 0 or an error.
3371  */
3372 
3373 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3374 {
3375 	return sock->ops->bind(sock, addr, addrlen);
3376 }
3377 EXPORT_SYMBOL(kernel_bind);
3378 
3379 /**
3380  *	kernel_listen - move socket to listening state (kernel space)
3381  *	@sock: socket
3382  *	@backlog: pending connections queue size
3383  *
3384  *	Returns 0 or an error.
3385  */
3386 
3387 int kernel_listen(struct socket *sock, int backlog)
3388 {
3389 	return sock->ops->listen(sock, backlog);
3390 }
3391 EXPORT_SYMBOL(kernel_listen);
3392 
3393 /**
3394  *	kernel_accept - accept a connection (kernel space)
3395  *	@sock: listening socket
3396  *	@newsock: new connected socket
3397  *	@flags: flags
3398  *
3399  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3400  *	If it fails, @newsock is guaranteed to be %NULL.
3401  *	Returns 0 or an error.
3402  */
3403 
3404 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3405 {
3406 	struct sock *sk = sock->sk;
3407 	int err;
3408 
3409 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3410 			       newsock);
3411 	if (err < 0)
3412 		goto done;
3413 
3414 	err = sock->ops->accept(sock, *newsock, flags, true);
3415 	if (err < 0) {
3416 		sock_release(*newsock);
3417 		*newsock = NULL;
3418 		goto done;
3419 	}
3420 
3421 	(*newsock)->ops = sock->ops;
3422 	__module_get((*newsock)->ops->owner);
3423 
3424 done:
3425 	return err;
3426 }
3427 EXPORT_SYMBOL(kernel_accept);
3428 
3429 /**
3430  *	kernel_connect - connect a socket (kernel space)
3431  *	@sock: socket
3432  *	@addr: address
3433  *	@addrlen: address length
3434  *	@flags: flags (O_NONBLOCK, ...)
3435  *
3436  *	For datagram sockets, @addr is the address to which datagrams are sent
3437  *	by default, and the only address from which datagrams are received.
3438  *	For stream sockets, attempts to connect to @addr.
3439  *	Returns 0 or an error code.
3440  */
3441 
3442 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3443 		   int flags)
3444 {
3445 	return sock->ops->connect(sock, addr, addrlen, flags);
3446 }
3447 EXPORT_SYMBOL(kernel_connect);
3448 
3449 /**
3450  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3451  *	@sock: socket
3452  *	@addr: address holder
3453  *
3454  * 	Fills the @addr pointer with the address which the socket is bound.
3455  *	Returns 0 or an error code.
3456  */
3457 
3458 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3459 {
3460 	return sock->ops->getname(sock, addr, 0);
3461 }
3462 EXPORT_SYMBOL(kernel_getsockname);
3463 
3464 /**
3465  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3466  *	@sock: socket
3467  *	@addr: address holder
3468  *
3469  * 	Fills the @addr pointer with the address which the socket is connected.
3470  *	Returns 0 or an error code.
3471  */
3472 
3473 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3474 {
3475 	return sock->ops->getname(sock, addr, 1);
3476 }
3477 EXPORT_SYMBOL(kernel_getpeername);
3478 
3479 /**
3480  *	kernel_sendpage - send a &page through a socket (kernel space)
3481  *	@sock: socket
3482  *	@page: page
3483  *	@offset: page offset
3484  *	@size: total size in bytes
3485  *	@flags: flags (MSG_DONTWAIT, ...)
3486  *
3487  *	Returns the total amount sent in bytes or an error.
3488  */
3489 
3490 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3491 		    size_t size, int flags)
3492 {
3493 	if (sock->ops->sendpage) {
3494 		/* Warn in case the improper page to zero-copy send */
3495 		WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3496 		return sock->ops->sendpage(sock, page, offset, size, flags);
3497 	}
3498 	return sock_no_sendpage(sock, page, offset, size, flags);
3499 }
3500 EXPORT_SYMBOL(kernel_sendpage);
3501 
3502 /**
3503  *	kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3504  *	@sk: sock
3505  *	@page: page
3506  *	@offset: page offset
3507  *	@size: total size in bytes
3508  *	@flags: flags (MSG_DONTWAIT, ...)
3509  *
3510  *	Returns the total amount sent in bytes or an error.
3511  *	Caller must hold @sk.
3512  */
3513 
3514 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3515 			   size_t size, int flags)
3516 {
3517 	struct socket *sock = sk->sk_socket;
3518 
3519 	if (sock->ops->sendpage_locked)
3520 		return sock->ops->sendpage_locked(sk, page, offset, size,
3521 						  flags);
3522 
3523 	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3524 }
3525 EXPORT_SYMBOL(kernel_sendpage_locked);
3526 
3527 /**
3528  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3529  *	@sock: socket
3530  *	@how: connection part
3531  *
3532  *	Returns 0 or an error.
3533  */
3534 
3535 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3536 {
3537 	return sock->ops->shutdown(sock, how);
3538 }
3539 EXPORT_SYMBOL(kernel_sock_shutdown);
3540 
3541 /**
3542  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3543  *	@sk: socket
3544  *
3545  *	This routine returns the IP overhead imposed by a socket i.e.
3546  *	the length of the underlying IP header, depending on whether
3547  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3548  *	on at the socket. Assumes that the caller has a lock on the socket.
3549  */
3550 
3551 u32 kernel_sock_ip_overhead(struct sock *sk)
3552 {
3553 	struct inet_sock *inet;
3554 	struct ip_options_rcu *opt;
3555 	u32 overhead = 0;
3556 #if IS_ENABLED(CONFIG_IPV6)
3557 	struct ipv6_pinfo *np;
3558 	struct ipv6_txoptions *optv6 = NULL;
3559 #endif /* IS_ENABLED(CONFIG_IPV6) */
3560 
3561 	if (!sk)
3562 		return overhead;
3563 
3564 	switch (sk->sk_family) {
3565 	case AF_INET:
3566 		inet = inet_sk(sk);
3567 		overhead += sizeof(struct iphdr);
3568 		opt = rcu_dereference_protected(inet->inet_opt,
3569 						sock_owned_by_user(sk));
3570 		if (opt)
3571 			overhead += opt->opt.optlen;
3572 		return overhead;
3573 #if IS_ENABLED(CONFIG_IPV6)
3574 	case AF_INET6:
3575 		np = inet6_sk(sk);
3576 		overhead += sizeof(struct ipv6hdr);
3577 		if (np)
3578 			optv6 = rcu_dereference_protected(np->opt,
3579 							  sock_owned_by_user(sk));
3580 		if (optv6)
3581 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3582 		return overhead;
3583 #endif /* IS_ENABLED(CONFIG_IPV6) */
3584 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3585 		return overhead;
3586 	}
3587 }
3588 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3589